Who are the editors of the book?
What is the edition of the book?
Third Edition
What series is this book part of?
Prentice Hall Series in Artificial Intelligence
Who are some contributing writers?
Where was the book published?
Upper Saddle River, New Jersey, USA
Who is the Vice President and Editorial Director?
Marcia J. Horton
Who is the Editor-in-Chief?
Michael Hirsch
What is the ISBN-13 of the book?
978-0-13-604259-4
What is the ISBN-10 of the book?
0-13-604259-7
What does the book aim to cover?
The breadth of AI, including logic, probability, perception, reasoning, learning, and action.
What is the subtitle of the book?
A Modern Approach
What is the copyright year of the book?
2010, 2003, 1995
What is the purpose of the book?
To synthesize current knowledge into a common framework of AI.
What year was the last edition of the book published?
2003
What are some important applications of AI technology?
What significant algorithmic landmark was achieved in AI?
Solution of the game of checkers
What theoretical progress areas are mentioned in AI?
What new emphasis is placed in the book regarding environments?
Partially observable and nondeterministic environments
What concepts are introduced in the context of belief states?
What types of representations are discussed for agents?
What is covered in more depth regarding planning?
Contingent planning in partially observable environments
What new material was added on probabilistic models?
First-order probabilistic models and open-universe models
What chapter was completely rewritten in the book?
The introductory machine-learning chapter
What percentage of citations are from works published after 2003?
20%
What percentage of the material in the book is brand new?
20%
What is the main unifying theme of the book?
The idea of an intelligent agent
How is AI defined in the book?
Study of agents that receive percepts and perform actions
What does an agent implement in AI?
A function that maps percept sequences to actions
What is the role of learning in AI according to the book?
Extending the reach of the designer into unknown environments
How many chapters does the book contain?
27 chapters
What is the intended use of the book?
Undergraduate course or course sequence
What is the website for the book?
aima.cs.berkeley.edu
What is the prerequisite knowledge for the book?
Familiarity with basic computer science concepts at a sophomore level
What mathematical topics are useful for the book?
Freshman calculus and linear algebra
Where can you find exercises in the book?
At the end of each chapter
How are programming exercises indicated?
Marked with a keyboard icon
What is the purpose of the extensive index?
To make it easy to find things in the book
What does the cover of the book depict?
Final position from game 6 of the 1997 match between Garry Kasparov and DEEP BLUE
Who was forced to resign in the chess match?
Garry Kasparov
What does the cover feature besides Kasparov?
Asimo robot, Thomas Bayes, Mars Rover, Alan Turing, Shakey robot, Aristotle
Who contributed to Chapter 24 of the book?
Jitendra Malik and David Forsyth
What is the focus of Chapter 25?
Robotics
What is the role of Aristotle in the book's cover?
Pioneered the study of logic
What is depicted behind the title of the book?
A portion of the CPSC Bayesian network for medical diagnosis
What is the purpose of the book's website?
Contains algorithms, course materials, links, and discussion group info
What is marked in the margin when a new term is defined?
The new term itself
Who wrote Chapter 24 (computer vision)?
Jitendra Malik and David Forsyth
Who wrote Chapter 25 (robotics)?
Sebastian Thrun
Who wrote part of Chapter 22 (natural language)?
Vibhu Mittal
Who reviewed the manuscript?
Who formatted and improved the diagrams in this edition?
Jon Barron
Who helped with diagrams and algorithms in previous editions?
Who wrote and maintains the Java code examples on the website?
Who wrote the robotics chapter for the first edition?
John Canny
Who researched the historical notes?
Douglas Edwards
Who provided extensive improvements to every chapter?
Julie Sussman, P.P.A.
Who are Stuart's family members mentioned?
Who are Peter's family members mentioned?
Which libraries and developers are thanked for their contributions?
Who provided helpful comments on the book?
Who is Stuart Russell?
What awards has Stuart Russell received?
What are some works by Stuart Russell?
Who is Peter Norvig?
What was Peter Norvig's previous position?
What is the significance of the Anonymous Reviewer?
What is the Center for Intelligent Systems?
What did Stuart Russell achieve in 1998?
What is the Smith–Zadeh Chair?
How many papers has Stuart Russell published?
Who was the head of the Computational Sciences Division at NASA Ames Research Center?
He oversaw NASA’s research in artificial intelligence and robotics.
What is one of the first Internet information extraction services?
Developed by Junglee, where he was chief scientist.
What degrees did he receive from Brown University and UC Berkeley?
B.S. in applied mathematics and Ph.D. in computer science.
Which awards did he receive from UC Berkeley?
Distinguished Alumni and Engineering Innovation awards.
What medal did he receive from NASA?
Exceptional Achievement Medal.
Where has he been a professor?
University of Southern California.
What are the titles of his other books?
What is the first topic covered in the book?
Introduction to Artificial Intelligence.
What does 1.1 cover in the book?
What Is AI?
What does 1.2 discuss?
The Foundations of Artificial Intelligence.
What is covered in 1.3?
The History of Artificial Intelligence.
What is the focus of 1.4?
The State of the Art in AI.
What does 2.1 introduce?
Agents and Environments.
What concept is discussed in 2.2?
Good Behavior: The Concept of Rationality.
What is explored in 3.1?
Problem-Solving Agents.
What type of strategies are discussed in 3.4?
Uninformed Search Strategies.
What does 5.1 examine?
Games in AI.
What is explained in 6.1?
Defining Constraint Satisfaction Problems.
What is the definition of Constraint Satisfaction Problems?
Section 6.1: Defining Constraint Satisfaction Problems
What is constraint propagation in CSPs?
Section 6.2: Constraint Propagation: Inference in CSPs
What is the backtracking search method for CSPs?
Section 6.3: Backtracking Search for CSPs
What is the local search method for CSPs?
Section 6.4: Local Search for CSPs
What is discussed in the section on the structure of problems?
Section 6.5: The Structure of Problems
What does the summary section cover in CSPs?
Section 6.6: Summary, Bibliographical and Historical Notes, Exercises
What are logical agents?
Section 7: Logical Agents
What is the Wumpus World?
Section 7.2: The Wumpus World
What is the focus of the section on logic?
Section 7.3: Logic
What is propositional logic?
Section 7.4: Propositional Logic: A Very Simple Logic
What is propositional theorem proving?
Section 7.5: Propositional Theorem Proving
What is effective propositional model checking?
Section 7.6: Effective Propositional Model Checking
What are agents based on propositional logic?
Section 7.7: Agents Based on Propositional Logic
What is covered in the summary section for logical agents?
Section 7.8: Summary, Bibliographical and Historical Notes, Exercises
What is first-order logic?
Section 8: First-Order Logic
What does representation revisited refer to?
Section 8.1: Representation Revisited
What are the syntax and semantics of first-order logic?
Section 8.2: Syntax and Semantics of First-Order Logic
What is discussed in the section on using first-order logic?
Section 8.3: Using First-Order Logic
What is knowledge engineering in first-order logic?
Section 8.4: Knowledge Engineering in First-Order Logic
What does the summary section cover in first-order logic?
Section 8.5: Summary, Bibliographical and Historical Notes, Exercises
What is the difference between propositional and first-order inference?
Section 9.1: Propositional vs. First-Order Inference
What is unification and lifting?
Section 9.2: Unification and Lifting
What is forward chaining?
Section 9.3: Forward Chaining
What is backward chaining?
Section 9.4: Backward Chaining
What is resolution in inference?
Section 9.5: Resolution
What does the summary section cover in inference?
Section 9.6: Summary, Bibliographical and Historical Notes, Exercises
What is the definition of classical planning?
Section 10.1: Definition of Classical Planning
What are algorithms for planning as state-space search?
Section 10.2: Algorithms for Planning as State-Space Search
What are planning graphs?
Section 10.3: Planning Graphs
What are other classical planning approaches?
Section 10.4: Other Classical Planning Approaches
What is discussed in the analysis of planning approaches?
Section 10.5: Analysis of Planning Approaches
What does the summary section cover in classical planning?
Section 10.6: Summary, Bibliographical and Historical Notes, Exercises
What is discussed regarding planning and acting in the real world?
Section 11: Planning and Acting in the Real World
What are time, schedules, and resources in planning?
Section 11.1: Time, Schedules, and Resources
What is hierarchical planning?
Section 11.2: Hierarchical Planning
What is planning in nondeterministic domains?
Section 11.3: Planning and Acting in Nondeterministic Domains
What is multiagent planning?
Section 11.4: Multiagent Planning
What does the summary section cover in planning and acting?
Section 11.5: Summary, Bibliographical and Historical Notes, Exercises
What is knowledge representation?
Section 12: Knowledge Representation
What is ontological engineering?
Section 12.1: Ontological Engineering
What are categories and objects in knowledge representation?
Section 12.2: Categories and Objects
What are events in knowledge representation?
Section 12.3: Events
What are mental events and mental objects?
Section 12.4: Mental Events and Mental Objects
What are reasoning systems for categories?
Section 12.5: Reasoning Systems for Categories
What is reasoning with default information?
Section 12.6: Reasoning with Default Information
What is discussed in the Internet Shopping World?
Section 12.7: The Internet Shopping World
What does the summary section cover in knowledge representation?
Section 12.8: Summary, Bi
What is discussed in section 12.6?
Reasoning with Default Information
What is covered in section 12.7?
The Internet Shopping World
What does section 12.8 include?
Summary, Bibliographical and Historical Notes, Exercises
What is the title of section 13?
Quantifying Uncertainty
What is discussed in section 13.1?
Acting under Uncertainty
What is the focus of section 13.2?
Basic Probability Notation
What does section 13.3 cover?
Inference Using Full Joint Distributions
What is discussed in section 13.4?
Independence
What is the topic of section 13.5?
Bayes’ Rule and Its Use
What is revisited in section 13.6?
The Wumpus World
What does section 13.7 include?
Summary, Bibliographical and Historical Notes, Exercises
What is the title of section 14?
Probabilistic Reasoning
What is the focus of section 14.1?
Representing Knowledge in an Uncertain Domain
What is discussed in section 14.2?
The Semantics of Bayesian Networks
What does section 14.3 cover?
Efficient Representation of Conditional Distributions
What is the topic of section 14.4?
Exact Inference in Bayesian Networks
What is discussed in section 14.5?
Approximate Inference in Bayesian Networks
What is covered in section 14.6?
Relational and First-Order Probability Models
What does section 14.7 discuss?
Other Approaches to Uncertain Reasoning
What is included in section 14.8?
Summary, Bibliographical and Historical Notes, Exercises
What is the title of section 15?
Probabilistic Reasoning over Time
What does section 15.1 cover?
Time and Uncertainty
What is discussed in section 15.2?
Inference in Temporal Models
What is covered in section 15.3?
Hidden Markov Models
What is the focus of section 15.4?
Kalman Filters
What does section 15.5 discuss?
Dynamic Bayesian Networks
What is included in section 15.6?
Keeping Track of Many Objects
What does section 15.7 include?
Summary, Bibliographical and Historical Notes, Exercises
What is the title of section 16?
Making Simple Decisions
What is discussed in section 16.1?
Combining Beliefs and Desires under Uncertainty
What does section 16.2 cover?
The Basis of Utility Theory
What is the focus of section 16.3?
Utility Functions
What is discussed in section 16.4?
Multiattribute Utility Functions
What does section 16.5 cover?
Decision Networks
What is the topic of section 16.6?
The Value of Information
What is included in section 16.7?
Decision-Theoretic Expert Systems
What does section 16.8 include?
Summary, Bibliographical and Historical Notes, Exercises
What is the title of section 17?
Making Complex Decisions
What is discussed in section 17.1?
Sequential Decision Problems
What does section 17.2 cover?
Value Iteration
What is the focus of section 17.3?
Policy Iteration
What is discussed in section 17.4?
Partially Observable MDPs
What does section 17.5 cover?
Decisions with Multiple Agents: Game Theory
What is the topic of section 17.6?
Mechanism Design
What does section 17.7 include?
Summary, Bibliographical and Historical Notes, Exercises
What is the title of section 18?
Learning from Examples
What is discussed in section 18.1?
Forms of Learning
What does section 18.2 cover?
Supervised Learning
What is the focus of section 18.3?
Learning Decision Trees
What is the section 18.2 about?
Supervised Learning
What does section 18.3 cover?
Learning Decision Trees
What is discussed in section 18.4?
Evaluating and Choosing the Best Hypothesis
What is the focus of section 18.5?
The Theory of Learning
What does section 18.6 cover?
Regression and Classification with Linear Models
What is the topic of section 18.7?
Artificial Neural Networks
What does section 18.8 discuss?
Nonparametric Models
What is covered in section 18.9?
Support Vector Machines
What does section 18.10 focus on?
Ensemble Learning
What is discussed in section 18.11?
Practical Machine Learning
What does section 18.12 summarize?
Summary, Bibliographical and Historical Notes, Exercises
What is the topic of section 19.1?
A Logical Formulation of Learning
What does section 19.2 cover?
Knowledge in Learning
What is discussed in section 19.3?
Explanation-Based Learning
What does section 19.4 focus on?
Learning Using Relevance Information
What is the topic of section 19.5?
Inductive Logic Programming
What does section 19.6 summarize?
Summary, Bibliographical and Historical Notes, Exercises
What is covered in section 20.1?
Statistical Learning
What does section 20.2 discuss?
Learning with Complete Data
What is the focus of section 20.3?
Learning with Hidden Variables: The EM Algorithm
What is summarized in section 20.4?
Summary, Bibliographical and Historical Notes, Exercises
What does section 21.1 introduce?
Introduction to Reinforcement Learning
What is discussed in section 21.2?
Passive Reinforcement Learning
What does section 21.3 cover?
Active Reinforcement Learning
What is the focus of section 21.4?
Generalization in Reinforcement Learning
What is covered in section 21.5?
Policy Search
What does section 21.6 discuss?
Applications of Reinforcement Learning
What is summarized in section 21.7?
Summary, Bibliographical and Historical Notes, Exercises
What is the topic of section 22.1?
Language Models
What does section 22.2 cover?
Text Classification
What is discussed in section 22.3?
Information Retrieval
What does section 22.4 focus on?
Information Extraction
What is summarized in section 22.5?
Summary, Bibliographical and Historical Notes, Exercises
What is the topic of section 23.1?
Phrase Structure Grammars
What does section 23.2 cover?
Syntactic Analysis (Parsing)
What is discussed in section 23.3?
Augmented Grammars and Semantic Interpretation
What does section 23.4 focus on?
Machine Translation
What is the topic of section 23.5?
Speech Recognition
What is summarized in section 23.6?
Summary, Bibliographical and Historical Notes, Exercises
What does section 24.1 cover?
Image Formation
What is discussed in section 24.2?
Early Image-Processing Operations
What does section 24.3 focus on?
Object Recognition by Appearance
What is the topic of section 24.4?
Reconstructing the 3D World
What is the focus of Object Recognition by Appearance?
What does Reconstructing the 3D World involve?
What is Object Recognition from Structural Information about?
What is the topic of Using Vision?
What is the introduction to Robotics about?
What does Robot Hardware cover?
What is the focus of Robotic Perception?
What does Planning to Move involve?
What is the topic of Planning Uncertain Movements?
What does Moving refer to in robotics?
What are Robotic Software Architectures?
What are the Application Domains of robotics?
What does the Conclusions section summarize?
What is discussed in Weak AI: Can Machines Act Intelligently??
What does Strong AI: Can Machines Really Think? explore?
What are the Ethics and Risks of Developing Artificial Intelligence?
What does AI: The Present and Future discuss?
What are Agent Components in AI?
What do Agent Architectures refer to?
What question does Are We Going in the Right Direction? address?
What if AI Does Succeed?
What does the Mathematical background section include?
What is covered in Complexity Analysis and O() Notation?
What are Vectors, Matrices, and Linear Algebra used for?
What do Probability Distributions explain?
What is Defining Languages with Backus–Naur Form (BNF)?
What does Describing Algorithms with Pseudocode involve?
What is the purpose of Online Help in programming?
What does AI encompass?
A huge variety of subfields including: - Learning - Perception - Playing chess - Proving mathematical theorems - Writing poetry - Driving a car - Diagnosing diseases
What are the two dimensions of AI definitions?
What is rationality in AI?
A system is rational if it does the 'right thing' given what it knows.
What is the Turing Test?
A test to determine if a computer can mimic human responses so well that a human interrogator cannot tell the difference.
What capabilities must a computer have to pass the Turing Test?
Who proposed the Turing Test?
Alan Turing in 1950.
What is natural language processing?
The ability to communicate successfully in English.
What is knowledge representation?
Storing what a computer knows or hears.
What is automated reasoning?
Using stored information to answer questions and draw new conclusions.
What is machine learning?
The ability to adapt to new circumstances and detect patterns.
What are the four approaches to AI?
What is the focus of a human-centered approach to AI?
An empirical science involving observations and hypotheses about human behavior.
What does the term 'computational intelligence' refer to?
The study of the design of intelligent agents.
What is the goal of AI according to Bellman (1978)?
The automation of activities associated with human thinking, such as decision-making and problem-solving.
What did Kurzweil (1990) define AI as?
The art of creating machines that perform functions requiring intelligence when performed by people.
What is the first point made about human imperfection?
Not all chess players are grandmasters; not everyone gets an A on the exam.
Who cataloged systematic errors in human reasoning?
Kahneman et al. (1982)
What does the total Turing Test include?
A video signal to test perceptual abilities and the ability to pass physical objects through a hatch.
What two capabilities are required to pass the total Turing Test?
Why do AI researchers focus on studying intelligence rather than passing the Turing Test?
They believe it is more important to study the underlying principles of intelligence.
What analogy is used to explain the quest for artificial flight?
The Wright brothers succeeded when they stopped imitating birds and focused on aerodynamics.
What are three ways to determine how humans think?
What is the goal of cognitive science?
To construct precise and testable theories of the human mind using AI models and psychology techniques.
Who developed the General Problem Solver (GPS)?
Allen Newell and Herbert Simon
What distinguishes modern authors in the field of AI and cognitive science?
They separate claims of algorithm performance from models of human performance.
Who was one of the first to codify 'right thinking'?
The Greek philosopher Aristotle.
Who was one of the first to codify right thinking?
Aristotle
What are syllogisms?
Patterns for argument structures that yield correct conclusions from correct premises.
What initiated the field called logic?
The study of the laws of thought.
What notation did 19th-century logicians develop?
Precise notation for statements about objects and relations.
What does the logicist tradition in AI aim to do?
Build intelligent systems based on logical programs.
What is a major obstacle in the logicist approach to AI?
Translating informal knowledge into formal logical terms.
How does computational reasoning relate to the logicist tradition?
It faces obstacles like computational resource exhaustion.
What is an agent in AI?
Something that acts, expected to operate autonomously and adapt.
What defines a rational agent?
Acts to achieve the best outcome or expected outcome under uncertainty.
What is one way a rational agent can act?
By reasoning logically to achieve goals.
What is an example of a non-inferential rational action?
Recoiling from a hot stove.
What skills are needed for the Turing Test?
Knowledge representation, reasoning, and natural language generation.
What advantage does the rational-agent approach have over the laws of thought?
It is more general and amenable to scientific development.
What is the standard of rationality in the rational-agent approach?
Mathematically well defined and completely general.
What is the standard of rationality?
It is mathematically well defined and can be 'unpacked' to generate agent designs that provably achieve it.
How is human behavior defined?
By the sum total of all the things that humans do, adapted for a specific environment.
What is the main focus of the book?
General principles of rational agents and components for constructing them.
What is the issue with achieving perfect rationality?
It is not feasible in complicated environments due to high computational demands.
What is the working hypothesis of the book?
Perfect rationality is a good starting point for analysis.
What chapters deal with limited rationality?
Chapters 5 and 17.
What disciplines contributed to AI?
Philosophy, among others.
What did Aristotle contribute to rationality?
He formulated laws governing the rational part of the mind and developed syllogisms for reasoning.
What was Ramon Lull's idea regarding reasoning?
That useful reasoning could be carried out by a mechanical artifact.
What did Thomas Hobbes propose about reasoning?
That reasoning is like numerical computation, involving silent addition and subtraction.
Who designed the first known calculating machine?
Wilhelm Schickard around 1623.
What is the Pascaline?
A calculating machine built by Blaise Pascal in 1642, known for addition and subtraction.
What did Gottfried Wilhelm Leibniz build?
A calculator that could add, subtract, multiply, and take roots.
What concept did Thomas Hobbes suggest in his book Leviathan?
The idea of an 'artificial animal'.
What is the distinction discussed by René Descartes?
Distinction between mind and matter.
What problem arises from a purely physical conception of the mind?
It leaves little room for free will.
What philosophy advocates the power of reasoning?
Rationalism.
What is dualism according to Descartes?
A belief in a part of the mind that is outside of nature.
What is materialism?
The belief that the brain's operation constitutes the mind.
What is the essence of empiricism?
Knowledge arises from sensory experience.
Who proposed the principle of induction?
David Hume.
What does logical positivism combine?
It combines rationalism and empiricism.
Who led the Vienna Circle?
Rudolf Carnap.
What is the confirmation theory?
It analyzes the acquisition of knowledge from experience.
What is a vital question in AI regarding knowledge?
The connection between knowledge and action.
How did Aristotle justify actions?
By a logical connection between goals and knowledge of outcomes.
What is the outcome of the action discussed in the text?
The outcome is an action that results from premises, e.g., 'I need a cloak. I have to make a cloak.'
What does Aristotle suggest in Nicomachean Ethics regarding deliberation?
We deliberate not about ends, but about means.
What do doctors and orators assume in their deliberations?
They assume the end and consider how to attain it.
What happens if an impossibility is encountered during deliberation?
We give up the search.
Who implemented Aristotle's algorithm 2300 years later?
Newell and Simon in their GPS program.
What does goal-based analysis not address?
It does not say what to do when several actions will achieve the goal or when no action will achieve it completely.
Who described a quantitative formula for action decisions?
Antoine Arnauld (1612–1694).
What idea did John Stuart Mill promote in his book Utilitarianism?
Rational decision criteria in all spheres of human activity.
What are the three fundamental areas required for mathematical formalization?
Logic, computation, and probability.
Who worked on Boolean logic and when?
George Boole (1815–1864).
What did Gottlob Frege extend in 1879?
He extended Boole’s logic to include objects and relations, creating first-order logic.
Who introduced a theory of reference in logic?
Alfred Tarski (1902–1983).
What is considered the first nontrivial algorithm?
Euclid’s algorithm for computing greatest common divisors.
What does the term algorithm derive from?
It comes from al-Khowarazmi, a Persian mathematician of the 9th century.
What did Kurt Gödel show in 1930 regarding first-order logic?
There exists an effective procedure to prove any true statement, but it cannot capture mathematical induction.
What did Gödel's incompleteness theorem show?
In any formal theory as strong as Peano arithmetic, there are true statements that are undecidable, meaning they have no proof within the theory.
What does the Church–Turing thesis state?
The Turing machine is capable of computing any computable function.
What is an intractable problem?
A problem where the time required to solve instances grows exponentially with the size of the instances.
Who pioneered the theory of NP-completeness?
Steven Cook (1971) and Richard Karp (1972).
What is the significance of polynomial vs exponential growth?
Exponential growth means even moderately large instances cannot be solved in reasonable time.
Who first framed the idea of probability?
Gerolamo Cardano (1501–1576).
What did Blaise Pascal show in 1654?
How to predict the future of an unfinished gambling game and assign average payoffs to gamblers.
What is the role of probability in quantitative sciences?
