Introduction to Artificial Intelligence

http://www-courses.cs.uiuc.edu/~cs440/
Tuesday and Thursdays 2:00 to 3:15pm, 1404 Siebel Center Jean Ponce ponce@cs.uiuc.edu 2065 Beckman Institute 333-8864

TA
Tuna Oezer (oezer@uiuc.edu) 1121 Siebel Office hours: Mon. 2:00-3:00 and Th. 1:00-2:00.

This Lecture
Introduction to AI Course overview Examples: Agents, robots, and vision
Next time: Problem solving and search

What is AI?
AI is the study of ideas that enable computers to be intelligent. [P. Winston]

So, what is intelligence?

Fast thinking?
Knowing a lot? Being able to pass as a smart human? Being able to reason?

Being able to learn?

Being able to perceive and act upon ones environment? Writing poetry? Passing an AI class?

Operational Definitions of AI?

Thinking Humanly
`The automation of activities that we associate with human thinking, activities such as decision-making, problem solving, learning [Bellman, 1978]

Thinking Rationally
The study of mental faculties through the use of computational models. [Charniak & McDermott, 1985]

Acting Humanly
The study of how to make computers do things at which, at the moment, people are better. [Rich& Knight, 1991]

Acting Rationally
The branch of computer science that is concerned with the automation of intelligent behavior. [Luger+Stubblefield, 1993]

Which do we choose?

Thinking Humanly: Cognitive Science

1960s cognitive revolution: Information-processing psychology replaced prevailing orthodoxy of behaviorism. Requires scientific theories of internal activities of the brain: What level of abstraction? ``Knowledge'' or ``circuits''? How to validate? Requires: 1. Predicting and testing behavior of human subjects (topdown); 2. Direct identification from neurological data (bottom-up). Both approaches (roughly, Cognitive Science and Cognitive Neuroscience) are now distinct from AI.

Acting humanly: The Turing test

Turing (1950) ``Computing machinery and intelligence'': Can machines think?'' or Can machines behave intelligently? Operational test for intelligent behavior: the Imitation Game.

Predicted that by 2000, a machine might have a 30% chance of fooling a lay person for 5 minutes. Anticipated all major arguments against AI in following 50 years. Suggested major components of AI: knowledge, reasoning, language understanding, learning. Problem: Turing test is not reproducible, constructive, or amenable to mathematical analysis.

Thinking rationally: Laws of Thought

Aristotle: what are correct arguments/thought processes? Several Greek schools developed various forms of logic: Notation and rules of derivation for thoughts.

Direct line through mathematics and philosophy to modern AI.

Problems: 1. Not all intelligent behavior is mediated by logical deliberation. 2. What is the purpose of thinking? What thoughts should I have?

Acting rationally
Rational behavior: Doing the right thing.
The right thing: That which is expected to maximize goal achievement, given the available information. Doesn't necessarily involve thinking---e.g., blinking reflex---but thinking should be in the service of rational action.

Is AI a hard problem?
The meaning of words and sentences
John drove his sister to buy groceries. John drove his sister to commit suicide. John drove his car to commit suicide. John drove his rabbit to buy groceries.

AI Prehistory
Philosophy Mathematics Psychology Linguistics Neuroscience Control Theory

AI History
1943 1950 1950s McCulloch & Pitts: Boolean circuit model of brain Turing's ``Computing Machinery and Intelligence'' Early AI programs, including Samuel's checkers program, Newell & Simon's Logic Theorist, Gelernter's Geometry Engine 1956 McCarthy organizes Dartmouth meeting and includes Minsky, Shannon, Newell, Samuel, Simon Name ``Artificial Intelligence'' adopted 1957 General Problem Solver [Newell, Simon, Shaw @ CMU] 1958 Creation of the MIT AI Lab by Minsky and McCarthy 1958 LISP, [McCarthy], second high level language (MIT AI Memo 1) 1963 Creation of the Stanford AI Lab by McCarthy 1965 Robinson's complete algorithm for logical reasoning 1966-74 AI discovers computational complexity 1966-72 Shakey, SRIs Mobile Robot [Fikes, Nilson]

