Class 1: Introduction

August 30, 2004

Definition

AI deals with problems that appear to require “intelligence” to be solved.

Examples:

• Database
• Find student names
• Find buying patterns
• Game
• Chess
• Checkers
• Monopoly
• Packman/shooting, etc.
• Forms
• Bubble sheet
• Handwritten form

Other definitions:

• Systems that think like humans.
• The automation of activities that we associate with human thinking such as decision-making, problem solving, learning.
• Systems that think rationally.
• The study of computations that that make it possible to perceive, reason, and act.
• Systems that act like humans.
• The study of how to make computers do things at which, at the moment, people are better.
• Systems that act rationally.
• The study of the design of the intelligent agents.

AI History

• The gestation of AI (1943-1956)
• McCulloch and Pitts neurons (1943) drew on three sources: knowledge of basic physiology and function of neurons; formal analysis of propositional logic; and Turing’s theory of computation.
• Claude Shannon (1950) and Alan Turing (1953) were writing chess programs on conventional computers. Alan Turing once proposed an intelligence test for computer programs. In one variant of this test a human judge is allowed to interrogate a program through some sort of an interface. If the program can fool the human into believing it is another human responding rather than a computer, then the program is judged intelligent.
• The first neural network computer SNARC was built in 1951. This was the basis of Marvin Minsky’s Ph.D. thesis at the Princeton mathematics department.
• John McCarthy moved from Princeton to Dartmouth College and organized a two-month AI workshop there in 1956.
• Allen Newell and Herbert Simon from Carnegie Tech (now Carnegie Mellon University) presented a theorem-proving program called Logic Theorist.
• Early enthusiasm, great expectations (1952-1969)
• Newell and Simon followed LT with the General Problem Solver, or GPS. It was designed to imitate human problem-solving protocols. The order in which the program considered subgoals and possible actions was similar to the way humans approached the same problems. It was the first program to employ the “thinking humanly” approach.
• John McCarthy moved from Dartmouth to MIT and designed Lisp in 1958. Marvin Minsky moved to MIT in the same year. They defined the AI field together, but grew apart as McCarthy stressed representation and reasoning in formal logic, whereas Minsky was more interested in getting programs to work.
• Minsky supervised students who chose limited problems that appeared to require intelligence to solve known as microworlds. Examples:
• SAINT by James Slagle (1963) solved closed-form integration problems.
• ANALOGY by Tom Evan (1968) solved geometric analogy problems that appear in IQ tests.
• STUDENT by Daniel Bobrow solved algebra story problems.
• Neural networks research also continued. The work by Winograd and Cowan (1963) showed how a large number of elements could represent an individual concept; Frank Rosenblatt (1962) enhanced learning methods with perceptrons.
• A dose of reality (1966-1974)
• Herbert Simon in 1957: “It is not my aim to surprise or shock you – but the simplest way I can summarize is to say that there are now in the world machines that think, that learn, and that create. Moreover, their ability to do these things is going to increase rapidly until—in visible future—the range of problems they can handle will be coextensive with the range to which human mind has been applied”.
• The barrier that faced almost all AI research was that methods that sufficed for demonstration on one or two simple examples turned out to fail miserably on wider selections of problems. Most early programs contained little or no knowledge of their subject matter; they succeeded by means of simple syntactic manipulation.
• The second kind of difficulty was the intractability of many problems that AI was attempting to solve. Before the theory of NP-completeness was developed it was thought that “scaling up” means more powerful hardware and clever software.
• Lighthill report (1973) was the basis for the decision by British government to end support for AI research.
• Knowledge-based systems: The key to power? (1969-1979)
• Weak methods: general-purpose search algorithms to string together elementary reasoning methods. Their performance was weak for many complex domains. The only way around it is to use more specific knowledge to make larger reasoning steps.
• Ed Feigenbaum and others at Stanford began the Heuristic Programming Project (HPP) to investigate the extent to which the new methodology of expert systems could be applied. MYCIN was developed to diagnose blood infections. With about 450 rules MYCIN was able to perform as well as some experts, and considerably better than junior doctors.
• Demand increased for workable knowledge representation schemes. Prolog language became very popular and was based on formal logic. Minsky’s idea of frames (1975) adopted a more structured approach, assembling facts about a particular object and event types and arranging the types into a large taxonomic hierarchy.
• AI becomes an industry (1980-present)
• The first successful commercial expert system R1, began operation at Digital Equipment Corporation in 1982 (McDermott). The program helped configure orders for new computer systems, and by 1986, it was saving the company an estimated $40 million a years. DuPont had 100 expert systems in use by 1988.
• In 1981 the Japanese announced the “Fifth Generation” project, a 10 year project to build intelligent computers running Prolog as their machine language.
• Computer scientists neglected neural networks, but the work continued in other fields, particularly in physics. Hopfield (1982) used techniques from statistical mechanics to analyze the storage and optimization properties of networks. The reinvention of back-propagation algorithm caused the explosion of interest in neural networks (paper by Rumelhart and McClelland in 1986).
• AI becomes a science (1987-present)
• It is now more common to build on existing theories than to propose brand new ones, to base claims on rigorous theorems or hard experimental evidence, and to show relevance to real-world applications.
• Speech recognition problems use hidden Markov models based on rigorous mathematical theory.
• The belief network formalism was invented to allow efficient reasoning about the combination of uncertain evidence. This approach has come to dominate AI research on uncertain reasoning and expert systems.
• Neural networks field improved methodology and theoretical frameworks to arrive at an understanding in which neural nets can be compared with corresponding techniques from statistics, pattern recognition, and machine learning, and the most promising technique can be applied to each application. As a result of these developments, the data mining technology has spawned a vigorous new industry.
• Similar gentle revolutions occurred in robotics, computer vision, machine learning, and knowledge representation.
• The emergence of intelligent agents (1995 - present)
• An agent is something that acts. A rational agent is one that acts so as to achieve the best outcome.
• One of the most important environments for intelligent agents is the Internet (the "-bot" suffix has entered everyday language). AI technologies underlie many Internet tools such as search engines, recommender systems, and Web site construction systems.

Representation Example

Suppose we want to find out if a word exists in a text. In C++ we write:

bool word_exists(const string& word, const string& text)

{

stringstream ss(text);

string s;

bool flag = false;

while(ss)

{

ss >> s;

if (s == word)

{

flag = true;

break;

}

}

return flag;

}

In Lisp we write the following.

(defun word_exists (word text)

(if (null text)

NIL

(if (eql (first text) word)

T

(word_exists word (rest text))

)

)

)