Helps deal with uncertain measurements and incomplete theories.
Who were the early contributors to probability theory?
What did Thomas Bayes propose?
A rule for updating probabilities in light of new evidence.
What is the focus of economics?
Studying how people make choices that lead to preferred outcomes.
Who published 'An Inquiry into the Nature and Causes of the Wealth of Nations'?
Adam Smith (1723–1790)
What is utility in economics?
A mathematical treatment of 'preferred outcomes'.
Who first formalized the concept of utility?
Léon Walras (1834-1910)
What is decision theory?
Combines probability theory with utility theory for decisions under uncertainty.
What does game theory study?
Interactions in small economies where one player's actions affect another's utility.
Who developed game theory?
John von Neumann and Oskar Morgenstern in 'The Theory of Games and Economic Behavior' (1944).
What is operations research?
A field that emerged to optimize decisions, starting from World War II radar installations.
Who formalized Markov decision processes?
Richard Bellman (1957)
What is satisficing?
Making decisions that are 'good enough' instead of optimal.
Who won the Nobel Prize for work on satisficing?
Herbert Simon (1916–2001)
What is satisficing?
Choosing options that are 'good enough' rather than optimizing decisions.
What does neuroscience study?
The nervous system, particularly the brain.
What did Aristotle note about human brains?
Humans have the largest brain in proportion to their size.
What did Paul Broca demonstrate in 1861?
Localized areas of the brain are responsible for specific cognitive functions.
What is Broca's area responsible for?
Speech production.
Who developed a staining technique for observing neurons?
Camillo Golgi in 1873.
What did Santiago Ramon y Cajal study?
The brain's neuronal structures.
Who first applied mathematical models to the nervous system?
Nicolas Rashevsky in 1936 and 1938.
What are the main parts of a neuron?
What is the typical length of an axon?
1 cm, but can reach up to 1 meter.
How many neurons can a neuron connect with at synapses?
10 to 100,000 other neurons.
Where does most information processing occur in the brain?
In the cerebral cortex.
What is the basic organizational unit in the cerebral cortex?
A column of tissue about 0.5 mm in diameter containing about 20,000 neurons.
What can happen to brain mappings over time?
They can change radically over a few weeks.
Who invented the electroencephalograph (EEG)?
Hans Berger
What technology provides detailed images of brain activity?
Functional magnetic resonance imaging (fMRI)
What can individual neurons be stimulated by?
What is the comparison of computational units between a supercomputer and the human brain?
What is the cycle time of a brain compared to a supercomputer?
What is the memory update rate of a brain?
10^{14} updates/sec
Who is considered the father of experimental psychology?
Wilhelm Wundt
What is Hermann von Helmholtz known for?
Applying the scientific method to the study of human vision
What did Helmholtz's Handbook of Physiological Optics contribute to?
It is described as the most important treatise on the physics and physiology of human vision
What did Wundt emphasize in his psychology experiments?
Carefully controlled experiments and introspection
What is a major limitation of Wundt's experimental psychology?
The subjective nature of the data
What did behaviorism, led by John Watson, reject?
Any theory involving mental processes
Who wrote the influential work 'Behavior of the Lower Organisms'?
H. S. Jennings (1906)
What does cognitive psychology view the brain as?
An information-processing device
Who is associated with the early works of cognitive psychology?
William James (1842–1910)
What did Helmholtz insist about perception?
It involved a form of unconscious logical inference
Who directed Cambridge's Applied Psychology Unit?
Frederic Bartlett (1886–1969)
What are the three key steps of a knowledge-based agent according to Craik?
What was modeled in Donald Broadbent's book 'Perception and Communication'?
Psychological phenomena as information processing
What workshop is considered the start of the field of cognitive science?
A workshop in September 1956 at MIT
What did George Miller present at the 1956 workshop?
The Magic Number Seven
What is essential for artificial intelligence to succeed?
Intelligence and an artifact
What is the modern digital electronic computer considered in AI?
The artifact of choice
What was the first operational computer?
The Heath Robinson, built in 1940 by Alan Turing's team.
What was the purpose of the Heath Robinson?
To decipher German messages during World War II.
What was developed in 1943 by Turing's team?
The Colossus, a powerful general-purpose machine.
Who invented the first operational programmable computer?
Konrad Zuse invented the Z-3 in 1941.
What significant concepts did Zuse invent?
Floating-point numbers and the first high-level programming language, Plankalkül.
Who assembled the first electronic computer?
John Atanasoff and Clifford Berry assembled the ABC between 1940 and 1942.
What was the most influential forerunner of modern computers?
The ENIAC, developed at the University of Pennsylvania.
What trend has occurred in computer hardware since its inception?
An increase in speed and capacity, and a decrease in price.
What was the first programmable machine?
A loom devised by Joseph Marie Jacquard in 1805.
What was the purpose of Babbage's Difference Engine?
To compute mathematical tables for engineering and scientific projects.
What was the Analytical Engine?
Babbage's machine that included addressable memory, stored programs, and conditional jumps.
Who is considered the world's first programmer?
Ada Lovelace, who wrote programs for the Analytical Engine.
What programming language is named after Ada Lovelace?
The programming language Ada.
What contributions has AI made to computer science?
Pioneered ideas like time sharing, interactive interpreters, and personal computers.
Who built automated machines in the 17th century?
Ktesibios of Alexandria built early automated machines.
Who built the first self-controlling machine?
Ktesibios of Alexandria (c. 250 B.C.)
What was the first self-controlling machine?
A water clock with a regulator.
Who created the steam engine governor?
James Watt (1736–1819)
Who invented the thermostat?
Cornelis Drebbel (1572–1633)
What was developed in the 19th century?
The mathematical theory of stable feedback systems.
Who is the central figure in control theory?
Norbert Wiener (1894–1964)
What book did Norbert Wiener publish in 1948?
Cybernetics
Who challenged the behaviorist orthodoxy in psychology?
Wiener, Arturo Rosenblueth, Julian Bigelow
What did Wiener and colleagues view purposive behavior as?
A regulatory mechanism minimizing error.
What is the goal of modern control theory?
Design systems that maximize an objective function over time.
What mathematical tools are used in control theory?
Calculus and matrix algebra.
Who published Verbal Behavior in 1957?
B. F. Skinner
Who wrote a famous review of Skinner's Verbal Behavior?
Noam Chomsky
What did Chomsky's Syntactic Structures address?
Creativity in language.
What theory did Chomsky criticize for not addressing creativity in language?
The behaviorist theory
Who was the Indian linguist whose models influenced Chomsky's theory?
Panini (c. 350 B.C.)
What hybrid field emerged from modern linguistics and AI?
Computational linguistics or natural language processing
What is necessary for understanding language beyond sentence structure?
Understanding of subject matter and context
Who are credited with the first recognized work in artificial intelligence?
Warren McCulloch and Walter Pitts (1943)
What did McCulloch and Pitts propose about artificial neurons?
Each neuron is either 'on' or 'off', responding to neighboring neurons
What is the rule proposed by Donald Hebb for modifying connection strengths?
Hebbian learning
What was the name of the first neural network computer built by Minsky and Edmonds?
SNARC
What influential test did Alan Turing introduce in his 1950 article?
The Turing Test
What did Turing propose instead of simulating the adult mind?
To simulate the child's mind
Who proposed the Dartmouth workshop for artificial intelligence?
John McCarthy
When was the Dartmouth workshop held?
Summer of 1956
What was the main goal of the Dartmouth workshop?
To study artificial intelligence and simulate aspects of learning and intelligence.
Who were some attendees of the Dartmouth workshop?
Which two researchers from Carnegie Tech were prominent at the workshop?
What program did Newell and Simon create?
Logic Theorist (LT)
What did the Logic Theorist claim to solve?
The mind-body problem.
What language did Newell and Simon invent for writing LT?
IPL (Information Processing Language)
What theorem did the Logic Theorist prove that impressed Russell?
A theorem shorter than the one in Principia Mathematica.
Why did AI become a separate field?
It aimed to duplicate human faculties like creativity and self-improvement.
What does AI aim to duplicate?
Why isn't AI a branch of mathematics?
AI embraces human faculties and is a branch of computer science, focusing on autonomous machines.
What was the early period of AI known for?
Full of successes despite primitive computers and programming tools.
Who referred to the early AI period as 'Look, Ma, no hands'?
John McCarthy
What is the General Problem Solver (GPS)?
A program designed to imitate human problem-solving protocols.
What hypothesis did Newell and Simon formulate?
Physical symbol system hypothesis: Intelligence operates by manipulating data structures of symbols.
What did Herbert Gelernter create?
Geometry Theorem Prover: Proved complex mathematical theorems.
Who developed checkers programs that learned to play?
Arthur Samuel
What did Arthur Samuel's program demonstrate?
Computers can learn and improve beyond their initial programming.
What significant contributions did John McCarthy make in 1958?
What is Lisp?
A high-level programming language defined by John McCarthy.
What was the Advice Taker?
A hypothetical program seen as the first complete AI system.
What is the Advice Taker?
A hypothetical program described by McCarthy as the first complete AI system.
What did the Advice Taker aim to do?
Use knowledge to search for solutions to problems and embody general world knowledge.
What principle does the Advice Taker embody?
Knowledge representation and reasoning through formal, explicit representation.
When did Marvin Minsky move to MIT?
In 1958.
What was McCarthy's focus in AI?
Representation and reasoning in formal logic.
What did Minsky develop an outlook against?
An anti-logic outlook.
What did McCarthy start in 1963?
The AI lab at Stanford.
What is the resolution method?
A complete theorem-proving algorithm for first-order logic discovered by J. A. Robinson in 1965.
What was the first project to integrate logical reasoning and physical activity?
The Shakey robotics project at SRI.
What are microworlds?
Limited domains chosen by students that appeared to require intelligence to solve.
What problem did James Slagle's SAINT program solve?
Closed-form calculus integration problems.
What did Tom Evans's ANALOGY program solve?
Geometric analogy problems that appear in IQ tests.
What kind of problems did Daniel Bobrow's STUDENT program solve?
Algebra story problems.
What is the blocks world?
A microworld consisting of solid blocks placed on a tabletop for rearrangement tasks.
Who worked on the vision project in the blocks world?
David Huffman in 1971.
What did Winograd and Cowan's work show?
A large number of elements could collectively represent an individual concept.
Who enhanced Hebb's learning methods with adalines?
Bernie Widrow.
Who introduced the concept of perceptrons?
Frank Rosenblatt in 1962.
Who enhanced learning methods in AI with adalines?
Bernie Widrow (Widrow and Hoff, 1960; Widrow, 1962)
Who developed perceptrons in AI?
Frank Rosenblatt (1962)
What does the perceptron convergence theorem state?
The learning algorithm can adjust connection strengths to match any input data, provided a match exists.
Who made predictions about AI's future in 1957?
Herbert Simon
What did Herbert Simon predict about computers within 10 years?
A computer would be chess champion and a significant mathematical theorem would be proved by machine.
How long did it actually take for Simon's predictions to come true?
Approximately 40 years.
What was a major difficulty of early AI programs?
Most early programs knew nothing of their subject matter; they succeeded through simple syntactic manipulations.
What was a notable failure in early machine translation efforts?
Accurate translation requires background knowledge to resolve ambiguity.
What did a 1966 report by an advisory committee conclude about machine translation?
There has been no machine translation of general scientific text, and none is in immediate prospect.
What was the second kind of difficulty faced by AI?
The intractability of many problems AI attempted to solve.
What strategy did early AI programs use to solve problems?
Trying out different combinations of steps until a solution was found.
What misconception existed before the theory of computational complexity?
Scaling up to larger problems was thought to be a matter of faster hardware and larger memories.
What dampened optimism in resolution theorem proving?
Researchers failed to prove theorems involving more than a few dozen facts.
What is the illusion of unlimited computational power related to?
It was not confined to problem-solving programs and was evident in early experiments with machine evolution (genetic algorithms).
Who were the pioneers of genetic algorithms?
Friedberg (1958) and Friedberg et al. (1959).
What was the goal of genetic algorithms?
To generate a program with good performance for a specific task through random mutations and a selection process.
What was a major criticism of AI in the Lighthill report?
Failure to address the 'combinatorial explosion'.
What happened to AI research funding after the Lighthill report?
The British government ended support for AI research in all but two universities.
What did Minsky and Papert's book Perceptrons (1969) demonstrate?
Perceptrons could learn anything they could represent, but they represented very little.
What limitation did a two-input perceptron have?
It could not be trained to recognize when its two inputs were different.
What caused a decline in funding for neural-net research?
Results from Minsky and Papert's research on perceptrons.
What learning algorithm caused a resurgence in neural-net research in the late 1980s?
Back-propagation learning algorithms.
What are weak methods in AI?
General-purpose search mechanisms that do not scale up to large or difficult problem instances.
What is the alternative to weak methods in problem-solving?
Using more powerful, domain-specific knowledge for larger reasoning steps.
What was the DENDRAL program used for?
Inferring molecular structure from mass spectrometer data.
Who were the key developers of the DENDRAL program?
Ed Feigenbaum, Bruce Buchanan, and Joshua Lederberg.
What input did the DENDRAL program require?
The elementary formula of the molecule and the mass spectrum of its fragments.
What was a naive approach of the DENDRAL program?
Generating all possible structures consistent with the formula and predicting mass spectra for comparison.
What was the purpose of the DENDRAL researchers?
To analyze molecular structures using spectral data and identify common substructures.
What rule is used to recognize a ketone subgroup?
If two peaks at x1 and x2 satisfy: - x1 + x2 = M + 28 - x1 - 28 is a high peak - x2 - 28 is a high peak - At least one of x1 or x2 is high.
What was significant about DENDRAL?
It was the first successful knowledge-intensive system using many special-purpose rules.
What project did Feigenbaum and others start at Stanford?
The Heuristic Programming Project (HPP) to apply expert systems methodology to various fields.
What was the purpose of MYCIN?
To diagnose blood infections using about 450 rules, performing as well as some experts.
How did MYCIN differ from DENDRAL?
MYCIN’s rules were acquired through expert interviews and reflected medical uncertainty with certainty factors.
What did Winograd's SHRDLU system demonstrate?
It showed that natural language understanding could overcome ambiguity but was limited to a specific domain (blocks world).
What did Roger Schank emphasize in understanding natural language?
He claimed that robust understanding requires general knowledge and methods beyond syntax.
What was the focus of Schank and his students' programs?
They focused on representing and reasoning with knowledge required for language understanding.
What were the problems of language understanding identified?
What led to an increase in knowledge representation schemes?
The growth of applications to real-world problems.
What logic-based languages became popular for AI?
What is the frames approach in AI?
A structured method assembling facts about object and event types into a taxonomic hierarchy.
What was the first successful commercial expert system?
R1 at Digital Equipment Corporation.
How much did R1 save DEC annually by 1986?
$40 million a year.
How many expert systems did DuPont have in use by 1988?
100 expert systems in use and 500 in development.
What was the Fifth Generation project announced by Japan?
A 10-year plan to build intelligent computers running Prolog.
What was the purpose of the Microelectronics and Computer Technology Corporation (MCC)?
To assure national competitiveness in AI research.
What happened to the AI industry from 1980 to 1988?
It grew from a few million dollars to billions of dollars.
What is the AI Winter?
A period when many AI companies failed to deliver on promises.
What algorithm was reinvented in the mid-1980s?
The back-propagation learning algorithm.
What did the Parallel Distributed Processing collection cause?
Great excitement in connectionist models of intelligent systems.
How do connectionist and symbolic approaches relate?
They are complementary, not competing.
What are the two fields modern neural network research has bifurcated into?
What has revolutionized AI methodology in recent years?
Building on existing theories, rigorous theorems, and real-world applications.
What did David McAllester state about AI's early isolationism?
AI was isolated from computer science but is now embracing fields like control theory and statistics.
What is required for hypotheses in AI to be accepted?
Hypotheses must undergo rigorous empirical experiments and statistical analysis.
What has dominated the field of speech recognition in recent years?
Approaches based on hidden Markov models (HMMs).
What are the two relevant aspects of HMMs in speech recognition?
What is the significance of HMMs in blind tests?
HMMs have been steadily improving their scores in rigorous blind tests.
How is machine translation related to speech recognition?
Both have evolved from initial enthusiasm to rigorous mathematical frameworks.
What characterized the shift in AI methodology towards neatness?
The field has reached stability and maturity, favoring mathematical rigor.
What is the role of HMMs in understanding speech recognition?
HMMs provide a mathematical framework for the problem without claiming human similarity.
What was the initial approach to machine translation in the 1950s?
Based on sequences of words with models learned according to information theory.
When did the approach to machine translation return to favor?
In the late 1990s.
What was the focus of neural networks research in the 1980s?
To explore capabilities and differences from traditional techniques.
What did Judea Pearl's 1988 work contribute to AI?
Led to acceptance of probability and decision theory in AI.
What is the Bayesian network formalism used for?
Efficient representation and reasoning with uncertain knowledge.
What do normative expert systems do?
Act rationally according to decision theory without imitating human thought.
What has led to the emergence of intelligent agents since 1995?
Progress in solving subproblems of AI and the whole agent problem.
Who worked on the complete agent architecture called SOAR?
Allen Newell, John Laird, and Paul Rosenbloom.
What is one important environment for intelligent agents?
The Internet.
What is a consequence of building complete agents in AI?
Need to reorganize isolated subfields to integrate results.
What is a challenge for reasoning and planning systems in AI?
Handling uncertainty from sensory systems.
What must reasoning and planning systems handle?
Uncertainty
What fields has AI come into closer contact with?
What recent progress has been made in robotic cars?
Who expressed discontent with AI progress?
What do some AI founders believe AI should focus on?
What is the term for the effort to achieve human-level AI?
Human-level AI (HLAI)
When was the first symposium on human-level AI held?
2004
What is the subfield of AI that seeks a universal algorithm?
Artificial General Intelligence (AGI)
When was the first AGI conference held?
2008
What is a concern related to AI creation?
Friendly AI
What has been the main subject of study in computer science?
Algorithms
What does recent AI work suggest is more important than algorithms?
Data
What significant data sources have become available since 2001?
What did Yarowsky's 1995 work achieve?
96% accuracy in word-sense disambiguation without labeled examples
What example did Banko and Brill discuss regarding data performance?
Performance increases with data volume, surpassing algorithm differences
What problem did Hays and Efros address?
Filling in holes in photographs
What algorithm did Hays and Efros define for filling in masked areas in photos?
An algorithm that searches through a collection of photos to find a match for the background.
What was the performance threshold of Hays and Efros' algorithm?
Performance improved from poor with 10,000 photos to excellent with 2 million photos.
What does the knowledge bottleneck in AI refer to?
The challenge of expressing all necessary knowledge for a system.
What have reporters noted about AI's development?
They noted that 'AI Winter' may be yielding to a new Spring.
What is an example of a robotic vehicle?
STANLEY, a driverless car that won the 2005 DARPA Grand Challenge.
What technology did STANLEY use?
Cameras, radar, and laser rangefinders to sense the environment.
What is an application of speech recognition in AI?
Automated systems guiding conversations for booking flights.
What was NASA's Remote Agent program?
The first on-board autonomous planning program for spacecraft operations.
What did IBM's DEEP BLUE achieve?
Defeated world champion Garry Kasparov in chess.
What is a significant impact of DEEP BLUE's victory?
IBM's stock increased by $18 billion.
What role do learning algorithms play in spam fighting?
They classify over a billion messages as spam daily.
What challenge do spam fighters face?
Spammers continually update their tactics, making static approaches ineffective.
What was a logistical application during the Persian Gulf crisis of 1991?
Dynamic analysis for U.S. forces deployment.
What does DART stand for in logistics planning?
Dynamic Analysis and Replanning Tool
How many vehicles, cargo, and people did DART account for during the Persian Gulf crisis?
Up to 50,000
What was the impact of DART on planning time?
Generated a plan in hours that took weeks with older methods
What organization stated that DART paid back their 30-year investment in AI?
Defense Advanced Research Project Agency (DARPA)
How many Roomba robotic vacuum cleaners has iRobot sold?
Over two million
What is the PackBot used for in Iraq and Afghanistan?
Handle hazardous materials, clear explosives, identify snipers
What does the machine translation program translate from Arabic to?
English
What statistical model is used in the Arabic to English translation program?
Built from examples of Arabic-to-English translations and English text totaling two trillion words
What is the main focus of intelligence according to the book?
Rational action
What philosophical idea made AI conceivable?
The mind is like a machine operating on knowledge encoded in an internal language
What did mathematicians provide for AI?
Tools to manipulate logical and probabilistic statements
What formalized the decision-making process in AI?
Economists
What did neuroscientists discover about the brain?
How it works and its similarities and differences from computers
How do psychologists view humans and animals?
As information-processing machines
What role do computer engineers play in AI?
Provide powerful machines for AI applications
What does control theory deal with?
Designing devices that act optimally based on environmental feedback
What has influenced the history of AI?
Cycles of success, optimism, cutbacks, and new creative approaches
What recent advancement has contributed to the rapid progress of AI?
Greater use of scientific methods
What has contributed to the rapid advancement of AI in the past decade?
Greater use of the scientific method in experimenting and comparing approaches.
How have theoretical understanding and real systems progressed in AI?
They have improved hand in hand with each other.
What is the title of Herb Simon's work on AI methodology?
The Sciences of the Artificial (1981)
Who discusses the Turing Test and criticizes its usefulness?
Shieber (1994)
Which book gives a readable account of the philosophical problems of AI?
Artificial Intelligence: The Very Idea by John Haugeland (1985)
What does the Encyclopedia of AI contain?
Survey articles on almost every topic in AI.
What are the major AI conferences mentioned?
What are the main professional societies for AI?
What does AAAI’s AI Magazine contain?
Topical and tutorial articles.
What is the focus of exercise 1.1?
Define in your own words:
- intelligence
- artificial intelligence
- agent
- rationality
- logical reasoning
What does exercise 1.2 ask about Turing's original paper?
Discuss objections to his test for intelligence and their current relevance.
What question does exercise 1.3 pose about reflex actions?
Are reflex actions rational? Are they intelligent?
What is a reflex action?
A reflex action is an automatic response to a stimulus, such as flinching from a hot stove.
Are reflex actions rational or intelligent?
Reflex actions are not considered rational or intelligent as they occur automatically without conscious thought.
What happens if Evans's ANALOGY program scores 200 on an IQ test?
It does not necessarily mean the program is more intelligent than a human; intelligence is multifaceted and not solely based on IQ.
Why is the sea slug Aplysia studied?
Aplysia has about 20,000 large neurons, making it easier to study neural structures and functions.
How does Aplysia's computational power compare to high-end computers?
The comparison depends on the cycle time for neuron updates and the architecture of the high-end computer.
What is introspection?
Introspection is the examination of one's own thoughts and feelings.
Can introspection be inaccurate?
Yes, individuals can misinterpret or be unaware of their true thoughts.
Are supermarket bar code scanners AI?
They are a form of AI as they automate a task but lack advanced intelligence.
Are web search engines instances of AI?
Yes, they utilize algorithms to process and retrieve information intelligently.
Are voice-activated telephone menus AI?
Yes, they use speech recognition and processing to interact with users.
Are internet routing algorithms AI?
Yes, they adapt to network conditions dynamically, showcasing intelligent behavior.
Do humans learn complex mathematics for cognitive activities?
Most humans do not learn complex mathematics before college, yet cognitive models may involve such computations.
Why does evolution favor rational systems?
Rational systems enhance survival by effectively achieving goals related to reproduction and resource acquisition.