AI History (Cont.)
1969 1969-79 1970 1971 1980-88 1981 Publication of Perceptrons [Minsky & Papert], Neural network research almost disappears Early development of knowledge-based systems SHRDLU, Winograds natural language system MACSYMA, an symbolic algebraic manipulation system Expert systems industry booms Japan: Fifth generation project US: Microelectronics and Computer Technology Corp. UK: Alvey Expert systems industry busts: ``AI Winter'' Neural networks return to popularity Resurgence of probabilistic and decision-theoretic methods Computational learning theory ``Nouvelle AI'': ALife, GAs, soft computing

1988-93 1985-95 1988-

AI research and its spinoffs

AAAI Conference IJCAI Conference AI Journal

Spinoffs
Robotics (ICRA, ISRR, IROS) Computer Vision (ICCV, CVPR, ECCV) Neural Networks (NIPS, ) Machine Learning (MLS, ) Speech Natural language understanding

Text: Stuart Russell and Peter Norvig, Artificial Intelligence: A Modern Approach (2nd edition), Prentice-Hall, 2003.
Grading: Problem Sets and MPs (in C): 45% Midterm: 20% Final Exam: 35% Newsgroup: http://www-courses.cs.uiuc.edu/~cs440/

Graduate students
Will be graded on a different curve than undergraduates. They are expected to do better for the same grade.
May receive or 1 unit of credit. To receive 1 unit they will have to do a programming project, and should contact my before the Spring break so we can pick a topic.

Syllabus
Introduction (Ch. 1) Problem solving and search (Ch. 3) Informed search methods (Ch. 4) Logical agents (Ch. 7) First-order logic and inference (Ch. 8 and 9) Resolution and planning (Ch. 9 and 11) Uncertainty (Ch. 13) Bayesian networks (Ch. 14) Learning (Ch. 18 and 19) Neural networks (Ch. 20) Vision and robotics (Ch. 24 and 25)

Policies
Cheating: You are expected to do all of the work on your own. You may discuss concepts with your classmates, but the homeworks must be done on your own. The penalty for cheating on any assignment is straightforward. On the first occurrence, you will receive a zero for the assignment, and then our course grade will be reduced by one full letter grade. A second occurrence will result in course failure. Late homework: Unless announced in advance, solutions will be posted no sooner than two days after the due date. Homework will be accepted until that point with a penalty of 10% per day that it is late. No assignments will be accepted after the solutions have been posted. To time stamp them, late homeworks will only be accepted in class, during office hours, or electronically by email to the TA's.

Intelligent Agents

An agent is an entity that perceives and acts. Abstractly, an agent is a function from percept histories to actions:

For any given class of environments and tasks, we seek the agent (or class of agents) with the best performance.

Designing an agent: (e.g. a taxi)

Percept, Action, Goal, Environment (PAGE) Percepts Video, accelerometers, engine sensors, keyboard, voice, sound, GPS, Actions Steer, accelerate, brake, horn, speak/display, Goals Safety, reach destination, maximize profits, obey laws, passenger comfort, Environment US urban streets, freeways, traffic, pedestrians, weather, customers,

A more realistic goal: programming a robot to move boxes

Percepts Vision, haptics, hearing Actions Walk, grab, lift, drop Goals Stack boxes at destination Environment 1404 Siebel

Robots

Honda Humanoid Robots

P2 P1 Asimo

Ten years $50,000,000

Honda Asimo

http://world.honda.com/ASIMO/

Sony QRIO

http://www.sony.net/SonyInfo/QRIO/

Toyota trumpet playing robot

http://www.toyota.co.jp/en/special/robot/

Vision

(Nalwa, 1993)

A challenge: object recognition

What we can do today (Rothganger et al. 2004)

ILM

Courtesy of S. Leigh

What is it all for?