Is AI a science or engineering?
AI encompasses both science and engineering, as it involves theoretical research and practical application.
Can computers be intelligent?
Computers follow programmed instructions but can exhibit intelligent behavior through complex algorithms.
Can animals be intelligent?
Animals act based on instincts and learned behaviors, indicating a form of intelligence.
Can the laws of physics dictate intelligence?
While all systems operate under physical laws, intelligence arises from complex interactions, not just physical dictates.
What tasks can computers currently solve?
Tasks include web searching, basic games, and limited driving but not complex tasks like surgery or creative writing.
What are some currently infeasible tasks for AI?
Examples include playing competitive bridge, complex surgery, and creative storytelling.
What is the purpose of AI contests?
AI contests encourage innovation by defining tasks for researchers to solve, advancing the field.
What is the DARPA Grand Challenge?
A contest for robotic cars to navigate autonomously, showcasing advancements in robotics and AI.
What is the International Planning Competition?
A contest that challenges AI systems to solve planning problems effectively.
What are intelligent agents?
Systems that can be viewed as perceiving their environment through sensors and acting upon it through actuators.
What is a rational agent?
An agent that behaves as well as possible based on its environment.
What is a percept?
The agent’s perceptual inputs at any given instant.
What is a percept sequence?
The complete history of everything the agent has ever perceived.
What does the agent function do?
Maps any given percept sequence to an action.
What is the difference between agent function and agent program?
Agent function is an abstract mathematical description; agent program is a concrete implementation.
What is an example of a simple agent?
A vacuum-cleaner agent in a made-up world.
What do agents use to interact with their environment?
Sensors to perceive and actuators to act.
What does the vacuum agent perceive in the vacuum-cleaner world?
What actions can the vacuum agent choose from?
What is a simple agent function in the vacuum-cleaner world?
If the current square is dirty, then suck; otherwise, move to the other square.
What defines the goodness of an agent?
Filling out the agent function table correctly.
What is a rational agent?
An agent that does the right thing according to its performance measure.
What does the performance measure evaluate?
The desirability of the sequence of environment states generated by the agent's actions.
What is an important aspect to define success for an agent?
Success should be defined in terms of the environment states, not the agent's opinion of its performance.
What can affect an agent's perception of its performance?
An agent can delude itself about its performance, leading to a false sense of rationality.
What is the goal of analyzing agents?
To understand how artifacts interact with the world, particularly in AI with significant computational resources.
What is a performance measure in agent design?
A criterion to evaluate success, tailored to the task and agent.
How can a vacuum-cleaner agent maximize its performance measure?
By cleaning dirt and then dumping it, repeating this process.
What is a better performance measure for a vacuum-cleaner agent?
Reward for clean squares, with penalties for electricity and noise.
What defines a rational agent?
An agent that selects actions to maximize its performance measure based on percept sequences and knowledge.
What factors influence the rationality of an agent?
What is the performance measure for the vacuum-cleaner agent example?
One point for each clean square per time step over 1000 time steps.
What actions can the vacuum-cleaner agent perform?
Left, Right, and Suck.
What does the vacuum-cleaner agent perceive?
Its location and whether that location contains dirt.
When would the vacuum-cleaner agent be considered irrational?
When it oscillates back and forth after all dirt is cleaned.
What philosophical questions arise from agent performance measures?
What happens when the agent oscillates needlessly after cleaning?
It fares poorly due to a penalty for each movement left or right.
What should a better agent do once all squares are clean?
It should do nothing if sure all squares are clean.
What should an agent do if clean squares can become dirty again?
It should occasionally check and re-clean them if needed.
What must an agent do in an unknown environment?
It needs to explore rather than stick to known squares.
What is the difference between rationality and omniscience?
Rationality maximizes expected performance; omniscience knows actual outcomes.
What example illustrates the difference between rationality and perfection?
Crossing the street while a cargo door falls off an airliner.
What does rationality not require according to the definition?
It does not require omniscience.
What should a rational agent do before crossing a busy road?
It should look both ways before crossing.
What is an important part of rationality related to percepts?
Gathering information to modify future percepts.
What is the role of information gathering in rationality?
It helps maximize expected performance.
What does a rational agent need to do with its perceptions?
It must learn as much as possible from what it perceives.
What happens in extreme cases where the environment is completely known?
The agent need not perceive or learn; it simply acts correctly.
What is an example of a fragile agent mentioned in the text?
The dung beetle that continues its task even if the dung ball is removed.
What does the dung beetle do to plug its nest?
Fetches a ball of dung from a nearby heap.
What happens if the dung ball is removed from the beetle's grasp?
The beetle continues to pantomime plugging the nest.
What does the female sphex wasp do after stinging a caterpillar?
Drags the caterpillar to its burrow.
What does the sphex wasp do if the caterpillar is moved?
Reverts to the 'drag' step of its plan.
What does the lack of autonomy in an agent mean?
It relies on prior knowledge of its designer.
What should a rational agent do to be considered autonomous?
Learn to compensate for partial or incorrect prior knowledge.
How can a vacuum-cleaning agent improve its performance?
By learning to foresee where and when additional dirt will appear.
What is the PEAS description?
Performance, Environment, Actuators, Sensors.
What is essential for specifying a task environment?
To specify the performance measure, environment, actuators, and sensors.
What is an example of a complex task environment?
An automated taxi driver.
Why is the automated taxi driver task considered open-ended?
Novel combinations of circumstances can arise.
What does PEAS stand for in the context of the taxi's task environment?
Performance Measure, Environment, Actuators, Sensors
What is the performance measure for the automated taxi driver?
What are the elements of the taxi's environment?
What actuators does an automated taxi have?
What sensors are used in an automated taxi?
What are some desirable qualities for the automated taxi's performance measure?
What types of roads must the taxi driver deal with?
What optional driving conditions might the taxi face?
What communication methods might the automated taxi use?
What is a key feature of the sensors in an automated taxi?
They include controllable video cameras and sensors for distances to obstacles.
What is an example of a simple environment for a robot?
A robot inspecting parts on a conveyor belt.
What is the significance of the relationship between agent behavior and environment?
It determines the complexity of the agent's performance measure.
What are the possible actions for simple software agents?
What is a softbot designed to do?
What abilities does a softbot operator need?
What characterizes the Internet environment?
What dimensions categorize task environments in AI?
What is a fully observable environment?
An environment where sensors provide complete state information at all times.
What defines a partially observable environment?
An environment where some state aspects are missing due to sensor limitations.
What is an unobservable environment?
An environment where the agent has no sensors at all.
What is the difference between single-agent and multiagent environments?
Single-agent involves one agent (e.g., solving a puzzle), multiagent involves multiple agents (e.g., playing chess).
What type of environment is an agent solving a crossword puzzle in?
Single-agent environment
What type of environment is an agent playing chess in?
Two-agent environment
What is the key distinction for treating another vehicle as an agent?
Whether its behavior maximizes a performance measure dependent on agent A's behavior
What type of environment is chess classified as?
Competitive multiagent environment
What type of environment is taxi driving classified as?
Partially cooperative multiagent environment
What is a characteristic of communication in multiagent environments?
It often emerges as a rational behavior
What defines a deterministic environment?
Next state is completely determined by current state and agent's action
What defines a stochastic environment?
Next state is not completely determined; uncertainty exists
What is the implication of an uncertain environment?
It is not fully observable or not deterministic
What is a nondeterministic environment?
Actions characterized by possible outcomes without probabilities attached
What is an episodic task environment?
Agent's experience is divided into atomic episodes where actions are independent
What is a sequential task environment?
Next episode depends on actions taken in previous episodes
What is the difference between sequential and episodic environments?
What characterizes a static environment?
The environment does not change while the agent is deciding on an action.
What characterizes a dynamic environment?
The environment can change while the agent is deliberating.
What is a semidynamic environment?
The environment does not change with time, but the agent’s performance score does.
What is the difference between discrete and continuous environments?
What does a known environment imply?
Outcomes for all actions are given, or their probabilities if stochastic.
What does an unknown environment imply?
The agent must learn how the environment works to make good decisions.
What is the hardest case for an environment?
Partially observable, multiagent, stochastic, sequential, dynamic, continuous, and unknown.
How is the part-picking robot described in terms of environment?
It is described as episodic, as it normally considers each part in isolation.
What is the antonym of 'parallel' in computer science?
'ial'
What are the characteristics of a crossword puzzle task environment?
What are the characteristics of chess with a clock task environment?
What are the characteristics of poker task environment?
What are the characteristics of backgammon task environment?
What are the characteristics of taxi driving task environment?
What are the characteristics of medical diagnosis task environment?
What is the role of the code repository in the context of environments?
It includes implementations of environments and a simulator to evaluate agent behavior.
What does the environment generator do?
It selects particular environments for running agents with specific likelihoods.
What is a rational agent for a given environment class?
It maximizes performance across various environments.
What does a rational agent maximize in an environment class?
Average performance
What is the formula for an agent?
agent = architecture + program
What does an agent program implement?
The agent function
What does the architecture provide to the agent program?
Percepts from sensors
What is the input for agent programs?
Current percept from sensors
What is the difference between agent program and agent function?
Agent program takes current percept; agent function takes entire percept history.
What is the purpose of the table in the agent program?
To represent the agent function and decide actions.
What is a potential failure of the table-driven approach to agent construction?
Exponential growth of entries in the lookup table.
What does the function TABLE-DRIVEN-AGENT do?
Returns an action based on the current percept.
What is the persistent variable in TABLE-DRIVEN-AGENT?
percepts, a sequence initially empty
What is the data rate of visual input from a single camera in an automated taxi?
27 megabytes per second
What is the resolution of the camera in the automated taxi?
640×480 pixels
How many entries does the lookup table have for an hour's driving?
Over 10^50,000,000,000 entries
How many entries would a chess lookup table have?
At least 10^150 entries
What is the size comparison of the lookup table to the number of atoms in the observable universe?
Less than 10^80 atoms
What are the four basic kinds of agent programs?
What does a simple reflex agent base its actions on?
Current percept, ignoring percept history
What is the action of the vacuum agent when the status is dirty?
Return Suck
What happens when the location is A for the vacuum agent?
Return Right
What happens when the location is B for the vacuum agent?
Return Left
What is the advantage of simple reflex agents compared to lookup tables?
They require less space and processing
What is a notable example of replacing large tables with a program?
Newton’s method for square roots
What happens when the brake lights come on?
Initiate braking.
What do we call the connection between condition and action?
Condition–action rule.
How is a condition–action rule expressed?
if car-in-front-is-braking then initiate-braking.
What are some connections humans have?
Learned responses and innate reflexes.
What is a simple reflex agent?
Acts according to a rule matching the current state.
What does the INTERPRET-INPUT function do?
Generates an abstracted description of the current state.
What does the RULE-MATCH function return?
The first rule that matches the given state description.
What is a limitation of simple reflex agents?
Limited intelligence; works only with fully observable environments.
What happens if the braking rule cannot determine the condition?
The agent may brake unnecessarily or not at all.
What happens to a reflex vacuum agent without a location sensor?
It may get stuck in infinite loops.
What are the possible percepts for a reflex vacuum agent?
[Dirty] and [Clean].
What does a reflex vacuum agent do in response to [Dirty]?
Sucks up the dirt.
What is the risk of moving left or right in a vacuum environment?
It may fail if starting in certain squares (A or B).
What happens if the agent starts in square A and receives [Clean]?
Moving Left fails forever.
What happens if the agent starts in square B and receives [Clean]?
Moving Right fails forever.
What can help escape from infinite loops in simple reflex agents?
Randomizing actions.
What might a vacuum agent do when it perceives [Clean]?
Flip a coin to choose between Left and Right.
How many steps does it take on average for the agent to reach the other square?
Two steps.
When is randomization usually not rational?
In single-agent environments.
What do model-based reflex agents keep track of?
The part of the world they can't see now.
What is the internal state of a braking problem agent?
The previous frame from the camera.
What knowledge is needed to update internal state information?
What is a model of the world in the context of agents?
Knowledge about how the world works.
What does a model-based reflex agent use to generate the current state?
The current percept and the old internal state.
What is the function that updates the state in a model-based reflex agent?
UPDATE-STATE.
What does the MODEL-BASED-REFLEX-AGENT function return?
An action.
What does the agent's model describe?
How the next state depends on current state and action.
What does the RULE-MATCH function do?
It matches the current state with the rules and determines the action to take.
What does a model-based reflex agent do?
It keeps track of the current state of the world and chooses actions like a reflex agent.
What is responsible for creating the new internal state description?
The agent's internal model.
How does the representation of models and states vary?
It varies widely depending on the environment type and agent design technology.
What does the box labeled 'what the world is like now' represent?
It represents the agent's 'best guess' about the current state.
Why might uncertainty about the current state be unavoidable?
Because the agent may not have complete information about the environment.
What does a model-based agent's internal state not have to describe literally?
It does not have to describe 'what the world is like now' literally.
What does a goal-based agent keep track of?
It keeps track of the world state and a set of goals to achieve.
What action does a goal-based agent choose?
It chooses an action that will eventually lead to achieving its goals.
What might influence a taxi's decision to fill up with gas?
The taxi's rule about having at least half a tank of gas.
What is the title of the book?
Artificial Intelligence: A Modern Approach
Who are the authors of the book?
What is the subtitle of the book?
A Modern Approach
What is the focus of the book?
Explores logic, probability, perception, reasoning, learning, action.
What does the subtitle 'A Modern Approach' imply?
Synthesizes current knowledge into a common framework.
What is the ISBN-13 of the book?
978-0-13-604259-4
What is the ISBN-10 of the book?
0-13-604259-7
What is the purpose of this book according to the authors?
To explore the full breadth of AI, including depth in various areas.
What year was the copyright for this edition?
2010, 2003, 1995
Who is the Vice President and Editorial Director?
Marcia J. Horton
What is the role of Michael Hirsch?
Editor-in-Chief
What are some of the contributing writers?
What is the publication's protection status?
Protected by copyright; permissions required for reproduction.
What does the book encompass?
Microelectronic devices to robotic planetary explorers.
What is the significance of the Library of Congress Cataloging?
Data on file for the publication.
What is stated about the authors' efforts in preparing the book?
Best efforts in development, research, and testing effectiveness.
What does the book apologize for?
Subfields being less recognizable due to the common framework approach.
What are some important applications of AI technology?
What are the algorithmic landmarks mentioned in AI?
What is the emphasis on in partially observable environments?
What are the three types of representations for agents?
What new material has been added regarding probabilistic models?
What has been rewritten in the machine-learning chapter?
What percentage of citations in the book are from works published after 2003?
20%
What is the main theme of the book related to intelligent agents?
AI is the study of agents that receive percepts and perform actions.
How many chapters does the book contain?
27 chapters
What is the primary aim of the book?
Convey ideas from 50 years of AI research and 2000 years of related work.
What does the book include to make key ideas concrete?
Pseudocode algorithms
What type of course is the book primarily intended for?
Undergraduate course or course sequence
What is the role of learning in agent design?
Extends the designer's reach into unknown environments.
What is the prerequisite for the book?
Familiarity with basic concepts of computer science at a sophomore level.
What are the useful mathematical topics mentioned?
Where can you find exercises for each chapter?
At the end of each chapter in the book.
What icon marks exercises requiring programming?
A keyboard icon.
What does the pointing icon signify in the book?
Important points are marked with it.
How extensive is the index in the book?
Around 6,000 items.
What can you find on the Web site aima.cs.berkeley.edu?
What is depicted on the cover of the book?
Final position from the game 6 of Kasparov vs. DEEP BLUE chess match.
Who was forced to resign in the chess match?
Garry Kasparov.
What does the cover feature alongside Kasparov?
Who wrote Chapter 24 on computer vision?
Jitendra Malik and David Forsyth.
Who contributed to Chapter 25 on robotics?
Sebastian Thrun.
Which chapter was partially written by Vibhu Mittal?
Chapter 22 on natural language.
Who wrote Chapter 24 on computer vision?
Jitendra Malik and David Forsyth
Who wrote Chapter 25 on robotics?
Sebastian Thrun
Who contributed to Chapter 22 on natural language?
Vibhu Mittal
Who reviewed the manuscript?
Who formatted and improved the diagrams in this edition?
Jon Barron
Who helped with diagrams and algorithms in previous editions?
Who wrote and maintains the Java code examples on the Web site?
Who wrote the robotics chapter for the first edition?
John Canny
Who researched the historical notes?
Douglas Edwards
Who provided extensive improvements by reading every chapter?
Julie Sussman, P.P.A.
Who are the authors thanking for their support and encouragement?
Which institutions' librarians are thanked for their help?
What tools revolutionized the way research is done?
Who provided especially helpful comments on the book?
Who is Stuart Russell?
What awards has Stuart Russell received?
What are some notable works of Stuart Russell?
Who is Peter Norvig?
What was Peter Norvig's role at NASA?
What is the focus of the authors' work?
Key concepts related to machine learning and artificial intelligence
What is the significance of the listed names?
They are contributors and experts in the field of artificial intelligence and machine learning.
What is the role of the Association for Computing Machinery?
Promotes computing as a science and profession.
Where was the Computational Sciences Division led by the individual?
NASA Ames Research Center.
What field did he oversee at NASA?
Research and development in artificial intelligence and robotics.
What company was he chief scientist at?
Junglee.
What did he help develop at Junglee?
One of the first Internet information extraction services.
Which degree did he receive from Brown University?
B.S. in applied mathematics.
What degree did he obtain from the University of California at Berkeley?
Ph.D. in computer science.
What awards did he receive from Berkeley?
Distinguished Alumni and Engineering Innovation awards.
What medal did he receive from NASA?
Exceptional Achievement Medal.
Where has he been a professor?
University of Southern California.
What is one of his books on AI programming?
Paradigms of AI Programming: Case Studies in Common Lisp.
What is the title of his book on translation systems?
Verbmobil: A Translation System for Face-to-Face Dialog.
What is the focus of Section I in the contents?
Artificial Intelligence.
What does 2.1 cover in the contents?
Agents and Environments.
What is discussed in 3.1 of the contents?
Problem-Solving Agents.
What does 4.1 focus on?
Local Search Algorithms and Optimization Problems.
What is the focus of 5.1 in the contents?
Games.
What concept is introduced in 6.1?
Defining Constraint Satisfaction Problems.
What is a Constraint Satisfaction Problem (CSP)?
A problem where the goal is to find values for variables that satisfy a set of constraints.
What are the main methods for solving CSPs?
What does Constraint Propagation do in CSPs?
Infers variable values based on constraints to reduce the search space.
What is Backtracking Search?
A depth-first search algorithm that incrementally builds candidates for solutions and abandons candidates as soon as it determines they cannot lead to a valid solution.
What is Local Search in CSPs?
An optimization technique that starts with a complete assignment and iteratively makes small changes to find a better solution.
What is the Wumpus World?
A standard problem in AI used to illustrate the concepts of knowledge-based agents and reasoning.
What is Propositional Logic?
A branch of logic dealing with propositions that can be true or false, used in knowledge representation.
What is First-Order Logic?
An extension of propositional logic that includes quantifiers and predicates, allowing for more expressive statements.
What is Inference in First-Order Logic?
The process of deriving new information from known facts using logical rules.
What is Classical Planning?
The process of creating a sequence of actions to achieve specific goals based on a model of the world.
What are Planning Graphs?
Data structures that represent the planning problem, showing actions and their effects over time.
What is Hierarchical Planning?
A planning approach that breaks down tasks into subtasks, allowing for more manageable planning.
What is Ontological Engineering?
The process of creating and managing ontologies to represent knowledge in a domain.
What are Mental Events in knowledge representation?
Events that occur in the mind, such as beliefs and desires, used to model cognitive processes.
What is Reasoning with Default Information?
A method of reasoning that allows for conclusions to be drawn in the absence of complete information.
What is discussed in section 12.6?
Reasoning with Default Information
What is the topic of section 12.7?
The Internet Shopping World
What is covered in section 12.8?
Summary, Bibliographical and Historical Notes, Exercises
What is the focus of section 13?
Quantifying Uncertainty
What does section 13.1 address?
Acting under Uncertainty
What is explained in section 13.2?
Basic Probability Notation
What is the subject of section 13.3?
Inference Using Full Joint Distributions
What does section 13.4 discuss?
Independence
What is the focus of section 13.5?
Bayes’ Rule and Its Use
What is revisited in section 13.6?
The Wumpus World
What does section 13.7 summarize?
Summary, Bibliographical and Historical Notes, Exercises
What is the main topic of section 14?
Probabilistic Reasoning
What does section 14.1 cover?
Representing Knowledge in an Uncertain Domain
What is explained in section 14.2?
The Semantics of Bayesian Networks
What does section 14.3 address?
Efficient Representation of Conditional Distributions
What is the topic of section 14.4?
Exact Inference in Bayesian Networks
What does section 14.5 discuss?
Approximate Inference in Bayesian Networks
What is covered in section 14.6?
Relational and First-Order Probability Models
What does section 14.7 focus on?
Other Approaches to Uncertain Reasoning
What is summarized in section 14.8?
Summary, Bibliographical and Historical Notes, Exercises
What is the main topic of section 15?
Probabilistic Reasoning over Time
What does section 15.1 address?
Time and Uncertainty
What is discussed in section 15.2?
Inference in Temporal Models
What is the focus of section 15.3?
Hidden Markov Models
What does section 15.4 explain?
Kalman Filters
What is covered in section 15.5?
Dynamic Bayesian Networks
What does section 15.6 focus on?
Keeping Track of Many Objects
What is summarized in section 15.7?
Summary, Bibliographical and Historical Notes, Exercises
What is the main topic of section 16?
Making Simple Decisions
What does section 16.1 discuss?
Combining Beliefs and Desires under Uncertainty
What is explained in section 16.2?
The Basis of Utility Theory
What is the focus of section 16.3?
Utility Functions
What does section 16.4 cover?
Multiattribute Utility Functions
What is discussed in section 16.5?
Decision Networks
What is the topic of section 16.6?
The Value of Information
What does section 16.7 focus on?
Decision-Theoretic Expert Systems
What is summarized in section 16.8?
Summary, Bibliographical and Historical Notes, Exercises
What is the main topic of section 17?
Making Complex Decisions
What does section 17.1 cover?
Sequential Decision Problems
What is explained in section 17.2?
Value Iteration
What is the focus of section 17.3?
Policy Iteration
What does section 17.4 discuss?
Partially Observable MDPs
What is the topic of section 17.5?
Decisions with Multiple Agents: Game Theory
What is covered in section 17.6?
Mechanism Design
What is summarized in section 17.7?
Summary, Bibliographical and Historical Notes, Exercises
What is the main topic of section 18?
Learning from Examples
What does section 18.1 discuss?
Forms of Learning
What is explained in section 18.2?
Supervised Learning
What does section 18.3 cover?
Learning Decision Trees
What is Supervised Learning?
A type of machine learning where a model is trained on labeled data.
What are Decision Trees used for in machine learning?
To model decisions and their possible consequences.
What is the purpose of Evaluating and Choosing the Best Hypothesis?
To select the most effective model based on performance metrics.
What does the Theory of Learning encompass?
Principles and frameworks that describe how learning occurs.
What is Regression in machine learning?
A method for predicting continuous outcomes based on input variables.
What are Artificial Neural Networks?
Computational models inspired by the human brain, used for various tasks.
What are Nonparametric Models?
Models that do not assume a fixed form for the underlying data distribution.
What are Support Vector Machines?
A supervised learning model used for classification and regression tasks.
What is Ensemble Learning?
Combining multiple models to improve performance and robustness.
What does Practical Machine Learning involve?
Application of machine learning techniques to real-world problems.
What is the focus of Knowledge in Learning?
Understanding how knowledge is represented and utilized in learning.
What is Explanation-Based Learning?
A method where learning is guided by explanations of the data.
What is Inductive Logic Programming?
A method of machine learning that uses logic programming as a uniform representation.
What is Statistical Learning?
A framework for understanding the relationships among data through statistics.
What is the EM Algorithm?
An iterative method for finding maximum likelihood estimates in models with hidden variables.
What is Reinforcement Learning?
A type of learning where an agent learns to make decisions by receiving rewards or penalties.
What does Natural Language Processing focus on?
The interaction between computers and human language.
What is a Language Model?
A statistical model that predicts the next word in a sequence based on prior words.
What is Text Classification?
The process of assigning predefined categories to text data.
What is Machine Translation?
Automatic translation of text from one language to another using algorithms.
What is Speech Recognition?
The ability of a machine to identify and process human speech.
What does Perception in AI refer to?
The process of interpreting sensory information to understand the environment.
What is Object Recognition?
The ability of a system to identify and classify objects within images.
What is the significance of 3D Reconstruction?
Creating a 3D model from 2D images or data to understand spatial relationships.
What is the focus of Object Recognition by Appearance?
Pages: 942
What is covered in Reconstructing the 3D World?
Pages: 947
What does Object Recognition from Structural Information entail?
Pages: 957
What is discussed in Using Vision?
Pages: 961
What is the content of Robotics section?
Pages: 971
What does Robot Hardware cover?
Pages: 973
What is the focus of Robotic Perception?
Pages: 978
What is discussed in Planning to Move?
Pages: 986
What does Planning Uncertain Movements address?
Pages: 993
What is covered in Moving?
Pages: 997
What is discussed in Robotic Software Architectures?
Pages: 1003
What are the Application Domains in robotics?
Pages: 1006
What does the Philosophical Foundations section explore?
Pages: 1020
What is discussed in Weak AI?
Pages: 1020
What does Strong AI refer to?
Pages: 1026
What are the Ethics and Risks of AI development?
Pages: 1034
What is covered in the AI: The Present and Future section?
Pages: 1044
What are the Agent Components?
Pages: 1044
What does Agent Architectures entail?
Pages: 1047
What is discussed in Are We Going in the Right Direction?
Pages: 1049
What does What If AI Does Succeed? explore?
Pages: 1051
What is included in the Mathematical Background?
Pages: 1053
What does Complexity Analysis and O() Notation cover?
Pages: 1053
What are Vectors, Matrices, and Linear Algebra?
Pages: 1055
What does Probability Distributions discuss?
Pages: 1057
What is covered in Notes on Languages and Algorithms?
Pages: 1060
What does Defining Languages with Backus–Naur Form (BNF) entail?
Pages: 1060
What is discussed in Describing Algorithms with Pseudocode?
Pages: 1061
What does Online Help cover?
Pages: 1062
What is included in the Bibliography?
Pages: 1063
What is found in the Index?
Pages: 1095
What is the significance of intelligence according to the text?
Homo sapiens—man the wise
What is the main goal of artificial intelligence?
To build intelligent entities.
When was the term artificial intelligence coined?
In 1956.
What is the perception of AI among scientists in other disciplines?
It is often cited as the field they would most like to be in.
What does AI still have openings for?
Several full-time Einsteins and Edisons.
What does AI encompass?
A huge variety of subfields.
What does AI encompass?
AI encompasses a variety of subfields including: - Learning - Perception - Playing chess - Proving mathematical theorems - Writing poetry - Driving cars - Diagnosing diseases
What are the two dimensions for defining AI?
What does a rationality system do?
A rationality system does the 'right thing' given what it knows.
What is the Turing Test?
The Turing Test determines if a computer can mimic human responses well enough that a human cannot distinguish them from a person's.
What capabilities must a computer have to pass the Turing Test?
To pass the Turing Test, a computer must have: - Natural language processing - Knowledge representation - Automated reasoning - Machine learning
What is natural language processing?
Natural language processing enables a computer to communicate successfully in English.
What is knowledge representation?
Knowledge representation is the ability to store what a computer knows or hears.
What is automated reasoning?
Automated reasoning is using stored information to answer questions and draw new conclusions.
What is machine learning?
Machine learning allows a computer to adapt to new circumstances and detect and extrapolate patterns.
What are some systematic errors in human reasoning?
Cataloged by Kahneman et al. (1982).
What does Turing's test avoid?
Direct physical interaction between the interrogator and the computer.
What does the total Turing Test include?
A video signal and the ability to pass physical objects through a hatch.
What are the requirements to pass the total Turing Test?
What is the focus of AI researchers regarding the Turing Test?
Studying underlying principles of intelligence rather than duplicating an exemplar.
What is the cognitive modeling approach in AI?
Determining how humans think to express it as a computer program.
What are the three ways to understand human thought?
What is evidence that a program's mechanisms could operate in humans?
If the program’s input-output behavior matches corresponding human behavior.
Who developed the General Problem Solver (GPS)?
Allen Newell and Herbert Simon (1961).
What field combines AI models and psychology techniques?
Cognitive science.
What was often confused in the early days of AI?
The approaches of algorithm performance and human performance modeling.
Who attempted to codify 'right thinking'?
The Greek philosopher Aristotle.
Who was one of the first philosophers to codify 'right thinking'?
Aristotle
What are syllogisms used for?
Patterns for argument structures that yield correct conclusions.
What field did the study of 'laws of thought' initiate?
Logic
What notation did logicians develop in the 19th century?
A precise notation for statements about objects and their relations.
What is the logicist tradition in AI?
It aims to build intelligent systems based on logical programs.
What are the two main obstacles in the logicist approach?
What defines a rational agent?
An agent that acts to achieve the best outcome or best expected outcome under uncertainty.
What is one way a rational agent can act?
By reasoning logically to a conclusion that achieves its goals.
What is an example of a reflex action?
Recoiling from a hot stove.
What skills enable agents to act rationally?
Knowledge representation, reasoning, and learning.
What are the advantages of the rational-agent approach?
What standard is well defined in the rational-agent approach?
The standard of rationality.
What is the standard of rationality defined by?
Mathematically well defined and completely general.
How is human behavior described in the text?
Well adapted for one specific environment; defined by the sum total of human actions.
What does the book focus on?
General principles of rational agents and components for constructing them.
Is achieving perfect rationality feasible in complicated environments?
No, the computational demands are too high.
What is the working hypothesis regarding perfect rationality?
It simplifies the problem and provides a setting for foundational material.
What do Chapters 5 and 17 discuss?
Limited rationality—acting appropriately with insufficient time for computations.
What disciplines contributed to Artificial Intelligence?
Philosophy, among others, with various ideas and techniques.
What questions does the history of AI focus on?
Formal rules for conclusions, mind-physical brain relation, knowledge origins, knowledge leading to action.
Who was the first to formulate laws governing the rational mind?
Aristotle (384–322 B.C.).
What did Aristotle develop for proper reasoning?
An informal system of syllogisms.
What did Ramon Lull propose?
Useful reasoning could be carried out by a mechanical artifact.
What comparison did Thomas Hobbes make about reasoning?
He likened it to numerical computation.
Who designed a mechanical calculator around 1500?
Leonardo da Vinci (1452–1519).
What was the first known calculating machine?
Constructed around 1623 by Wilhelm Schickard.
What is the Pascaline known for?
Built in 1642 by Blaise Pascal; famous for its arithmetic capabilities.
What capabilities did Leibniz's calculator have?
Add, subtract, multiply, and take roots.
What idea did Thomas Hobbes suggest in Leviathan?
The concept of an 'artificial animal'.
What does the heart represent in the quote?
A spring
What do the nerves represent in the quote?
So many strings
What do the joints represent in the quote?
So many wheels
Who discussed the distinction between mind and matter?
René Descartes
What problem arises from a purely physical conception of the mind?
It limits free will.
What philosophy advocates the power of reasoning?
Rationalism
What is dualism?
The belief that a part of the mind is outside of nature.
What does materialism assert about the mind?
The mind is the brain's operation according to physical laws.
Who started the empiricism movement?
Francis Bacon
What did John Locke state about understanding?
Nothing is in the understanding, which was not first in the senses.
What principle did David Hume propose?
The principle of induction.
What doctrine did the Vienna Circle develop?
Logical positivism
What does logical positivism combine?
Rationalism and empiricism.
What is the confirmation theory?
It analyzes the acquisition of knowledge from experience.
What did Carnap's book define?
An explicit computational procedure for extracting knowledge.
What is vital for artificial intelligence?
The connection between knowledge and action.
How did Aristotle justify actions?
By a logical connection between goals and knowledge of outcomes.
What does Aristotle suggest about deliberation in the Nicomachean Ethics?
We deliberate not about ends, but about means.
What is the outcome of reasoning according to Aristotle's algorithm?
The conclusion is an action.
Who implemented Aristotle's algorithm 2300 years later?
Newell and Simon with their GPS program.
What is the main focus of goal-based analysis?
It is useful but does not specify actions when several actions achieve a goal or none completely.
Who described a quantitative formula for deciding actions?
Antoine Arnauld.
What did John Stuart Mill promote in his book Utilitarianism?
The idea of rational decision criteria in all spheres of human activity.
What are the three fundamental areas for mathematical formalization in AI?
Logic, computation, and probability.
Who worked on propositional logic and introduced Boolean logic?
George Boole.
What did Gottlob Frege extend in 1879?
Boole's logic to include objects and relations, creating first-order logic.
What did Alfred Tarski introduce?
A theory of reference to relate logic objects to real-world objects.
What is considered the first nontrivial algorithm?
Euclid's algorithm for computing greatest common divisors.
What does the word 'algorithm' originate from?
The writings of al-Khowarazmi, a Persian mathematician.
What did Kurt Gödel show in 1930 regarding first-order logic?
There exists an effective procedure to prove any true statement, but it cannot capture mathematical induction.
What did Gödel's incompleteness theorem show?
There are true statements in Peano arithmetic that are undecidable and have no proof within the theory.
What does the Church–Turing thesis state?
The Turing machine is capable of computing any computable function.
What is an intractable problem?
A problem where the time to solve instances grows exponentially with the size of the instances.
Who pioneered the theory of NP-completeness?
Steven Cook (1971) and Richard Karp (1972).
What is the significance of NP-complete problems?
Problems that can be reduced to NP-complete problems are likely to be intractable.
Who first framed the idea of probability?
Gerolamo Cardano (1501–1576).
What was Blaise Pascal's contribution to probability?
He showed how to predict the future of an unfinished gambling game and assign average payoffs.
What role does probability play in quantitative sciences?
Helps deal with uncertain measurements and incomplete theories.
Who advanced the theory of probability?
What did Thomas Bayes propose?
A rule for updating probabilities based on new evidence.
When did economics begin as a science?
In 1776, with Adam Smith's publication of 'An Inquiry into the Nature and Causes of the Wealth of Nations'.
What concept did Adam Smith introduce in economics?
Economies consist of individual agents maximizing their own economic well-being.
What is the mathematical treatment of utility?
First formalized by Léon Walras (1834-1910) and improved by Frank Ramsey and others.
What does decision theory combine?
Probability theory and utility theory.
What is the focus of game theory?
The actions of one player significantly affect the utility of another.
What is a key aspect of operations research?
Emerging from WWII to optimize decisions, such as radar installations.
What are Markov decision processes?
A class of sequential decision problems formalized by Richard Bellman (1957).
What concept did Herbert Simon introduce in AI?
Satisficing—making decisions that are 'good enough' rather than optimal.
What is the concept of satisficing?
Choosing options that are 'good enough' instead of striving for an optimal decision.
Who contributed to the resurgence of interest in decision-theoretic techniques?
Wellman, 1995
What is neuroscience?
The study of the nervous system, especially the brain.
What did Aristotle note about human brains?
Humans have the largest brain in proportion to body size.
When was the brain recognized as the seat of consciousness?
Middle of the 18th century.
What did Paul Broca study in 1861?
Aphasia in brain-damaged patients, showing localized brain functions.
What is Broca’s area responsible for?
Speech production.
Who developed a staining technique for observing neurons?
Camillo Golgi in 1873.
What did Santiago Ramon y Cajal study?
The brain’s neuronal structures.
Who was the first to apply mathematical models to the nervous system?
Nicolas Rashevsky (1936, 1938).
What are the main parts of a neuron?
What is the typical length of an axon?
1 cm, but can reach up to 1 meter.
How do neurons communicate?
Through synapses via electrochemical reactions.
Where does most information processing occur in the brain?
In the cerebral cortex.
What is the basic organizational unit in the cerebral cortex?
A column of tissue about 0.5 mm in diameter containing about 20,000 neurons.
What can change in the brain over a few weeks?
Mappings between areas of the brain and body parts they control.
Who invented the electroencephalograph (EEG)?
Hans Berger (1929)
What imaging technique provides detailed images of brain activity?
Functional magnetic resonance imaging (fMRI)
What allows mapping of neuronal input-output relationships?
Single-cell recording
What is the comparison of computational units between supercomputers, PCs, and the human brain?
What is the cycle time of a supercomputer compared to the human brain?
Supercomputer: 10^{-9} sec Human Brain: 10^{-3} sec
Who are the key figures in the origins of scientific psychology?
What was Helmholtz's significant contribution to psychology?
Handbook of Physiological Optics
When did Wundt open the first laboratory of experimental psychology?
1879
What was Wundt's method for making psychology a science?
Carefully controlled experiments
What is the challenge in disconfirming theories in psychology?
Subjective nature of data
What is the mystical alternative theory to understanding minds?
Mysticism
What is the main difference in properties between brains and digital computers?
Cycle time and storage capacity
What do futurists predict about the future of computers?
Approaching singularity
What did behaviorism reject?
Any theory involving mental processes
Who led the behaviorism movement?
John Watson (1878–1958)
What was the focus of behaviorism?
Studying only objective measures of stimuli and responses
Who is associated with cognitive psychology?
William James (1842–1910)
What did Helmholtz insist about perception?
It involved unconscious logical inference
What key steps did Craik identify for a knowledge-based agent?
What did Craik argue about mental terms?
They are as scientific as physical terms
Who continued Craik's work after his death?
Donald Broadbent
What was the title of Broadbent's influential book?
Perception and Communication (1958)
What workshop is considered the start of cognitive science?
A workshop at MIT in September 1956
What papers were presented at the MIT workshop?
What is a common view among psychologists regarding cognitive theories?
They should describe a detailed information-processing mechanism like a computer program
What are the two essential components for AI to succeed?
Intelligence and an artifact
What artifact has been the choice for AI development?
The modern digital electronic computer
What was the first operational computer?
The Heath Robinson, built in 1940 by Alan Turing’s team.
What was the purpose of the Heath Robinson?
To decipher German messages during World War II.
What was developed in 1943 by Turing’s team?
The Colossus, a powerful general-purpose machine.
Who invented the first operational programmable computer?
Konrad Zuse with the Z-3 in 1941.
What did Zuse invent besides the Z-3?
Floating-point numbers and the first high-level programming language, Plankalkül.
Who assembled the first electronic computer?
John Atanasoff and Clifford Berry between 1940 and 1942.
What was the ENIAC?
A secret military project at the University of Pennsylvania, influential for modern computers.
What trend occurred in computer hardware after 2005?
Manufacturers started multiplying the number of CPU cores instead of increasing clock speed.
Who devised the first programmable machine?
Joseph Marie Jacquard in 1805, using punched cards.
What was the purpose of Babbage’s Difference Engine?
To compute mathematical tables for engineering and scientific projects.
What was the Analytical Engine?
Babbage's machine with addressable memory and stored programs, capable of universal computation.
Who is considered the first programmer?
Ada Lovelace, who wrote programs for Babbage's Analytical Engine.
What programming language is named after Ada Lovelace?
The programming language Ada.
What contributions has AI made to computer science?
Ideas like time sharing, interactive interpreters, and object-oriented programming.
Who built the first automated machines?
Machines dating from the 17th century.
What was Turing's interest in computers post-World War II?
Using them for AI research, including chess programs.
What is control theory concerned with?
How artifacts operate under their own control.
Who built the first self-controlling machine?
Ktesibios of Alexandria (c. 250 B.C.)
What was the first self-controlling machine?
A water clock with a regulator.
Who created the steam engine governor?
James Watt (1736–1819)
Who invented the thermostat?
Cornelis Drebbel (1572–1633)
What did Norbert Wiener contribute to?
Control theory and its connection to cognition.
What book did Norbert Wiener publish in 1948?
Cybernetics
What concept did Wiener and his colleagues challenge?
Behaviorist orthodoxy
What did Ashby’s Design for a Brain elaborate on?
Intelligence created by homeostatic devices.
What is the goal of modern control theory?
Design systems that maximize an objective function over time.
Who published Verbal Behavior in 1957?
B. F. Skinner
Who wrote a famous review of Skinner's Verbal Behavior?
Noam Chomsky
What did Chomsky's Syntactic Structures address?
The creativity in language.
Who pointed out the limitations of the behaviorist theory in language?
Noam Chomsky
What notion did the behaviorist theory fail to address?
Creativity in language
Whose syntactic models influenced Chomsky's theory?
Panini (c. 350 B.C.)
What hybrid field emerged from modern linguistics and AI?
Computational linguistics or natural language processing
What is required for understanding language beyond sentence structure?
Understanding of subject matter and context
What study involves putting knowledge into a computable form?
Knowledge representation
Who is recognized for the first work in AI (1943)?
Warren McCulloch and Walter Pitts
What three sources influenced McCulloch and Pitts' work?
What characterizes the state of a neuron in their model?
On or Off
What learning rule did Donald Hebb propose?
Hebbian learning
Who built the first neural network computer in 1950?
Marvin Minsky and Dean Edmonds
What was the name of the first neural network computer?
SNARC
Who articulated the Turing Test and related concepts?
Alan Turing
What did Turing propose instead of simulating the adult mind?
To simulate the child's mind
In what year is the birth of artificial intelligence recognized?
1956
Who is considered an influential figure in AI and moved to Dartmouth College?
John McCarthy
What year is known as the birth of artificial intelligence?
1956
What was the purpose of the Dartmouth workshop in 1956?
To study artificial intelligence and simulate aspects of learning and intelligence.
How long did the Dartmouth workshop last?
2 months
How many attendees were at the Dartmouth workshop?
10 attendees
Which two researchers from Carnegie Tech gained significant attention during the workshop?
Allen Newell and Herbert Simon
What program did Newell and Simon create that was capable of non-numerical thinking?
Logic Theorist (LT)
What language did Newell and Simon invent to write LT?
IPL (Information Processing Language)
What theorem did the Logic Theorist prove that impressed Russell?
A theorem from Principia Mathematica
Why did AI need to become a separate field?
It aimed to duplicate human faculties like creativity and self-improvement.
What was the official proposal for the Dartmouth workshop's conjecture?
Every aspect of learning can be described for machine simulation.
What major figures were introduced to each other at the Dartmouth workshop?
John McCarthy, Minsky, Claude Shannon, Nathaniel Rochester, Newell, Simon, and others.
What is a key objective of AI?
To duplicate human faculties such as creativity, self-improvement, and language use.
Why isn't AI considered a branch of mathematics?
AI embraces the idea of duplicating human faculties, which other fields do not address.
What methodology distinguishes AI from other fields?
AI is a branch of computer science and attempts to build autonomous machines in complex environments.
What period is referred to as the 'Look, Ma, no hands!' era?
The early years of AI from 1952 to 1969.
What was the General Problem Solver (GPS)?
A program designed to imitate human problem-solving protocols.
What does the physical symbol system hypothesis state?
Any intelligent system must manipulate data structures composed of symbols.
Who constructed the Geometry Theorem Prover?
Herbert Gelernter in 1959.
What did Arthur Samuel's checkers program demonstrate?
It learned to play at a strong amateur level, disproving that computers can only do what they are told.
What significant contributions did John McCarthy make in 1958?
Defined the Lisp programming language, invented time sharing, and described the Advice Taker program.
What is Lisp?
A high-level programming language defined by John McCarthy, dominant in AI for 30 years.
What was the Advice Taker?
A hypothetical program described by McCarthy as the first complete AI system.
What did John McCarthy describe in his 1958 paper?
The Advice Taker, a hypothetical program seen as the first complete AI system.
What was the primary function of the Advice Taker?
To use knowledge to search for solutions to problems and embody general knowledge of the world.
How could the Advice Taker improve its competence?
By accepting new axioms during operation without needing reprogramming.
What central principles did the Advice Taker embody?
Knowledge representation and reasoning.
What year did Marvin Minsky move to MIT?
1958.
What was McCarthy's focus compared to Minsky's?
McCarthy focused on representation and reasoning, while Minsky focused on getting programs to work.
What significant discovery did J. A. Robinson make in 1965?
The resolution method, a complete theorem-proving algorithm for first-order logic.
What did the Shakey robotics project demonstrate?
The integration of logical reasoning and physical activity.
What are microworlds in AI?
Limited domains that require intelligence to solve problems.
What problem could Slagle’s SAINT program solve?
Closed-form calculus integration problems.
What type of problems did Evans’s ANALOGY program solve?
Geometric analogy problems found in IQ tests.
What kind of problems did Bobrow’s STUDENT program address?
Algebra story problems.
What is the blocks world in AI?
A microworld with solid blocks for tasks like rearranging using a robot hand.
Who worked on the vision project in the blocks world?
David Huffman (1971).
What did Winograd and Cowan's work demonstrate?
A large number of elements can represent an individual concept, increasing robustness and parallelism.
What are adalines in AI?
Networks enhanced by Bernie Widrow, based on Hebb's learning methods.
What did Frank Rosenblatt develop in 1962?
Perceptrons.
Who enhanced learning methods in AI?
What does the perceptron convergence theorem state?
The learning algorithm can adjust connection strengths to match input data, if such a match exists.
What did Herbert Simon predict in 1957?
Machines that think, learn, and create will increase rapidly in capability.
What was Simon's prediction about computers within 10 years?
A computer would be a chess champion and prove a significant mathematical theorem.
What was a major issue with early AI programs?
They often knew nothing of their subject matter and relied on simple syntactic manipulations.
What problem arose in early machine translation efforts?
Accurate translation requires background knowledge to resolve ambiguity.
What was the outcome of the 1966 report on machine translation?
No machine translation of general scientific text was achieved; funding was canceled.
What was a limitation of early AI problem-solving methods?
They relied on trying different combinations of steps, which failed with larger problems.
What illusion did early AI researchers have about computational power?
They believed scaling up problems was just a matter of faster hardware and larger memories.
What was the issue with resolution theorem proving in AI?
Researchers struggled to prove theorems involving more than a few dozen facts.
What is the illusion of unlimited computational power in AI?
It refers to the belief that problem-solving programs can solve any problem effectively, which has proven to be incorrect.
What are genetic algorithms?
They are methods based on making small mutations to machine-code programs to generate effective solutions for specific tasks.
What was a major criticism of AI in the Lighthill report?
It highlighted the failure to address the 'combinatorial explosion' in AI research, leading to reduced funding.
Who authored the book Perceptrons?
Marvin Minsky and Seymour Papert in 1969.
What limitation did Perceptrons have?
A two-input perceptron could not be trained to recognize when its two inputs were different.
What are weak methods in AI?
General-purpose search mechanisms that struggle with large or complex problem instances.
What is the DENDRAL program?
An early AI program developed to infer molecular structures from mass spectrometry data.
Who were the key contributors to the DENDRAL program?
Ed Feigenbaum, Bruce Buchanan, and Joshua Lederberg.
What input does the DENDRAL program require?
Elementary formula of the molecule and the mass spectrum from a mass spectrometer.
What is a challenge faced by the DENDRAL program?
Generating all possible structures consistent with the formula is intractable for moderate-sized molecules.
What was the purpose of DENDRAL?
To identify molecular structures using spectral analysis.
What rule is used to recognize a ketone subgroup?
If x1 + x2 = M + 28, x1 - 28 is high, x2 - 28 is high, and at least one of x1 or x2 is high.
What was significant about DENDRAL in AI history?
It was the first successful knowledge-intensive system.
What project did Feigenbaum and others start after DENDRAL?
The Heuristic Programming Project (HPP).
What system was developed for medical diagnosis?
MYCIN.
How many rules did MYCIN use?
About 450 rules.
What was a major difference between MYCIN and DENDRAL?
MYCIN's rules were acquired from expert interviews, not derived from a theoretical model.
What did MYCIN incorporate to handle uncertainty?
A calculus of uncertainty called certainty factors.
What was Winograd's SHRDLU system designed for?
Understanding natural language in the blocks world.
What did Roger Schank emphasize about language understanding?
Robust understanding requires general knowledge about the world.
What was the focus of Schank's programs?
Representing and reasoning with knowledge for language understanding.
What are the problems related to language understanding?
What led to the growth of knowledge representation schemes?
The increase in applications to real-world problems.
Which logic-based language became popular in Europe?
Prolog
What is the PLANNER family?
A family of languages developed in the United States based on logic.
What did Minsky propose in 1975?
The idea of frames for knowledge representation.
What was the first successful commercial expert system?
R1 at Digital Equipment Corporation.
How much did R1 save DEC annually by 1986?
$40 million
How many expert systems did DuPont have in use by 1988?
100 expert systems
What was the Fifth Generation project?
A 10-year plan announced by Japan to build intelligent computers running Prolog.
What was the purpose of the Microelectronics and Computer Technology Corporation (MCC)?
To assure national competitiveness in AI research.
What was the AI Winter?
A period where many AI companies failed to deliver on promises.
What algorithm was reinvented in the mid-1980s?
Back-propagation learning algorithm
What did the collection Parallel Distributed Processing cause?
Great excitement in the field of AI and psychology.
What are connectionist models seen as competitors to?
Symbolic models and logicist approaches.
What does Terrence Deacon's book suggest about humans?
Symbol manipulation is a defining characteristic of humans.
What is the current view on connectionist and symbolic approaches?
They are complementary, not competing.
What is the current view of connectionist and symbolic approaches in AI?
They are considered complementary, not competing.
What are the two fields modern neural network research has bifurcated into?
What methodology has AI adopted since 1987?
AI has embraced the scientific method, emphasizing rigorous empirical experiments and statistical analysis.
What has changed in AI's approach to theories?
It is more common to build on existing theories rather than propose brand-new ones.
How does current AI research differ from earlier periods?
It is now based on rigorous theorems and hard experimental evidence, showing relevance to real-world applications.
What did David McAllester say about AI's early isolationism?
AI was largely separated from other fields like computer science, but this isolationism is being abandoned.
What is the significance of hidden Markov models (HMMs) in speech recognition?
HMMs are based on rigorous mathematical theory and trained on large real speech data, ensuring robust performance.
What is the trend in the field of speech technology?
It is transitioning to widespread industrial and consumer applications.
What characterizes the change in AI methodology?
A victory of the neats over the scruffies, implying a shift towards mathematical rigor and stability.
What parallels exist between machine translation and speech recognition?
Both have evolved from initial enthusiasm for simplistic approaches to more rigorous, data-driven methods.
What was the initial approach to machine translation in the 1950s?
Based on sequences of words, with models learned according to information theory principles.
When did the approach to machine translation return in popularity?
In the late 1990s.
What was the focus of neural networks in the 1980s?
To explore capabilities and differences from traditional techniques.
What has the work on neural networks allowed for?
Comparison with techniques from statistics, pattern recognition, and machine learning.
What industry has emerged due to developments in data mining technology?
A vigorous new industry.
What did Judea Pearl’s 1988 work contribute to AI?
New acceptance of probability and decision theory in AI.
What formalism was invented for efficient representation of uncertain knowledge?
Bayesian network.
What major problem does the Bayesian network approach overcome?
Problems of probabilistic reasoning systems from the 1960s and 1970s.
What do normative expert systems act according to?
Laws of decision theory.
What areas have seen revolutions similar to AI?
Robotics, computer vision, and knowledge representation.
What has led to significant benefits in AI recently?
Reintegration of isolated subfields and tools from machine learning.
Who worked on the SOAR agent architecture?
Allen Newell, John Laird, and Paul Rosenbloom.
What is a significant environment for intelligent agents?
The Internet.
What suffix has entered everyday language due to AI systems?
'-bot'.
What must reasoning and planning systems handle due to unreliable sensory systems?
Uncertainty.
What must reasoning and planning systems handle?
Uncertainty
Which fields has AI come into closer contact with?
Control theory, Economics
What has recent progress in robotic cars derived from?
Better sensors, control-theoretic integration, high-level planning
What do influential AI founders express discontent with?
Progress of AI
What do they believe AI should strive for?
Machines that think, learn, and create
What is the term for the effort to achieve human-level AI?
Human-level AI (HLAI)
What is needed for human-level AI?
Very large knowledge bases
What subfield looks for a universal algorithm for learning?
Artificial General Intelligence (AGI)
What does AGI aim to achieve?
Universal algorithm for any environment
What is a concern in AI development?
Creating Friendly AI
What has been the main subject of study in computer science?
Algorithms
What is increasingly available that affects AI development?
Very large data sets
What did Yarowsky's work focus on?
Word-sense disambiguation
What accuracy did Yarowsky achieve without labeled examples?
Above 96%
What do Banko and Brill show about using more data?
Performance increases exceed algorithm choice differences
What problem do Hays and Efros discuss?
Filling in holes in photographs
What is the purpose of the algorithm defined by Hays and Efros?
To find something that matches the background in a photograph after masking out an object.
What was the performance of Hays and Efros' algorithm with 10,000 photos?
Poor performance.
What is the performance of Hays and Efros' algorithm with 2 million photos?
Excellent performance.
What does the 'knowledge bottleneck' in AI refer to?
The problem of expressing all the knowledge a system needs.
What is suggested to solve the 'knowledge bottleneck' in many applications?
Learning methods rather than hand-coded knowledge engineering.
What did reporters say about 'AI Winter'?
It may be yielding to a new Spring.
What did Kurzweil say about AI applications?
Many thousands of AI applications are deeply embedded in industry infrastructure.
What is STANLEY?
A driverless robotic car that won the 2005 DARPA Grand Challenge.
What technology does STANLEY use to sense its environment?
Cameras, radar, and laser rangefinders.
What was CMU's BOSS?
A robotic vehicle that won the Urban Challenge by safely driving in traffic.
What does automated speech recognition do for travelers?
Guides the entire conversation when booking a flight.
What was NASA’s Remote Agent program?
The first on-board autonomous planning program for spacecraft.
What did REMOTE AGENT do?
Generated plans from high-level goals and monitored their execution.
What is MAPGEN?
A program that plans daily operations for NASA’s Mars Exploration Rovers.
What achievement is IBM’s DEEP BLUE known for?
Defeating world champion Garry Kasparov in chess.
What was the impact of DEEP BLUE's victory on IBM's stock?
Increased by $18 billion.
How do learning algorithms help in spam fighting?
They classify over a billion messages as spam daily.
What challenge do spammers present to static programmed approaches?
Spammers continually update their tactics, making it hard to keep up.
What was significant about logistics planning during the Persian Gulf crisis of 1991?
U.S. forces deployed a Dynamic Analysis for effective logistics.
What tool did U.S. forces use for logistics planning during the Persian Gulf crisis of 1991?
Dynamic Analysis and Replanning Tool (DART)
How many vehicles, cargo, and people did DART account for at a time?
Up to 50,000
What did DARPA state about the investment in AI for DART?
It paid back DARPA’s 30-year investment in AI.
How many Roomba robotic vacuum cleaners has iRobot sold?
Over two million
What is the purpose of the PackBot used by iRobot?
Handle hazardous materials, clear explosives, identify snipers.
What does the machine translation program automatically translate?
From Arabic to English
What statistical model is used in the machine translation program?
Built from examples of Arabic-to-English translations and English text totaling two trillion words.
What is the main concern of intelligence according to this book?
Rational action
What philosophical idea made AI conceivable?
The mind is like a machine that operates on knowledge encoded in an internal language.
What did mathematicians provide for AI development?
Tools to manipulate logical and probabilistic statements.
How do economists contribute to AI?
By formalizing decision-making to maximize expected outcomes.
What insights have neuroscientists provided for AI?
Facts about how the brain works and its similarities/differences from computers.
What perspective do psychologists adopt regarding humans and animals?
They consider them as information-processing machines.
What role do computer engineers play in AI?
They provide powerful machines that enable AI applications.
What does control theory deal with in relation to AI?
Designing devices that act optimally based on environmental feedback.
What has the history of AI experienced over time?
Cycles of success, optimism, cutbacks, and new creative approaches.
What has accelerated the advancement of AI in the past decade?
Greater use of scientific methods.
What has advanced AI rapidly in the past decade?
Greater use of the scientific method in experimenting and comparing approaches.
What has improved alongside the understanding of intelligence?
Improvements in the capabilities of real systems.
Which book discusses the methodological status of AI?
The Sciences of the Artificial by Herb Simon (1981).
What does Herb Simon's book explain about AI?
AI can be viewed as both science and mathematics.
Who criticized the usefulness of the Turing Test?
Shieber (1994) criticized its instantiation in the Loebner Prize competition.
What is the title of John Haugeland's book on AI?
Artificial Intelligence: The Very Idea (1985).
What does the Encyclopedia of AI contain?
Survey articles on almost every topic in AI.
What are the major AI conferences?
What are the major journals for general AI?
What are the main professional societies for AI?
What is the purpose of the exercises in the text?
To stimulate discussion and possibly set as term projects.
What does the exercise 1.1 ask for?
Define in your own words: intelligence, artificial intelligence, agent, rationality, logical reasoning.
What prediction did Turing make about AI by the year 2000?
A computer would have a 30% chance of passing a five-minute Turing Test with an unskilled interrogator.
What does exercise 1.3 question about reflex actions?
Are reflex actions rational? Are they intelligent?
What is suggested to extend in exercise 1.4?
Extend Evans’s ANALOGY program to score.
What are reflex actions?
Automatic responses to stimuli, like flinching from a hot stove.
Are reflex actions rational or intelligent?
They are not considered rational or intelligent; they are instinctive.
What happens if a program scores 200 on an IQ test?
It doesn't necessarily mean the program is more intelligent than a human.
Why is the sea slug Aplysia studied?
It has about 20,000 neurons that are large and easily manipulated.
What is the significance of neuron cycle time in Aplysia?
Assuming it's similar to humans, it allows for computational comparisons.
How can introspection be inaccurate?
Individuals may misreport their inner thoughts or be unaware of them.
What are examples of artificial intelligence systems?
What complex operations do cognitive models involve?
Operations like convolving an image with a Gaussian or minimizing entropy.
Why does evolution favor rational systems?
To achieve survival and reproductive goals effectively.
Is AI a science, engineering, or both?
AI encompasses elements of both science and engineering.
Can computers be intelligent?
They can only perform tasks as programmed, raising questions about true intelligence.
Can animals be intelligent?
They operate based on genetic programming, but this doesn't negate intelligence.
What can be said about intelligence and physical laws?
All entities act according to physical laws, but this does not define intelligence.
What tasks can computers currently perform?
What is the purpose of AI contests?
To define tasks and push advancements in AI research.
What should be examined regarding AI contests?
Progress made and their impact on the state of AI and new ideas.
What is the central concept of artificial intelligence discussed in Chapter 1?
Rational agents
What can rationality be applied to?
A wide variety of agents in any environment
What influences how well an agent can behave?
The nature of the environment
What is an agent?
Anything that perceives its environment through sensors and acts through actuators
What do we refer to the agent's perceptual inputs as?
Percept
What is the complete history of everything an agent has ever perceived?
Percept sequence
How is an agent's behavior mathematically described?
By the agent function that maps percept sequences to actions
What is the difference between agent function and agent program?
Agent function is an abstract mathematical description; agent program is a concrete implementation
What is an example used to illustrate agent concepts?
The vacuum-cleaner world
What is the vacuum-cleaner world described in the text?
A simple world with two locations (squares A and B) where a vacuum agent can perceive its location and dirt status.
What actions can the vacuum agent choose from?
What is a simple agent function for the vacuum agent?
If the current square is dirty, then suck; otherwise, move to the other square.
What does a rational agent do?
A rational agent does the right thing according to its environment and percepts.
How is good behavior defined for an agent?
By the desirability of the sequence of actions it generates in its environment.
What is a performance measure?
It evaluates the desirability of a sequence of environment states based on the agent's actions.
What can affect an agent's perception of its own performance?
An agent might delude itself about its performance leading to a flawed self-assessment.
What is the main focus of AI according to the text?
Designing artifacts with significant computational resources for nontrivial decision-making.
What is a performance measure in the context of agents?
A metric designed by a designer to evaluate the success of an agent's actions in specific circumstances.
How might a vacuum-cleaner agent maximize its performance measure?
By cleaning dirt, dumping it on the floor, and cleaning it again.
What is a more suitable performance measure for a vacuum-cleaner agent?
Reward the agent for each clean square and penalize for electricity and noise.
What is the definition of a rational agent?
An agent that selects actions to maximize its performance measure based on percept sequences and built-in knowledge.
What four factors determine what is rational at any time?
What is the performance measure for the vacuum-cleaner agent in the example?
One point for each clean square at each time step over 1000 time steps.
What does the vacuum-cleaner agent do if it finds a dirty square?
Cleans the square and moves to the other square if it's clean.
Under what condition is the vacuum-cleaner agent considered rational?
When its expected performance is at least as high as any other agent's under given circumstances.
What happens to the vacuum-cleaner agent when all dirt is cleaned?
It oscillates back and forth needlessly.
What penalty might affect the performance measure of the vacuum-cleaner agent?
A penalty of one point for each movement left or right.
What happens when all the dirt is cleaned up by the agent?
The agent will oscillate needlessly back and forth.
What should an agent do if clean squares can become dirty again?
The agent should occasionally check and re-clean them if needed.
What does an omniscient agent know?
The actual outcome of its actions.
Is omniscience possible in reality?
No, omniscience is impossible in reality.
What is the difference between rationality and perfection?
Rationality maximizes expected performance; perfection maximizes actual performance.
What does rationality depend on?
The percept sequence to date.
What is a key aspect of rational behavior when crossing a busy road?
An agent should look both ways before crossing.
What is information gathering in the context of rationality?
Actions taken to modify future percepts.
What must a rational agent do according to the definition?
Gather information and learn from what it perceives.
What happens if an environment is completely known a priori?
The agent need not perceive or learn; it simply acts correctly.
What is an example of a fragile agent mentioned in the text?
The dung beetle.
What does the dung beetle do after laying its eggs?
Fetches a ball of dung to plug the entrance.
What happens if the dung ball is removed from the dung beetle?
It continues pantomiming plugging the nest without noticing the dung ball is missing.
What is the behavior of the sphex wasp during its nesting process?
Digs a burrow, stings a caterpillar, drags it inside, and lays eggs.
What does the sphex wasp do if the caterpillar is moved?
Reverts to the dragging step and continues without modification.
What does it mean for an agent to lack autonomy?
It relies on prior knowledge of its designer rather than its own percepts.
What is an example of a rational agent that can learn?
A vacuum-cleaning agent that learns to foresee where dirt will appear.
What is the PEAS acronym in AI?
Performance, Environment, Actuators, Sensors.
Why is the automated taxi driver considered a complex problem?
It has open-ended tasks with novel combinations of circumstances.
What must be specified when designing an agent?
The task environment, including performance measure, environment, actuators, and sensors.
What does PEAS stand for in the context of an automated taxi?
Performance Measure, Environment, Actuators, Sensors
What are the performance measures for the automated taxi?
What types of environments does the taxi driver face?
What are the actuators used by an automated taxi?
What are the basic sensors for an automated taxi?
What are some desirable qualities for the automated taxi's performance?
What optional choices may affect the taxi's driving environment?
What is the significance of the complexity of the relationship in agent environments?
It affects the agent's behavior, percept sequence, and performance measure.
What distinguishes real environments from artificial ones in AI?
The complexity of the relationship among agent behavior, percept sequence, and performance measure.
What is an example of a simple 'real' environment for a robot?
A robot inspecting parts on a conveyor belt.
What are the communication outputs needed for an automated taxi?
What are the two possible actions for software agents?
What is a softbot designed to do?
What abilities does a softbot need to operate effectively?
What is the complexity of the Internet compared to?
The physical world
What dimensions categorize task environments in AI?
What is a fully observable environment?
An environment where sensors access the complete state at all times.
What causes a task environment to be partially observable?
What is an unobservable environment?
An environment where the agent has no sensors at all.
What is the difference between single-agent and multiagent environments?
What is an example of a single-agent environment?
Solving a crossword puzzle by itself.
What is an example of a multiagent environment?
Playing chess against another player.
What is a single-agent environment?
An environment where an agent, like one solving a crossword puzzle, operates independently.
What is a two-agent environment?
An environment where two agents interact, such as in a game of chess.
What distinguishes an agent from an object?
An agent's behavior maximizes a performance measure based on another entity's behavior.
What is a competitive multiagent environment?
An environment where agents, like in chess, compete to maximize their performance measures at the expense of others.
What is a partially cooperative multiagent environment?
An environment where agents work together to avoid collisions, like in taxi driving, while also competing for resources.
What defines a deterministic environment?
An environment where the next state is completely determined by the current state and the agent's action.
What defines a stochastic environment?
An environment where the next state cannot be precisely predicted due to uncertainty.
What is an uncertain environment?
An environment that is either not fully observable or not deterministic.
What is a nondeterministic environment?
An environment characterized by possible outcomes of actions without attached probabilities.
What is an episodic task environment?
An environment where the agent's experience is divided into episodes, with each episode independent of others.
What is a sequential task environment?
An environment where the agent's actions in previous episodes affect future episodes.
What is the difference between sequential and episodic environments?
In sequential environments, current decisions affect future decisions. In episodic environments, decisions are independent of past actions.
What defines a static environment?
A static environment does not change while an agent is deliberating, allowing the agent to focus on decision-making without external changes.
What is a dynamic environment?
A dynamic environment changes while an agent is deliberating, requiring continuous decision-making.
What is a semidynamic environment?
A semidynamic environment does not change over time, but the agent's performance score may change.
How do discrete and continuous environments differ?
Discrete environments have a finite number of states, while continuous environments have states that vary smoothly over time.
What characterizes a known environment?
In a known environment, the outcomes for all actions are known, allowing the agent to make informed decisions.
What is an unknown environment?
An unknown environment requires the agent to learn how it works, as the outcomes of actions are not fully understood.
What is the hardest type of environment?
The hardest environment is partially observable, multiagent, stochastic, sequential, dynamic, continuous, and unknown.
How does taxi driving exemplify a dynamic environment?
Taxi driving is dynamic because the environment changes continuously with moving vehicles and requires ongoing decision-making.
What is an example of a static environment?
Crossword puzzles are considered static as they do not change once the puzzle is set.
What is the antonym of parallel in computer science?
ial
What type of agents are involved in a crossword puzzle task environment?
Fully Single Deterministic Sequential Static Discrete
What type of agents are involved in chess with a clock?
Fully Multi Deterministic Sequential Semi Discrete
What type of agents are involved in poker?
Partially Multi Stochastic Sequential Static Discrete
What type of agents are involved in backgammon?
Fully Multi Stochastic Sequential Static Discrete
What type of agents are involved in taxi driving?
Partially Multi Stochastic Sequential Dynamic Continuous
What type of agents are involved in medical diagnosis?
Partially Single Stochastic Sequential Dynamic Continuous
What type of agents are involved in image analysis?
Fully Single Deterministic Episodic Semi Continuous
What type of agents are involved in a part-picking robot?
Partially Single Stochastic Episodic Dynamic Continuous
What type of agents are involved in a refinery controller?
Partially Single Stochastic Sequential Dynamic Continuous
What type of agents are involved in an interactive English tutor?
Partially Multi Stochastic Sequential Dynamic Discrete
What is the task environment for medical diagnosis considered?
Single-agent
What is the nature of environments in chess tournaments?
Episodic
What does the code repository include for environments?
Implementations and a general-purpose environment simulator
What is the purpose of running many simulations in different traffic conditions?
To evaluate a taxi driver in simulated traffic
What does the environment generator do?
Selects particular environments for running the agent
What is the average performance of an agent measured over?
The environment class
What does a rational agent maximize in an environment?
Average performance over the environment class.
What is the formula for an agent?
agent = architecture + program.
What does an agent program implement?
The agent function—mapping from percepts to actions.
What does the architecture provide to the program?
Percepts from sensors, runs the program, feeds actions to actuators.
What is the difference between agent program and agent function?
Agent program takes current percept; agent function takes entire percept history.
What is the purpose of the TABLE-DRIVEN-AGENT program?
To track percept sequence and decide actions based on a table of actions.
Why is the table-driven approach to agent construction problematic?
It requires a massive lookup table for all possible percept sequences, which is impractical.
What is the persistent variable in the TABLE-DRIVEN-AGENT program?
percepts, a sequence initially empty.
What does the action variable in the TABLE-DRIVEN-AGENT program represent?
The action returned based on the lookup from the percept sequence table.
What is the data rate from a single camera in an automated taxi?
27 megabytes per second
How many entries does the lookup table have for an hour’s driving?
Over 10^50,000,000,000 entries
What is the number of entries for a chess lookup table?
At least 10^150 entries
What is the number of atoms in the observable universe?
Less than 10^80
What does the TABLE-DRIVEN-AGENT implement?
The desired agent function
What is the key challenge for AI?
To produce rational behavior from a small program rather than a vast table
What replaced huge tables of square roots before the 1970s?
A five-line program for Newton’s method
What are the four basic kinds of agent programs?
What does a simple reflex agent base its actions on?
The current percept, ignoring the rest of the percept history
What is the action of the vacuum agent when the status is dirty?
Return Suck
What happens when the car in front brakes?
The driver should notice and initiate braking
What initiates the action 'initiate braking'?
If car-in-front-is-braking then initiate-braking.
What are condition–action rules also known as?
They are also called situation–action rules, productions, or if–then rules.
What is the function of SIMPLE-REFLEX-AGENT?
It acts according to a rule whose condition matches the current state from the percept.
What does the INTERPRET-INPUT function do?
It generates an abstracted description of the current state from the percept.
What is the limitation of simple reflex agents?
They are of limited intelligence and work only if the environment is fully observable.
What happens if the car in front has different taillight configurations?
A simple reflex agent may either brake unnecessarily or fail to brake at all.
What are the possible percepts for a simple reflex vacuum agent with only a dirt sensor?
The percepts are [Dirty] and [Clean].
What happens if a vacuum agent starts in square A and tries to move Left?
Moving Left fails forever if it starts in square A.
What happens if a vacuum agent starts in square B and tries to move Right?
Moving Right fails forever if it starts in square B.
What happens if a simple reflex agent starts in square A and receives [Clean]?
Moving Left fails (forever).
What happens if a simple reflex agent starts in square B and receives [Clean]?
Moving Right fails (forever).
How can escape from infinite loops be achieved in simple reflex agents?
By randomizing actions, such as flipping a coin to choose between Left and Right.
What is the average number of steps for a randomized agent to reach the other square?
Two steps.
What is the main benefit of a randomized simple reflex agent over a deterministic one?
It might outperform a deterministic simple reflex agent in some situations.
What is a model-based reflex agent?
An agent that keeps track of the part of the world it can’t see using an internal state.
What does the internal state of a model-based reflex agent depend on?
The percept history reflecting unobserved aspects of the current state.
What is an example of internal state for a braking problem?
The previous frame from the camera detecting red lights.
What knowledge is required for updating the internal state of a model-based agent?
Knowledge of how the world evolves and how the agent's actions affect the world.
What is a model of the world in the context of AI agents?
Knowledge about how the world works, used by model-based agents.
What does the function UPDATE-STATE do in a model-based reflex agent?
It updates the agent's internal state based on current percept and action.
What is the purpose of the MODEL-BASED-REFLEX-AGENT function?
To return an action based on the agent's current state and rules.
What is the purpose of the RULE-MATCH function?
To match the current state with rules and determine the action to take.
What does a model-based reflex agent do?
Keeps track of the current state of the world and chooses actions based on that state.
What is a key characteristic of model-based agents?
They maintain an internal model of the world to guide decision-making.
What is often unavoidable in partially observable environments?
Uncertainty about the current state of the environment.
What does the box labeled 'what the world is like now' represent?
The agent's best guess about the current state of the environment.
What is an example of uncertainty in an automated taxi?
Not being able to see around a large truck blocking the view.
What does a goal-based agent track?
The world state and a set of goals it aims to achieve.
How does a goal-based agent choose its actions?
By selecting actions that will lead to the achievement of its goals.
What aspect of the world state might a taxi consider when deciding its actions?
Its destination, which is part of the agent's internal state.
Who are some contributing writers?
What does the book aim to cover?
The breadth of AI, including logic, probability, perception, reasoning, learning, and action.
What are some important applications of AI technology?
What theoretical progress areas are mentioned in AI?
What new emphasis is placed in the book regarding environments?
Partially observable and nondeterministic environments
What types of representations are discussed for agents?
What is covered in more depth regarding planning?
Contingent planning in partially observable environments
What new material was added on probabilistic models?
First-order probabilistic models and open-universe models
What is the role of learning in AI according to the book?
Extending the reach of the designer into unknown environments
What is the prerequisite knowledge for the book?
Familiarity with basic computer science concepts at a sophomore level
What does the cover of the book depict?
Final position from game 6 of the 1997 match between Garry Kasparov and DEEP BLUE
What does the cover feature besides Kasparov?
Asimo robot, Thomas Bayes, Mars Rover, Alan Turing, Shakey robot, Aristotle
What is depicted behind the title of the book?
A portion of the CPSC Bayesian network for medical diagnosis
What is the purpose of the book's website?
Contains algorithms, course materials, links, and discussion group info
Who reviewed the manuscript?
Who are Stuart's family members mentioned?
Who are Peter's family members mentioned?
Which libraries and developers are thanked for their contributions?
Who provided helpful comments on the book?
Who is Stuart Russell?
What awards has Stuart Russell received?
What are some works by Stuart Russell?
Who is Peter Norvig?
What was Peter Norvig's previous position?
Who was the head of the Computational Sciences Division at NASA Ames Research Center?
He oversaw NASA’s research in artificial intelligence and robotics.
What is one of the first Internet information extraction services?
Developed by Junglee, where he was chief scientist.
What degrees did he receive from Brown University and UC Berkeley?
B.S. in applied mathematics and Ph.D. in computer science.
Which awards did he receive from UC Berkeley?
Distinguished Alumni and Engineering Innovation awards.
What are the titles of his other books?
What is the definition of Constraint Satisfaction Problems?
Section 6.1: Defining Constraint Satisfaction Problems
What is discussed in the section on the structure of problems?
Section 6.5: The Structure of Problems
What does the summary section cover in CSPs?
Section 6.6: Summary, Bibliographical and Historical Notes, Exercises
What is covered in the summary section for logical agents?
Section 7.8: Summary, Bibliographical and Historical Notes, Exercises
What are the syntax and semantics of first-order logic?
Section 8.2: Syntax and Semantics of First-Order Logic
What is knowledge engineering in first-order logic?
Section 8.4: Knowledge Engineering in First-Order Logic
What does the summary section cover in first-order logic?
Section 8.5: Summary, Bibliographical and Historical Notes, Exercises
What is the difference between propositional and first-order inference?
Section 9.1: Propositional vs. First-Order Inference
What does the summary section cover in inference?
Section 9.6: Summary, Bibliographical and Historical Notes, Exercises
What are algorithms for planning as state-space search?
Section 10.2: Algorithms for Planning as State-Space Search
What is discussed in the analysis of planning approaches?
Section 10.5: Analysis of Planning Approaches
What does the summary section cover in classical planning?
Section 10.6: Summary, Bibliographical and Historical Notes, Exercises
What is discussed regarding planning and acting in the real world?
Section 11: Planning and Acting in the Real World
What is planning in nondeterministic domains?
Section 11.3: Planning and Acting in Nondeterministic Domains
What does the summary section cover in planning and acting?
Section 11.5: Summary, Bibliographical and Historical Notes, Exercises
What is the focus of Object Recognition by Appearance?
What is Object Recognition from Structural Information about?
What is discussed in Weak AI: Can Machines Act Intelligently??
What does Strong AI: Can Machines Really Think? explore?
What are the Ethics and Risks of Developing Artificial Intelligence?
What question does Are We Going in the Right Direction? address?
What is covered in Complexity Analysis and O() Notation?
What are Vectors, Matrices, and Linear Algebra used for?
What is Defining Languages with Backus–Naur Form (BNF)?
What does AI encompass?
A huge variety of subfields including: - Learning - Perception - Playing chess - Proving mathematical theorems - Writing poetry - Driving a car - Diagnosing diseases
What is the Turing Test?
A test to determine if a computer can mimic human responses so well that a human interrogator cannot tell the difference.
What capabilities must a computer have to pass the Turing Test?
What are the four approaches to AI?
What is the focus of a human-centered approach to AI?
An empirical science involving observations and hypotheses about human behavior.
What does the term 'computational intelligence' refer to?
The study of the design of intelligent agents.
What is the goal of AI according to Bellman (1978)?
The automation of activities associated with human thinking, such as decision-making and problem-solving.
What did Kurzweil (1990) define AI as?
The art of creating machines that perform functions requiring intelligence when performed by people.
What is the first point made about human imperfection?
Not all chess players are grandmasters; not everyone gets an A on the exam.
What does the total Turing Test include?
A video signal to test perceptual abilities and the ability to pass physical objects through a hatch.
What two capabilities are required to pass the total Turing Test?
Why do AI researchers focus on studying intelligence rather than passing the Turing Test?
They believe it is more important to study the underlying principles of intelligence.
What analogy is used to explain the quest for artificial flight?
The Wright brothers succeeded when they stopped imitating birds and focused on aerodynamics.
What are three ways to determine how humans think?
What is the goal of cognitive science?
To construct precise and testable theories of the human mind using AI models and psychology techniques.
What distinguishes modern authors in the field of AI and cognitive science?
They separate claims of algorithm performance from models of human performance.
What are syllogisms?
Patterns for argument structures that yield correct conclusions from correct premises.
What notation did 19th-century logicians develop?
Precise notation for statements about objects and relations.
What does the logicist tradition in AI aim to do?
Build intelligent systems based on logical programs.
What is a major obstacle in the logicist approach to AI?
Translating informal knowledge into formal logical terms.
How does computational reasoning relate to the logicist tradition?
It faces obstacles like computational resource exhaustion.
What defines a rational agent?
Acts to achieve the best outcome or expected outcome under uncertainty.
What skills are needed for the Turing Test?
Knowledge representation, reasoning, and natural language generation.
What advantage does the rational-agent approach have over the laws of thought?
It is more general and amenable to scientific development.
What is the standard of rationality in the rational-agent approach?
Mathematically well defined and completely general.
What is the standard of rationality?
It is mathematically well defined and can be 'unpacked' to generate agent designs that provably achieve it.
How is human behavior defined?
By the sum total of all the things that humans do, adapted for a specific environment.
What is the main focus of the book?
General principles of rational agents and components for constructing them.
What is the issue with achieving perfect rationality?
It is not feasible in complicated environments due to high computational demands.
What is the working hypothesis of the book?
Perfect rationality is a good starting point for analysis.
What did Aristotle contribute to rationality?
He formulated laws governing the rational part of the mind and developed syllogisms for reasoning.
What was Ramon Lull's idea regarding reasoning?
That useful reasoning could be carried out by a mechanical artifact.
What did Thomas Hobbes propose about reasoning?
That reasoning is like numerical computation, involving silent addition and subtraction.
What is the Pascaline?
A calculating machine built by Blaise Pascal in 1642, known for addition and subtraction.
What did Gottfried Wilhelm Leibniz build?
A calculator that could add, subtract, multiply, and take roots.
What problem arises from a purely physical conception of the mind?
It leaves little room for free will.
What is the outcome of the action discussed in the text?
The outcome is an action that results from premises, e.g., 'I need a cloak. I have to make a cloak.'
What does Aristotle suggest in Nicomachean Ethics regarding deliberation?
We deliberate not about ends, but about means.
What do doctors and orators assume in their deliberations?
They assume the end and consider how to attain it.
What does goal-based analysis not address?
It does not say what to do when several actions will achieve the goal or when no action will achieve it completely.
What idea did John Stuart Mill promote in his book Utilitarianism?
Rational decision criteria in all spheres of human activity.
What are the three fundamental areas required for mathematical formalization?
Logic, computation, and probability.
What did Gottlob Frege extend in 1879?
He extended Boole’s logic to include objects and relations, creating first-order logic.
What is considered the first nontrivial algorithm?
Euclid’s algorithm for computing greatest common divisors.
What does the term algorithm derive from?
It comes from al-Khowarazmi, a Persian mathematician of the 9th century.
What did Kurt Gödel show in 1930 regarding first-order logic?
There exists an effective procedure to prove any true statement, but it cannot capture mathematical induction.
What did Gödel's incompleteness theorem show?
In any formal theory as strong as Peano arithmetic, there are true statements that are undecidable, meaning they have no proof within the theory.
What does the Church–Turing thesis state?
The Turing machine is capable of computing any computable function.
What is an intractable problem?
A problem where the time required to solve instances grows exponentially with the size of the instances.
What is the significance of polynomial vs exponential growth?
Exponential growth means even moderately large instances cannot be solved in reasonable time.
What did Blaise Pascal show in 1654?
How to predict the future of an unfinished gambling game and assign average payoffs to gamblers.
What is the role of probability in quantitative sciences?
Helps deal with uncertain measurements and incomplete theories.
Who were the early contributors to probability theory?
Who published 'An Inquiry into the Nature and Causes of the Wealth of Nations'?
Adam Smith (1723–1790)
What is decision theory?
Combines probability theory with utility theory for decisions under uncertainty.
What does game theory study?
Interactions in small economies where one player's actions affect another's utility.
Who developed game theory?
John von Neumann and Oskar Morgenstern in 'The Theory of Games and Economic Behavior' (1944).
What is operations research?
A field that emerged to optimize decisions, starting from World War II radar installations.
What did Aristotle note about human brains?
Humans have the largest brain in proportion to their size.
What did Paul Broca demonstrate in 1861?
Localized areas of the brain are responsible for specific cognitive functions.
What is the basic organizational unit in the cerebral cortex?
A column of tissue about 0.5 mm in diameter containing about 20,000 neurons.
What technology provides detailed images of brain activity?
Functional magnetic resonance imaging (fMRI)
What is the comparison of computational units between a supercomputer and the human brain?
What is the cycle time of a brain compared to a supercomputer?
What is Hermann von Helmholtz known for?
Applying the scientific method to the study of human vision
What did Helmholtz's Handbook of Physiological Optics contribute to?
It is described as the most important treatise on the physics and physiology of human vision
What did Wundt emphasize in his psychology experiments?
Carefully controlled experiments and introspection
What are the three key steps of a knowledge-based agent according to Craik?
What was modeled in Donald Broadbent's book 'Perception and Communication'?
Psychological phenomena as information processing
What workshop is considered the start of the field of cognitive science?
A workshop in September 1956 at MIT
What significant concepts did Zuse invent?
Floating-point numbers and the first high-level programming language, Plankalkül.
Who assembled the first electronic computer?
John Atanasoff and Clifford Berry assembled the ABC between 1940 and 1942.
What was the most influential forerunner of modern computers?
The ENIAC, developed at the University of Pennsylvania.
What trend has occurred in computer hardware since its inception?
An increase in speed and capacity, and a decrease in price.
What was the purpose of Babbage's Difference Engine?
To compute mathematical tables for engineering and scientific projects.
What was the Analytical Engine?
Babbage's machine that included addressable memory, stored programs, and conditional jumps.
Who is considered the world's first programmer?
Ada Lovelace, who wrote programs for the Analytical Engine.
What contributions has AI made to computer science?
Pioneered ideas like time sharing, interactive interpreters, and personal computers.
Who built automated machines in the 17th century?
Ktesibios of Alexandria built early automated machines.
What is the goal of modern control theory?
Design systems that maximize an objective function over time.
What hybrid field emerged from modern linguistics and AI?
Computational linguistics or natural language processing
What is necessary for understanding language beyond sentence structure?
Understanding of subject matter and context
Who are credited with the first recognized work in artificial intelligence?
Warren McCulloch and Walter Pitts (1943)
What did McCulloch and Pitts propose about artificial neurons?
Each neuron is either 'on' or 'off', responding to neighboring neurons
What was the main goal of the Dartmouth workshop?
To study artificial intelligence and simulate aspects of learning and intelligence.
Who were some attendees of the Dartmouth workshop?
What theorem did the Logic Theorist prove that impressed Russell?
A theorem shorter than the one in Principia Mathematica.
Why did AI become a separate field?
It aimed to duplicate human faculties like creativity and self-improvement.
Why isn't AI a branch of mathematics?
AI embraces human faculties and is a branch of computer science, focusing on autonomous machines.
What was the early period of AI known for?
Full of successes despite primitive computers and programming tools.
What is the General Problem Solver (GPS)?
A program designed to imitate human problem-solving protocols.
What hypothesis did Newell and Simon formulate?
Physical symbol system hypothesis: Intelligence operates by manipulating data structures of symbols.
What did Arthur Samuel's program demonstrate?
Computers can learn and improve beyond their initial programming.
What significant contributions did John McCarthy make in 1958?
What is the Advice Taker?
A hypothetical program described by McCarthy as the first complete AI system.
What did the Advice Taker aim to do?
Use knowledge to search for solutions to problems and embody general world knowledge.
What principle does the Advice Taker embody?
Knowledge representation and reasoning through formal, explicit representation.
What is the resolution method?
A complete theorem-proving algorithm for first-order logic discovered by J. A. Robinson in 1965.
What was the first project to integrate logical reasoning and physical activity?
The Shakey robotics project at SRI.
What are microworlds?
Limited domains chosen by students that appeared to require intelligence to solve.
What is the blocks world?
A microworld consisting of solid blocks placed on a tabletop for rearrangement tasks.
What did Winograd and Cowan's work show?
A large number of elements could collectively represent an individual concept.
Who enhanced learning methods in AI with adalines?
Bernie Widrow (Widrow and Hoff, 1960; Widrow, 1962)
What does the perceptron convergence theorem state?
The learning algorithm can adjust connection strengths to match any input data, provided a match exists.
What did Herbert Simon predict about computers within 10 years?
A computer would be chess champion and a significant mathematical theorem would be proved by machine.
What was a major difficulty of early AI programs?
Most early programs knew nothing of their subject matter; they succeeded through simple syntactic manipulations.
What was a notable failure in early machine translation efforts?
Accurate translation requires background knowledge to resolve ambiguity.
What did a 1966 report by an advisory committee conclude about machine translation?
There has been no machine translation of general scientific text, and none is in immediate prospect.
What was the second kind of difficulty faced by AI?
The intractability of many problems AI attempted to solve.
What strategy did early AI programs use to solve problems?
Trying out different combinations of steps until a solution was found.
What misconception existed before the theory of computational complexity?
Scaling up to larger problems was thought to be a matter of faster hardware and larger memories.
What dampened optimism in resolution theorem proving?
Researchers failed to prove theorems involving more than a few dozen facts.
What is the illusion of unlimited computational power related to?
It was not confined to problem-solving programs and was evident in early experiments with machine evolution (genetic algorithms).
What was the goal of genetic algorithms?
To generate a program with good performance for a specific task through random mutations and a selection process.
What was a major criticism of AI in the Lighthill report?
Failure to address the 'combinatorial explosion'.
What happened to AI research funding after the Lighthill report?
The British government ended support for AI research in all but two universities.
What did Minsky and Papert's book Perceptrons (1969) demonstrate?
Perceptrons could learn anything they could represent, but they represented very little.
What limitation did a two-input perceptron have?
It could not be trained to recognize when its two inputs were different.
What caused a decline in funding for neural-net research?
Results from Minsky and Papert's research on perceptrons.
What learning algorithm caused a resurgence in neural-net research in the late 1980s?
Back-propagation learning algorithms.
What are weak methods in AI?
General-purpose search mechanisms that do not scale up to large or difficult problem instances.
What is the alternative to weak methods in problem-solving?
Using more powerful, domain-specific knowledge for larger reasoning steps.
Who were the key developers of the DENDRAL program?
Ed Feigenbaum, Bruce Buchanan, and Joshua Lederberg.
What input did the DENDRAL program require?
The elementary formula of the molecule and the mass spectrum of its fragments.
What was a naive approach of the DENDRAL program?
Generating all possible structures consistent with the formula and predicting mass spectra for comparison.
What was the purpose of the DENDRAL researchers?
To analyze molecular structures using spectral data and identify common substructures.
What rule is used to recognize a ketone subgroup?
If two peaks at x1 and x2 satisfy: - x1 + x2 = M + 28 - x1 - 28 is a high peak - x2 - 28 is a high peak - At least one of x1 or x2 is high.
What was significant about DENDRAL?
It was the first successful knowledge-intensive system using many special-purpose rules.
What project did Feigenbaum and others start at Stanford?
The Heuristic Programming Project (HPP) to apply expert systems methodology to various fields.
What was the purpose of MYCIN?
To diagnose blood infections using about 450 rules, performing as well as some experts.
How did MYCIN differ from DENDRAL?
MYCIN’s rules were acquired through expert interviews and reflected medical uncertainty with certainty factors.
What did Winograd's SHRDLU system demonstrate?
It showed that natural language understanding could overcome ambiguity but was limited to a specific domain (blocks world).
What did Roger Schank emphasize in understanding natural language?
He claimed that robust understanding requires general knowledge and methods beyond syntax.
What was the focus of Schank and his students' programs?
They focused on representing and reasoning with knowledge required for language understanding.
What were the problems of language understanding identified?
What led to an increase in knowledge representation schemes?
The growth of applications to real-world problems.
What is the frames approach in AI?
A structured method assembling facts about object and event types into a taxonomic hierarchy.
How many expert systems did DuPont have in use by 1988?
100 expert systems in use and 500 in development.
What was the Fifth Generation project announced by Japan?
A 10-year plan to build intelligent computers running Prolog.
What was the purpose of the Microelectronics and Computer Technology Corporation (MCC)?
To assure national competitiveness in AI research.
What happened to the AI industry from 1980 to 1988?
It grew from a few million dollars to billions of dollars.
What did the Parallel Distributed Processing collection cause?
Great excitement in connectionist models of intelligent systems.
What are the two fields modern neural network research has bifurcated into?
What has revolutionized AI methodology in recent years?
Building on existing theories, rigorous theorems, and real-world applications.
What did David McAllester state about AI's early isolationism?
AI was isolated from computer science but is now embracing fields like control theory and statistics.
What is required for hypotheses in AI to be accepted?
Hypotheses must undergo rigorous empirical experiments and statistical analysis.
What has dominated the field of speech recognition in recent years?
Approaches based on hidden Markov models (HMMs).
What are the two relevant aspects of HMMs in speech recognition?
What is the significance of HMMs in blind tests?
HMMs have been steadily improving their scores in rigorous blind tests.
How is machine translation related to speech recognition?
Both have evolved from initial enthusiasm to rigorous mathematical frameworks.
What characterized the shift in AI methodology towards neatness?
The field has reached stability and maturity, favoring mathematical rigor.
What is the role of HMMs in understanding speech recognition?
HMMs provide a mathematical framework for the problem without claiming human similarity.
What was the initial approach to machine translation in the 1950s?
Based on sequences of words with models learned according to information theory.
What was the focus of neural networks research in the 1980s?
To explore capabilities and differences from traditional techniques.
What did Judea Pearl's 1988 work contribute to AI?
Led to acceptance of probability and decision theory in AI.
What is the Bayesian network formalism used for?
Efficient representation and reasoning with uncertain knowledge.
What do normative expert systems do?
Act rationally according to decision theory without imitating human thought.
What has led to the emergence of intelligent agents since 1995?
Progress in solving subproblems of AI and the whole agent problem.
Who worked on the complete agent architecture called SOAR?
Allen Newell, John Laird, and Paul Rosenbloom.
What is a consequence of building complete agents in AI?
Need to reorganize isolated subfields to integrate results.
What is a challenge for reasoning and planning systems in AI?
Handling uncertainty from sensory systems.
What recent progress has been made in robotic cars?
What do some AI founders believe AI should focus on?
What significant data sources have become available since 2001?
What did Yarowsky's 1995 work achieve?
96% accuracy in word-sense disambiguation without labeled examples
What example did Banko and Brill discuss regarding data performance?
Performance increases with data volume, surpassing algorithm differences
What algorithm did Hays and Efros define for filling in masked areas in photos?
An algorithm that searches through a collection of photos to find a match for the background.
What was the performance threshold of Hays and Efros' algorithm?
Performance improved from poor with 10,000 photos to excellent with 2 million photos.
What does the knowledge bottleneck in AI refer to?
The challenge of expressing all necessary knowledge for a system.
What have reporters noted about AI's development?
They noted that 'AI Winter' may be yielding to a new Spring.
What is an example of a robotic vehicle?
STANLEY, a driverless car that won the 2005 DARPA Grand Challenge.
What is an application of speech recognition in AI?
Automated systems guiding conversations for booking flights.
What was NASA's Remote Agent program?
The first on-board autonomous planning program for spacecraft operations.
What role do learning algorithms play in spam fighting?
They classify over a billion messages as spam daily.
What challenge do spam fighters face?
Spammers continually update their tactics, making static approaches ineffective.
What was a logistical application during the Persian Gulf crisis of 1991?
Dynamic analysis for U.S. forces deployment.
How many vehicles, cargo, and people did DART account for during the Persian Gulf crisis?
Up to 50,000
What was the impact of DART on planning time?
Generated a plan in hours that took weeks with older methods
What organization stated that DART paid back their 30-year investment in AI?
Defense Advanced Research Project Agency (DARPA)
What is the PackBot used for in Iraq and Afghanistan?
Handle hazardous materials, clear explosives, identify snipers
What statistical model is used in the Arabic to English translation program?
Built from examples of Arabic-to-English translations and English text totaling two trillion words
What philosophical idea made AI conceivable?
The mind is like a machine operating on knowledge encoded in an internal language
What did neuroscientists discover about the brain?
How it works and its similarities and differences from computers
What does control theory deal with?
Designing devices that act optimally based on environmental feedback
What has influenced the history of AI?
Cycles of success, optimism, cutbacks, and new creative approaches
What recent advancement has contributed to the rapid progress of AI?
Greater use of scientific methods
What has contributed to the rapid advancement of AI in the past decade?
Greater use of the scientific method in experimenting and comparing approaches.
How have theoretical understanding and real systems progressed in AI?
They have improved hand in hand with each other.
Which book gives a readable account of the philosophical problems of AI?
Artificial Intelligence: The Very Idea by John Haugeland (1985)
What are the major AI conferences mentioned?
What is the focus of exercise 1.1?
Define in your own words:
- intelligence
- artificial intelligence
- agent
- rationality
- logical reasoning
What does exercise 1.2 ask about Turing's original paper?
Discuss objections to his test for intelligence and their current relevance.
What question does exercise 1.3 pose about reflex actions?
Are reflex actions rational? Are they intelligent?
What is a reflex action?
A reflex action is an automatic response to a stimulus, such as flinching from a hot stove.
Are reflex actions rational or intelligent?
Reflex actions are not considered rational or intelligent as they occur automatically without conscious thought.
What happens if Evans's ANALOGY program scores 200 on an IQ test?
It does not necessarily mean the program is more intelligent than a human; intelligence is multifaceted and not solely based on IQ.
Why is the sea slug Aplysia studied?
Aplysia has about 20,000 large neurons, making it easier to study neural structures and functions.
How does Aplysia's computational power compare to high-end computers?
The comparison depends on the cycle time for neuron updates and the architecture of the high-end computer.
Can introspection be inaccurate?
Yes, individuals can misinterpret or be unaware of their true thoughts.
Are supermarket bar code scanners AI?
They are a form of AI as they automate a task but lack advanced intelligence.
Are web search engines instances of AI?
Yes, they utilize algorithms to process and retrieve information intelligently.
Are voice-activated telephone menus AI?
Yes, they use speech recognition and processing to interact with users.
Are internet routing algorithms AI?
Yes, they adapt to network conditions dynamically, showcasing intelligent behavior.
Do humans learn complex mathematics for cognitive activities?
Most humans do not learn complex mathematics before college, yet cognitive models may involve such computations.
Why does evolution favor rational systems?
Rational systems enhance survival by effectively achieving goals related to reproduction and resource acquisition.
Is AI a science or engineering?
AI encompasses both science and engineering, as it involves theoretical research and practical application.
Can computers be intelligent?
Computers follow programmed instructions but can exhibit intelligent behavior through complex algorithms.
Can animals be intelligent?
Animals act based on instincts and learned behaviors, indicating a form of intelligence.
Can the laws of physics dictate intelligence?
While all systems operate under physical laws, intelligence arises from complex interactions, not just physical dictates.
What tasks can computers currently solve?
Tasks include web searching, basic games, and limited driving but not complex tasks like surgery or creative writing.
What are some currently infeasible tasks for AI?
Examples include playing competitive bridge, complex surgery, and creative storytelling.
What is the purpose of AI contests?
AI contests encourage innovation by defining tasks for researchers to solve, advancing the field.
What is the DARPA Grand Challenge?
A contest for robotic cars to navigate autonomously, showcasing advancements in robotics and AI.
What is the International Planning Competition?
A contest that challenges AI systems to solve planning problems effectively.
What are intelligent agents?
Systems that can be viewed as perceiving their environment through sensors and acting upon it through actuators.
What is the difference between agent function and agent program?
Agent function is an abstract mathematical description; agent program is a concrete implementation.
What does the vacuum agent perceive in the vacuum-cleaner world?
What is a simple agent function in the vacuum-cleaner world?
If the current square is dirty, then suck; otherwise, move to the other square.
What does the performance measure evaluate?
The desirability of the sequence of environment states generated by the agent's actions.
What is an important aspect to define success for an agent?
Success should be defined in terms of the environment states, not the agent's opinion of its performance.
What can affect an agent's perception of its performance?
An agent can delude itself about its performance, leading to a false sense of rationality.
What is the goal of analyzing agents?
To understand how artifacts interact with the world, particularly in AI with significant computational resources.
What is a performance measure in agent design?
A criterion to evaluate success, tailored to the task and agent.
How can a vacuum-cleaner agent maximize its performance measure?
By cleaning dirt and then dumping it, repeating this process.
What is a better performance measure for a vacuum-cleaner agent?
Reward for clean squares, with penalties for electricity and noise.
What defines a rational agent?
An agent that selects actions to maximize its performance measure based on percept sequences and knowledge.
What factors influence the rationality of an agent?
What is the performance measure for the vacuum-cleaner agent example?
One point for each clean square per time step over 1000 time steps.
When would the vacuum-cleaner agent be considered irrational?
When it oscillates back and forth after all dirt is cleaned.
What philosophical questions arise from agent performance measures?
What happens when the agent oscillates needlessly after cleaning?
It fares poorly due to a penalty for each movement left or right.
What should a better agent do once all squares are clean?
It should do nothing if sure all squares are clean.
What should an agent do if clean squares can become dirty again?
It should occasionally check and re-clean them if needed.
What must an agent do in an unknown environment?
It needs to explore rather than stick to known squares.
What is the difference between rationality and omniscience?
Rationality maximizes expected performance; omniscience knows actual outcomes.
What example illustrates the difference between rationality and perfection?
Crossing the street while a cargo door falls off an airliner.
What should a rational agent do before crossing a busy road?
It should look both ways before crossing.
What is an important part of rationality related to percepts?
Gathering information to modify future percepts.
What does a rational agent need to do with its perceptions?
It must learn as much as possible from what it perceives.
What happens in extreme cases where the environment is completely known?
The agent need not perceive or learn; it simply acts correctly.
What is an example of a fragile agent mentioned in the text?
The dung beetle that continues its task even if the dung ball is removed.
What happens if the dung ball is removed from the beetle's grasp?
The beetle continues to pantomime plugging the nest.
What does the female sphex wasp do after stinging a caterpillar?
Drags the caterpillar to its burrow.
What should a rational agent do to be considered autonomous?
Learn to compensate for partial or incorrect prior knowledge.
How can a vacuum-cleaning agent improve its performance?
By learning to foresee where and when additional dirt will appear.
What is essential for specifying a task environment?
To specify the performance measure, environment, actuators, and sensors.
Why is the automated taxi driver task considered open-ended?
Novel combinations of circumstances can arise.
What does PEAS stand for in the context of the taxi's task environment?
Performance Measure, Environment, Actuators, Sensors
What is the performance measure for the automated taxi driver?
What sensors are used in an automated taxi?
What are some desirable qualities for the automated taxi's performance measure?
What optional driving conditions might the taxi face?
What communication methods might the automated taxi use?
What is a key feature of the sensors in an automated taxi?
They include controllable video cameras and sensors for distances to obstacles.
What is an example of a simple environment for a robot?
A robot inspecting parts on a conveyor belt.
What is the significance of the relationship between agent behavior and environment?
It determines the complexity of the agent's performance measure.
What is a softbot designed to do?
What abilities does a softbot operator need?
What characterizes the Internet environment?
What dimensions categorize task environments in AI?
What is a fully observable environment?
An environment where sensors provide complete state information at all times.
What defines a partially observable environment?
An environment where some state aspects are missing due to sensor limitations.
What is the difference between single-agent and multiagent environments?
Single-agent involves one agent (e.g., solving a puzzle), multiagent involves multiple agents (e.g., playing chess).
What is the key distinction for treating another vehicle as an agent?
Whether its behavior maximizes a performance measure dependent on agent A's behavior
What type of environment is taxi driving classified as?
Partially cooperative multiagent environment
What is a characteristic of communication in multiagent environments?
It often emerges as a rational behavior
What defines a deterministic environment?
Next state is completely determined by current state and agent's action
What is the implication of an uncertain environment?
It is not fully observable or not deterministic
What is a nondeterministic environment?
Actions characterized by possible outcomes without probabilities attached
What is an episodic task environment?
Agent's experience is divided into atomic episodes where actions are independent
What is the difference between sequential and episodic environments?
What characterizes a static environment?
The environment does not change while the agent is deciding on an action.
What characterizes a dynamic environment?
The environment can change while the agent is deliberating.
What is a semidynamic environment?
The environment does not change with time, but the agent’s performance score does.
What is the difference between discrete and continuous environments?
What does a known environment imply?
Outcomes for all actions are given, or their probabilities if stochastic.
What does an unknown environment imply?
The agent must learn how the environment works to make good decisions.
What is the hardest case for an environment?
Partially observable, multiagent, stochastic, sequential, dynamic, continuous, and unknown.
How is the part-picking robot described in terms of environment?
It is described as episodic, as it normally considers each part in isolation.
What are the characteristics of a crossword puzzle task environment?
What are the characteristics of chess with a clock task environment?
What are the characteristics of poker task environment?
What are the characteristics of backgammon task environment?
What are the characteristics of taxi driving task environment?
What are the characteristics of medical diagnosis task environment?
What is the role of the code repository in the context of environments?
It includes implementations of environments and a simulator to evaluate agent behavior.
What does the environment generator do?
It selects particular environments for running agents with specific likelihoods.
What is a rational agent for a given environment class?
It maximizes performance across various environments.
What is the difference between agent program and agent function?
Agent program takes current percept; agent function takes entire percept history.
What is the purpose of the table in the agent program?
To represent the agent function and decide actions.
What is a potential failure of the table-driven approach to agent construction?
Exponential growth of entries in the lookup table.
What is the data rate of visual input from a single camera in an automated taxi?
27 megabytes per second
What is the size comparison of the lookup table to the number of atoms in the observable universe?
Less than 10^80 atoms
What are the four basic kinds of agent programs?
What is the advantage of simple reflex agents compared to lookup tables?
They require less space and processing
What is a notable example of replacing large tables with a program?
Newton’s method for square roots
What does the INTERPRET-INPUT function do?
Generates an abstracted description of the current state.
What is a limitation of simple reflex agents?
Limited intelligence; works only with fully observable environments.
What happens if the braking rule cannot determine the condition?
The agent may brake unnecessarily or not at all.
What happens to a reflex vacuum agent without a location sensor?
It may get stuck in infinite loops.
What is the risk of moving left or right in a vacuum environment?
It may fail if starting in certain squares (A or B).
What might a vacuum agent do when it perceives [Clean]?
Flip a coin to choose between Left and Right.
What knowledge is needed to update internal state information?
What does a model-based reflex agent use to generate the current state?
The current percept and the old internal state.
What does the RULE-MATCH function do?
It matches the current state with the rules and determines the action to take.
What does a model-based reflex agent do?
It keeps track of the current state of the world and chooses actions like a reflex agent.
How does the representation of models and states vary?
It varies widely depending on the environment type and agent design technology.
What does the box labeled 'what the world is like now' represent?
It represents the agent's 'best guess' about the current state.
Why might uncertainty about the current state be unavoidable?
Because the agent may not have complete information about the environment.
What does a model-based agent's internal state not have to describe literally?
It does not have to describe 'what the world is like now' literally.
What does a goal-based agent keep track of?
It keeps track of the world state and a set of goals to achieve.
What action does a goal-based agent choose?
It chooses an action that will eventually lead to achieving its goals.
What might influence a taxi's decision to fill up with gas?
The taxi's rule about having at least half a tank of gas.
What is the focus of the book?
Explores logic, probability, perception, reasoning, learning, action.
What does the subtitle 'A Modern Approach' imply?
Synthesizes current knowledge into a common framework.
What is the purpose of this book according to the authors?
To explore the full breadth of AI, including depth in various areas.
What are some of the contributing writers?
What is the publication's protection status?
Protected by copyright; permissions required for reproduction.
What is stated about the authors' efforts in preparing the book?
Best efforts in development, research, and testing effectiveness.
What does the book apologize for?
Subfields being less recognizable due to the common framework approach.
What are some important applications of AI technology?
What are the three types of representations for agents?
What new material has been added regarding probabilistic models?
What has been rewritten in the machine-learning chapter?
What is the main theme of the book related to intelligent agents?
AI is the study of agents that receive percepts and perform actions.
What is the primary aim of the book?
Convey ideas from 50 years of AI research and 2000 years of related work.
What is the role of learning in agent design?
Extends the designer's reach into unknown environments.
What is the prerequisite for the book?
Familiarity with basic concepts of computer science at a sophomore level.
What can you find on the Web site aima.cs.berkeley.edu?
What is depicted on the cover of the book?
Final position from the game 6 of Kasparov vs. DEEP BLUE chess match.
What does the cover feature alongside Kasparov?
Who reviewed the manuscript?
Who are the authors thanking for their support and encouragement?
Who provided especially helpful comments on the book?
Who is Stuart Russell?
What awards has Stuart Russell received?
What are some notable works of Stuart Russell?
Who is Peter Norvig?
What was Peter Norvig's role at NASA?
What is the focus of the authors' work?
Key concepts related to machine learning and artificial intelligence
What is the significance of the listed names?
They are contributors and experts in the field of artificial intelligence and machine learning.
What is the role of the Association for Computing Machinery?
Promotes computing as a science and profession.
What field did he oversee at NASA?
Research and development in artificial intelligence and robotics.
What is one of his books on AI programming?
Paradigms of AI Programming: Case Studies in Common Lisp.
What is the title of his book on translation systems?
Verbmobil: A Translation System for Face-to-Face Dialog.
What is a Constraint Satisfaction Problem (CSP)?
A problem where the goal is to find values for variables that satisfy a set of constraints.
What does Constraint Propagation do in CSPs?
Infers variable values based on constraints to reduce the search space.
What is Backtracking Search?
A depth-first search algorithm that incrementally builds candidates for solutions and abandons candidates as soon as it determines they cannot lead to a valid solution.
What is Local Search in CSPs?
An optimization technique that starts with a complete assignment and iteratively makes small changes to find a better solution.
What is the Wumpus World?
A standard problem in AI used to illustrate the concepts of knowledge-based agents and reasoning.
What is Propositional Logic?
A branch of logic dealing with propositions that can be true or false, used in knowledge representation.
What is First-Order Logic?
An extension of propositional logic that includes quantifiers and predicates, allowing for more expressive statements.
What is Inference in First-Order Logic?
The process of deriving new information from known facts using logical rules.
What is Classical Planning?
The process of creating a sequence of actions to achieve specific goals based on a model of the world.
What are Planning Graphs?
Data structures that represent the planning problem, showing actions and their effects over time.
What is Hierarchical Planning?
A planning approach that breaks down tasks into subtasks, allowing for more manageable planning.
What is Ontological Engineering?
The process of creating and managing ontologies to represent knowledge in a domain.
What are Mental Events in knowledge representation?
Events that occur in the mind, such as beliefs and desires, used to model cognitive processes.
What is Reasoning with Default Information?
A method of reasoning that allows for conclusions to be drawn in the absence of complete information.
What are Decision Trees used for in machine learning?
To model decisions and their possible consequences.
What is the purpose of Evaluating and Choosing the Best Hypothesis?
To select the most effective model based on performance metrics.
What does the Theory of Learning encompass?
Principles and frameworks that describe how learning occurs.
What is Regression in machine learning?
A method for predicting continuous outcomes based on input variables.
What are Artificial Neural Networks?
Computational models inspired by the human brain, used for various tasks.
What are Nonparametric Models?
Models that do not assume a fixed form for the underlying data distribution.
What are Support Vector Machines?
A supervised learning model used for classification and regression tasks.
What does Practical Machine Learning involve?
Application of machine learning techniques to real-world problems.
What is the focus of Knowledge in Learning?
Understanding how knowledge is represented and utilized in learning.
What is Inductive Logic Programming?
A method of machine learning that uses logic programming as a uniform representation.
What is Statistical Learning?
A framework for understanding the relationships among data through statistics.
What is the EM Algorithm?
An iterative method for finding maximum likelihood estimates in models with hidden variables.
What is Reinforcement Learning?
A type of learning where an agent learns to make decisions by receiving rewards or penalties.
What does Natural Language Processing focus on?
The interaction between computers and human language.
What is a Language Model?
A statistical model that predicts the next word in a sequence based on prior words.
What is Machine Translation?
Automatic translation of text from one language to another using algorithms.
What does Perception in AI refer to?
The process of interpreting sensory information to understand the environment.
What is the significance of 3D Reconstruction?
Creating a 3D model from 2D images or data to understand spatial relationships.
What is the perception of AI among scientists in other disciplines?
It is often cited as the field they would most like to be in.
What does AI encompass?
AI encompasses a variety of subfields including: - Learning - Perception - Playing chess - Proving mathematical theorems - Writing poetry - Driving cars - Diagnosing diseases
What is the Turing Test?
The Turing Test determines if a computer can mimic human responses well enough that a human cannot distinguish them from a person's.
What capabilities must a computer have to pass the Turing Test?
To pass the Turing Test, a computer must have: - Natural language processing - Knowledge representation - Automated reasoning - Machine learning
What is natural language processing?
Natural language processing enables a computer to communicate successfully in English.
What is knowledge representation?
Knowledge representation is the ability to store what a computer knows or hears.
What is automated reasoning?
Automated reasoning is using stored information to answer questions and draw new conclusions.
What is machine learning?
Machine learning allows a computer to adapt to new circumstances and detect and extrapolate patterns.
What does Turing's test avoid?
Direct physical interaction between the interrogator and the computer.
What does the total Turing Test include?
A video signal and the ability to pass physical objects through a hatch.
What are the requirements to pass the total Turing Test?
What is the focus of AI researchers regarding the Turing Test?
Studying underlying principles of intelligence rather than duplicating an exemplar.
What is the cognitive modeling approach in AI?
Determining how humans think to express it as a computer program.
What are the three ways to understand human thought?
What is evidence that a program's mechanisms could operate in humans?
If the program’s input-output behavior matches corresponding human behavior.
What was often confused in the early days of AI?
The approaches of algorithm performance and human performance modeling.
What notation did logicians develop in the 19th century?
A precise notation for statements about objects and their relations.
What is the logicist tradition in AI?
It aims to build intelligent systems based on logical programs.
What are the two main obstacles in the logicist approach?
What defines a rational agent?
An agent that acts to achieve the best outcome or best expected outcome under uncertainty.
What is one way a rational agent can act?
By reasoning logically to a conclusion that achieves its goals.
What are the advantages of the rational-agent approach?
How is human behavior described in the text?
Well adapted for one specific environment; defined by the sum total of human actions.
What does the book focus on?
General principles of rational agents and components for constructing them.
Is achieving perfect rationality feasible in complicated environments?
No, the computational demands are too high.
What is the working hypothesis regarding perfect rationality?
It simplifies the problem and provides a setting for foundational material.
What do Chapters 5 and 17 discuss?
Limited rationality—acting appropriately with insufficient time for computations.
What disciplines contributed to Artificial Intelligence?
Philosophy, among others, with various ideas and techniques.
What questions does the history of AI focus on?
Formal rules for conclusions, mind-physical brain relation, knowledge origins, knowledge leading to action.
What is the Pascaline known for?
Built in 1642 by Blaise Pascal; famous for its arithmetic capabilities.
What does materialism assert about the mind?
The mind is the brain's operation according to physical laws.
What did John Locke state about understanding?
Nothing is in the understanding, which was not first in the senses.
What does Aristotle suggest about deliberation in the Nicomachean Ethics?
We deliberate not about ends, but about means.
What is the main focus of goal-based analysis?
It is useful but does not specify actions when several actions achieve a goal or none completely.
What did John Stuart Mill promote in his book Utilitarianism?
The idea of rational decision criteria in all spheres of human activity.
What are the three fundamental areas for mathematical formalization in AI?
Logic, computation, and probability.
What did Gottlob Frege extend in 1879?
Boole's logic to include objects and relations, creating first-order logic.
What did Alfred Tarski introduce?
A theory of reference to relate logic objects to real-world objects.
What is considered the first nontrivial algorithm?
Euclid's algorithm for computing greatest common divisors.
What does the word 'algorithm' originate from?
The writings of al-Khowarazmi, a Persian mathematician.
What did Kurt Gödel show in 1930 regarding first-order logic?
There exists an effective procedure to prove any true statement, but it cannot capture mathematical induction.
What did Gödel's incompleteness theorem show?
There are true statements in Peano arithmetic that are undecidable and have no proof within the theory.
What does the Church–Turing thesis state?
The Turing machine is capable of computing any computable function.
What is an intractable problem?
A problem where the time to solve instances grows exponentially with the size of the instances.
What is the significance of NP-complete problems?
Problems that can be reduced to NP-complete problems are likely to be intractable.
What was Blaise Pascal's contribution to probability?
He showed how to predict the future of an unfinished gambling game and assign average payoffs.
What role does probability play in quantitative sciences?
Helps deal with uncertain measurements and incomplete theories.
When did economics begin as a science?
In 1776, with Adam Smith's publication of 'An Inquiry into the Nature and Causes of the Wealth of Nations'.
What concept did Adam Smith introduce in economics?
Economies consist of individual agents maximizing their own economic well-being.
What is the mathematical treatment of utility?
First formalized by Léon Walras (1834-1910) and improved by Frank Ramsey and others.
What is the focus of game theory?
The actions of one player significantly affect the utility of another.
What is a key aspect of operations research?
Emerging from WWII to optimize decisions, such as radar installations.
What are Markov decision processes?
A class of sequential decision problems formalized by Richard Bellman (1957).
What concept did Herbert Simon introduce in AI?
Satisficing—making decisions that are 'good enough' rather than optimal.
What is the concept of satisficing?
Choosing options that are 'good enough' instead of striving for an optimal decision.
What did Aristotle note about human brains?
Humans have the largest brain in proportion to body size.
What did Paul Broca study in 1861?
Aphasia in brain-damaged patients, showing localized brain functions.
Who was the first to apply mathematical models to the nervous system?
Nicolas Rashevsky (1936, 1938).
What is the basic organizational unit in the cerebral cortex?
A column of tissue about 0.5 mm in diameter containing about 20,000 neurons.
What can change in the brain over a few weeks?
Mappings between areas of the brain and body parts they control.
What imaging technique provides detailed images of brain activity?
Functional magnetic resonance imaging (fMRI)
What is the comparison of computational units between supercomputers, PCs, and the human brain?
What is the cycle time of a supercomputer compared to the human brain?
Supercomputer: 10^{-9} sec Human Brain: 10^{-3} sec
Who are the key figures in the origins of scientific psychology?
What is the main difference in properties between brains and digital computers?
Cycle time and storage capacity
What key steps did Craik identify for a knowledge-based agent?
What papers were presented at the MIT workshop?
What is a common view among psychologists regarding cognitive theories?
They should describe a detailed information-processing mechanism like a computer program
What did Zuse invent besides the Z-3?
Floating-point numbers and the first high-level programming language, Plankalkül.
Who assembled the first electronic computer?
John Atanasoff and Clifford Berry between 1940 and 1942.
What was the ENIAC?
A secret military project at the University of Pennsylvania, influential for modern computers.
What trend occurred in computer hardware after 2005?
Manufacturers started multiplying the number of CPU cores instead of increasing clock speed.
What was the purpose of Babbage’s Difference Engine?
To compute mathematical tables for engineering and scientific projects.
What was the Analytical Engine?
Babbage's machine with addressable memory and stored programs, capable of universal computation.
Who is considered the first programmer?
Ada Lovelace, who wrote programs for Babbage's Analytical Engine.
What contributions has AI made to computer science?
Ideas like time sharing, interactive interpreters, and object-oriented programming.
What was Turing's interest in computers post-World War II?
Using them for AI research, including chess programs.
What is the goal of modern control theory?
Design systems that maximize an objective function over time.
What hybrid field emerged from modern linguistics and AI?
Computational linguistics or natural language processing
What is required for understanding language beyond sentence structure?
Understanding of subject matter and context
What three sources influenced McCulloch and Pitts' work?
What was the purpose of the Dartmouth workshop in 1956?
To study artificial intelligence and simulate aspects of learning and intelligence.
Which two researchers from Carnegie Tech gained significant attention during the workshop?
Allen Newell and Herbert Simon
What program did Newell and Simon create that was capable of non-numerical thinking?
Logic Theorist (LT)
What theorem did the Logic Theorist prove that impressed Russell?
A theorem from Principia Mathematica
Why did AI need to become a separate field?
It aimed to duplicate human faculties like creativity and self-improvement.
What was the official proposal for the Dartmouth workshop's conjecture?
Every aspect of learning can be described for machine simulation.
What major figures were introduced to each other at the Dartmouth workshop?
John McCarthy, Minsky, Claude Shannon, Nathaniel Rochester, Newell, Simon, and others.
What is a key objective of AI?
To duplicate human faculties such as creativity, self-improvement, and language use.
Why isn't AI considered a branch of mathematics?
AI embraces the idea of duplicating human faculties, which other fields do not address.
What methodology distinguishes AI from other fields?
AI is a branch of computer science and attempts to build autonomous machines in complex environments.
What period is referred to as the 'Look, Ma, no hands!' era?
The early years of AI from 1952 to 1969.
What was the General Problem Solver (GPS)?
A program designed to imitate human problem-solving protocols.
What does the physical symbol system hypothesis state?
Any intelligent system must manipulate data structures composed of symbols.
What did Arthur Samuel's checkers program demonstrate?
It learned to play at a strong amateur level, disproving that computers can only do what they are told.
What significant contributions did John McCarthy make in 1958?
Defined the Lisp programming language, invented time sharing, and described the Advice Taker program.
What is Lisp?
A high-level programming language defined by John McCarthy, dominant in AI for 30 years.
What was the Advice Taker?
A hypothetical program described by McCarthy as the first complete AI system.
What did John McCarthy describe in his 1958 paper?
The Advice Taker, a hypothetical program seen as the first complete AI system.
What was the primary function of the Advice Taker?
To use knowledge to search for solutions to problems and embody general knowledge of the world.
How could the Advice Taker improve its competence?
By accepting new axioms during operation without needing reprogramming.
What was McCarthy's focus compared to Minsky's?
McCarthy focused on representation and reasoning, while Minsky focused on getting programs to work.
What significant discovery did J. A. Robinson make in 1965?
The resolution method, a complete theorem-proving algorithm for first-order logic.
What did the Shakey robotics project demonstrate?
The integration of logical reasoning and physical activity.
What type of problems did Evans’s ANALOGY program solve?
Geometric analogy problems found in IQ tests.
What is the blocks world in AI?
A microworld with solid blocks for tasks like rearranging using a robot hand.
What did Winograd and Cowan's work demonstrate?
A large number of elements can represent an individual concept, increasing robustness and parallelism.
What does the perceptron convergence theorem state?
The learning algorithm can adjust connection strengths to match input data, if such a match exists.
What did Herbert Simon predict in 1957?
Machines that think, learn, and create will increase rapidly in capability.
What was Simon's prediction about computers within 10 years?
A computer would be a chess champion and prove a significant mathematical theorem.
What was a major issue with early AI programs?
They often knew nothing of their subject matter and relied on simple syntactic manipulations.
What problem arose in early machine translation efforts?
Accurate translation requires background knowledge to resolve ambiguity.
What was the outcome of the 1966 report on machine translation?
No machine translation of general scientific text was achieved; funding was canceled.
What was a limitation of early AI problem-solving methods?
They relied on trying different combinations of steps, which failed with larger problems.
What illusion did early AI researchers have about computational power?
They believed scaling up problems was just a matter of faster hardware and larger memories.
What was the issue with resolution theorem proving in AI?
Researchers struggled to prove theorems involving more than a few dozen facts.
What is the illusion of unlimited computational power in AI?
It refers to the belief that problem-solving programs can solve any problem effectively, which has proven to be incorrect.
What are genetic algorithms?
They are methods based on making small mutations to machine-code programs to generate effective solutions for specific tasks.
What was a major criticism of AI in the Lighthill report?
It highlighted the failure to address the 'combinatorial explosion' in AI research, leading to reduced funding.
What limitation did Perceptrons have?
A two-input perceptron could not be trained to recognize when its two inputs were different.
What are weak methods in AI?
General-purpose search mechanisms that struggle with large or complex problem instances.
What is the DENDRAL program?
An early AI program developed to infer molecular structures from mass spectrometry data.
Who were the key contributors to the DENDRAL program?
Ed Feigenbaum, Bruce Buchanan, and Joshua Lederberg.
What input does the DENDRAL program require?
Elementary formula of the molecule and the mass spectrum from a mass spectrometer.
What is a challenge faced by the DENDRAL program?
Generating all possible structures consistent with the formula is intractable for moderate-sized molecules.
What rule is used to recognize a ketone subgroup?
If x1 + x2 = M + 28, x1 - 28 is high, x2 - 28 is high, and at least one of x1 or x2 is high.
What was significant about DENDRAL in AI history?
It was the first successful knowledge-intensive system.
What project did Feigenbaum and others start after DENDRAL?
The Heuristic Programming Project (HPP).
What was a major difference between MYCIN and DENDRAL?
MYCIN's rules were acquired from expert interviews, not derived from a theoretical model.
What did MYCIN incorporate to handle uncertainty?
A calculus of uncertainty called certainty factors.
What did Roger Schank emphasize about language understanding?
Robust understanding requires general knowledge about the world.
What was the focus of Schank's programs?
Representing and reasoning with knowledge for language understanding.
What are the problems related to language understanding?
What led to the growth of knowledge representation schemes?
The increase in applications to real-world problems.
What was the Fifth Generation project?
A 10-year plan announced by Japan to build intelligent computers running Prolog.
What was the purpose of the Microelectronics and Computer Technology Corporation (MCC)?
To assure national competitiveness in AI research.
What did the collection Parallel Distributed Processing cause?
Great excitement in the field of AI and psychology.
What does Terrence Deacon's book suggest about humans?
Symbol manipulation is a defining characteristic of humans.
What is the current view on connectionist and symbolic approaches?
They are complementary, not competing.
What is the current view of connectionist and symbolic approaches in AI?
They are considered complementary, not competing.
What are the two fields modern neural network research has bifurcated into?
What methodology has AI adopted since 1987?
AI has embraced the scientific method, emphasizing rigorous empirical experiments and statistical analysis.
What has changed in AI's approach to theories?
It is more common to build on existing theories rather than propose brand-new ones.
How does current AI research differ from earlier periods?
It is now based on rigorous theorems and hard experimental evidence, showing relevance to real-world applications.
What did David McAllester say about AI's early isolationism?
AI was largely separated from other fields like computer science, but this isolationism is being abandoned.
What is the significance of hidden Markov models (HMMs) in speech recognition?
HMMs are based on rigorous mathematical theory and trained on large real speech data, ensuring robust performance.
What is the trend in the field of speech technology?
It is transitioning to widespread industrial and consumer applications.
What characterizes the change in AI methodology?
A victory of the neats over the scruffies, implying a shift towards mathematical rigor and stability.
What parallels exist between machine translation and speech recognition?
Both have evolved from initial enthusiasm for simplistic approaches to more rigorous, data-driven methods.
What was the initial approach to machine translation in the 1950s?
Based on sequences of words, with models learned according to information theory principles.
What was the focus of neural networks in the 1980s?
To explore capabilities and differences from traditional techniques.
What has the work on neural networks allowed for?
Comparison with techniques from statistics, pattern recognition, and machine learning.
What did Judea Pearl’s 1988 work contribute to AI?
New acceptance of probability and decision theory in AI.
What major problem does the Bayesian network approach overcome?
Problems of probabilistic reasoning systems from the 1960s and 1970s.
What areas have seen revolutions similar to AI?
Robotics, computer vision, and knowledge representation.
What has led to significant benefits in AI recently?
Reintegration of isolated subfields and tools from machine learning.
What has recent progress in robotic cars derived from?
Better sensors, control-theoretic integration, high-level planning
What do Banko and Brill show about using more data?
Performance increases exceed algorithm choice differences
What is the purpose of the algorithm defined by Hays and Efros?
To find something that matches the background in a photograph after masking out an object.
What does the 'knowledge bottleneck' in AI refer to?
The problem of expressing all the knowledge a system needs.
What is suggested to solve the 'knowledge bottleneck' in many applications?
Learning methods rather than hand-coded knowledge engineering.
What did Kurzweil say about AI applications?
Many thousands of AI applications are deeply embedded in industry infrastructure.
What does automated speech recognition do for travelers?
Guides the entire conversation when booking a flight.
What was NASA’s Remote Agent program?
The first on-board autonomous planning program for spacecraft.
How do learning algorithms help in spam fighting?
They classify over a billion messages as spam daily.
What challenge do spammers present to static programmed approaches?
Spammers continually update their tactics, making it hard to keep up.
What was significant about logistics planning during the Persian Gulf crisis of 1991?
U.S. forces deployed a Dynamic Analysis for effective logistics.
What tool did U.S. forces use for logistics planning during the Persian Gulf crisis of 1991?
Dynamic Analysis and Replanning Tool (DART)
What did DARPA state about the investment in AI for DART?
It paid back DARPA’s 30-year investment in AI.
What is the purpose of the PackBot used by iRobot?
Handle hazardous materials, clear explosives, identify snipers.
What statistical model is used in the machine translation program?
Built from examples of Arabic-to-English translations and English text totaling two trillion words.
What philosophical idea made AI conceivable?
The mind is like a machine that operates on knowledge encoded in an internal language.
What did mathematicians provide for AI development?
Tools to manipulate logical and probabilistic statements.
What insights have neuroscientists provided for AI?
Facts about how the brain works and its similarities/differences from computers.
What perspective do psychologists adopt regarding humans and animals?
They consider them as information-processing machines.
What role do computer engineers play in AI?
They provide powerful machines that enable AI applications.
What does control theory deal with in relation to AI?
Designing devices that act optimally based on environmental feedback.
What has the history of AI experienced over time?
Cycles of success, optimism, cutbacks, and new creative approaches.
What has advanced AI rapidly in the past decade?
Greater use of the scientific method in experimenting and comparing approaches.
What has improved alongside the understanding of intelligence?
Improvements in the capabilities of real systems.
Which book discusses the methodological status of AI?
The Sciences of the Artificial by Herb Simon (1981).
Who criticized the usefulness of the Turing Test?
Shieber (1994) criticized its instantiation in the Loebner Prize competition.
What are the major AI conferences?
What are the major journals for general AI?
What are the main professional societies for AI?
What is the purpose of the exercises in the text?
To stimulate discussion and possibly set as term projects.
What does the exercise 1.1 ask for?
Define in your own words: intelligence, artificial intelligence, agent, rationality, logical reasoning.
What prediction did Turing make about AI by the year 2000?
A computer would have a 30% chance of passing a five-minute Turing Test with an unskilled interrogator.
What does exercise 1.3 question about reflex actions?
Are reflex actions rational? Are they intelligent?
Are reflex actions rational or intelligent?
They are not considered rational or intelligent; they are instinctive.
What happens if a program scores 200 on an IQ test?
It doesn't necessarily mean the program is more intelligent than a human.
Why is the sea slug Aplysia studied?
It has about 20,000 neurons that are large and easily manipulated.
What is the significance of neuron cycle time in Aplysia?
Assuming it's similar to humans, it allows for computational comparisons.
How can introspection be inaccurate?
Individuals may misreport their inner thoughts or be unaware of them.
What are examples of artificial intelligence systems?
What complex operations do cognitive models involve?
Operations like convolving an image with a Gaussian or minimizing entropy.
Can computers be intelligent?
They can only perform tasks as programmed, raising questions about true intelligence.
Can animals be intelligent?
They operate based on genetic programming, but this doesn't negate intelligence.
What can be said about intelligence and physical laws?
All entities act according to physical laws, but this does not define intelligence.
What tasks can computers currently perform?
What should be examined regarding AI contests?
Progress made and their impact on the state of AI and new ideas.
What is an agent?
Anything that perceives its environment through sensors and acts through actuators
How is an agent's behavior mathematically described?
By the agent function that maps percept sequences to actions
What is the difference between agent function and agent program?
Agent function is an abstract mathematical description; agent program is a concrete implementation
What is the vacuum-cleaner world described in the text?
A simple world with two locations (squares A and B) where a vacuum agent can perceive its location and dirt status.
What is a simple agent function for the vacuum agent?
If the current square is dirty, then suck; otherwise, move to the other square.
What does a rational agent do?
A rational agent does the right thing according to its environment and percepts.
How is good behavior defined for an agent?
By the desirability of the sequence of actions it generates in its environment.
What is a performance measure?
It evaluates the desirability of a sequence of environment states based on the agent's actions.
What can affect an agent's perception of its own performance?
An agent might delude itself about its performance leading to a flawed self-assessment.
What is the main focus of AI according to the text?
Designing artifacts with significant computational resources for nontrivial decision-making.
What is a performance measure in the context of agents?
A metric designed by a designer to evaluate the success of an agent's actions in specific circumstances.
How might a vacuum-cleaner agent maximize its performance measure?
By cleaning dirt, dumping it on the floor, and cleaning it again.
What is a more suitable performance measure for a vacuum-cleaner agent?
Reward the agent for each clean square and penalize for electricity and noise.
What is the definition of a rational agent?
An agent that selects actions to maximize its performance measure based on percept sequences and built-in knowledge.
What four factors determine what is rational at any time?
What is the performance measure for the vacuum-cleaner agent in the example?
One point for each clean square at each time step over 1000 time steps.
What does the vacuum-cleaner agent do if it finds a dirty square?
Cleans the square and moves to the other square if it's clean.
Under what condition is the vacuum-cleaner agent considered rational?
When its expected performance is at least as high as any other agent's under given circumstances.
What happens to the vacuum-cleaner agent when all dirt is cleaned?
It oscillates back and forth needlessly.
What penalty might affect the performance measure of the vacuum-cleaner agent?
A penalty of one point for each movement left or right.
What happens when all the dirt is cleaned up by the agent?
The agent will oscillate needlessly back and forth.
What should an agent do if clean squares can become dirty again?
The agent should occasionally check and re-clean them if needed.
What is the difference between rationality and perfection?
Rationality maximizes expected performance; perfection maximizes actual performance.
What is a key aspect of rational behavior when crossing a busy road?
An agent should look both ways before crossing.
What is information gathering in the context of rationality?
Actions taken to modify future percepts.
What must a rational agent do according to the definition?
Gather information and learn from what it perceives.
What happens if an environment is completely known a priori?
The agent need not perceive or learn; it simply acts correctly.
What happens if the dung ball is removed from the dung beetle?
It continues pantomiming plugging the nest without noticing the dung ball is missing.
What is the behavior of the sphex wasp during its nesting process?
Digs a burrow, stings a caterpillar, drags it inside, and lays eggs.
What does the sphex wasp do if the caterpillar is moved?
Reverts to the dragging step and continues without modification.
What does it mean for an agent to lack autonomy?
It relies on prior knowledge of its designer rather than its own percepts.
What is an example of a rational agent that can learn?
A vacuum-cleaning agent that learns to foresee where dirt will appear.
Why is the automated taxi driver considered a complex problem?
It has open-ended tasks with novel combinations of circumstances.
What must be specified when designing an agent?
The task environment, including performance measure, environment, actuators, and sensors.
What does PEAS stand for in the context of an automated taxi?
Performance Measure, Environment, Actuators, Sensors
What are the performance measures for the automated taxi?
What are the basic sensors for an automated taxi?
What are some desirable qualities for the automated taxi's performance?
What optional choices may affect the taxi's driving environment?
What is the significance of the complexity of the relationship in agent environments?
It affects the agent's behavior, percept sequence, and performance measure.
What distinguishes real environments from artificial ones in AI?
The complexity of the relationship among agent behavior, percept sequence, and performance measure.
What is an example of a simple 'real' environment for a robot?
A robot inspecting parts on a conveyor belt.
What is a softbot designed to do?
What abilities does a softbot need to operate effectively?
What dimensions categorize task environments in AI?
What is a fully observable environment?
An environment where sensors access the complete state at all times.
What causes a task environment to be partially observable?
What is the difference between single-agent and multiagent environments?
What is a single-agent environment?
An environment where an agent, like one solving a crossword puzzle, operates independently.
What is a two-agent environment?
An environment where two agents interact, such as in a game of chess.
What distinguishes an agent from an object?
An agent's behavior maximizes a performance measure based on another entity's behavior.
What is a competitive multiagent environment?
An environment where agents, like in chess, compete to maximize their performance measures at the expense of others.
What is a partially cooperative multiagent environment?
An environment where agents work together to avoid collisions, like in taxi driving, while also competing for resources.
What defines a deterministic environment?
An environment where the next state is completely determined by the current state and the agent's action.
What defines a stochastic environment?
An environment where the next state cannot be precisely predicted due to uncertainty.
What is an uncertain environment?
An environment that is either not fully observable or not deterministic.
What is a nondeterministic environment?
An environment characterized by possible outcomes of actions without attached probabilities.
What is an episodic task environment?
An environment where the agent's experience is divided into episodes, with each episode independent of others.
What is a sequential task environment?
An environment where the agent's actions in previous episodes affect future episodes.
What is the difference between sequential and episodic environments?
In sequential environments, current decisions affect future decisions. In episodic environments, decisions are independent of past actions.
What defines a static environment?
A static environment does not change while an agent is deliberating, allowing the agent to focus on decision-making without external changes.
What is a dynamic environment?
A dynamic environment changes while an agent is deliberating, requiring continuous decision-making.
What is a semidynamic environment?
A semidynamic environment does not change over time, but the agent's performance score may change.
How do discrete and continuous environments differ?
Discrete environments have a finite number of states, while continuous environments have states that vary smoothly over time.
What characterizes a known environment?
In a known environment, the outcomes for all actions are known, allowing the agent to make informed decisions.
What is an unknown environment?
An unknown environment requires the agent to learn how it works, as the outcomes of actions are not fully understood.
What is the hardest type of environment?
The hardest environment is partially observable, multiagent, stochastic, sequential, dynamic, continuous, and unknown.
How does taxi driving exemplify a dynamic environment?
Taxi driving is dynamic because the environment changes continuously with moving vehicles and requires ongoing decision-making.
What is an example of a static environment?
Crossword puzzles are considered static as they do not change once the puzzle is set.
What type of agents are involved in a crossword puzzle task environment?
Fully Single Deterministic Sequential Static Discrete
What type of agents are involved in chess with a clock?
Fully Multi Deterministic Sequential Semi Discrete
What type of agents are involved in taxi driving?
Partially Multi Stochastic Sequential Dynamic Continuous
What type of agents are involved in medical diagnosis?
Partially Single Stochastic Sequential Dynamic Continuous
What type of agents are involved in image analysis?
Fully Single Deterministic Episodic Semi Continuous
What type of agents are involved in a part-picking robot?
Partially Single Stochastic Episodic Dynamic Continuous
What type of agents are involved in a refinery controller?
Partially Single Stochastic Sequential Dynamic Continuous
What type of agents are involved in an interactive English tutor?
Partially Multi Stochastic Sequential Dynamic Discrete
What does the code repository include for environments?
Implementations and a general-purpose environment simulator
What is the purpose of running many simulations in different traffic conditions?
To evaluate a taxi driver in simulated traffic
What does a rational agent maximize in an environment?
Average performance over the environment class.
What does the architecture provide to the program?
Percepts from sensors, runs the program, feeds actions to actuators.
What is the difference between agent program and agent function?
Agent program takes current percept; agent function takes entire percept history.
What is the purpose of the TABLE-DRIVEN-AGENT program?
To track percept sequence and decide actions based on a table of actions.
Why is the table-driven approach to agent construction problematic?
It requires a massive lookup table for all possible percept sequences, which is impractical.
What is the persistent variable in the TABLE-DRIVEN-AGENT program?
percepts, a sequence initially empty.
What does the action variable in the TABLE-DRIVEN-AGENT program represent?
The action returned based on the lookup from the percept sequence table.
What is the key challenge for AI?
To produce rational behavior from a small program rather than a vast table
What are the four basic kinds of agent programs?
What does a simple reflex agent base its actions on?
The current percept, ignoring the rest of the percept history
What are condition–action rules also known as?
They are also called situation–action rules, productions, or if–then rules.
What is the function of SIMPLE-REFLEX-AGENT?
It acts according to a rule whose condition matches the current state from the percept.
What does the INTERPRET-INPUT function do?
It generates an abstracted description of the current state from the percept.
What is the limitation of simple reflex agents?
They are of limited intelligence and work only if the environment is fully observable.
What happens if the car in front has different taillight configurations?
A simple reflex agent may either brake unnecessarily or fail to brake at all.
What are the possible percepts for a simple reflex vacuum agent with only a dirt sensor?
The percepts are [Dirty] and [Clean].
What happens if a vacuum agent starts in square A and tries to move Left?
Moving Left fails forever if it starts in square A.
What happens if a vacuum agent starts in square B and tries to move Right?
Moving Right fails forever if it starts in square B.
What happens if a simple reflex agent starts in square A and receives [Clean]?
Moving Left fails (forever).
What happens if a simple reflex agent starts in square B and receives [Clean]?
Moving Right fails (forever).
How can escape from infinite loops be achieved in simple reflex agents?
By randomizing actions, such as flipping a coin to choose between Left and Right.
What is the main benefit of a randomized simple reflex agent over a deterministic one?
It might outperform a deterministic simple reflex agent in some situations.
What is a model-based reflex agent?
An agent that keeps track of the part of the world it can’t see using an internal state.
What does the internal state of a model-based reflex agent depend on?
The percept history reflecting unobserved aspects of the current state.
What is an example of internal state for a braking problem?
The previous frame from the camera detecting red lights.
What knowledge is required for updating the internal state of a model-based agent?
Knowledge of how the world evolves and how the agent's actions affect the world.
What is a model of the world in the context of AI agents?
Knowledge about how the world works, used by model-based agents.
What does the function UPDATE-STATE do in a model-based reflex agent?
It updates the agent's internal state based on current percept and action.
What is the purpose of the MODEL-BASED-REFLEX-AGENT function?
To return an action based on the agent's current state and rules.
What is the purpose of the RULE-MATCH function?
To match the current state with rules and determine the action to take.
What does a model-based reflex agent do?
Keeps track of the current state of the world and chooses actions based on that state.
What is a key characteristic of model-based agents?
They maintain an internal model of the world to guide decision-making.
What is often unavoidable in partially observable environments?
Uncertainty about the current state of the environment.
What does the box labeled 'what the world is like now' represent?
The agent's best guess about the current state of the environment.
What is an example of uncertainty in an automated taxi?
Not being able to see around a large truck blocking the view.
How does a goal-based agent choose its actions?
By selecting actions that will lead to the achievement of its goals.
What aspect of the world state might a taxi consider when deciding its actions?
Its destination, which is part of the agent's internal state.
Are you sure you want to delete 0 flashcard(s)? This cannot be undone.
Select tags to remove from 0 selected flashcard(s):
Loading tags...