CST-301

Introduction to Artificial Intelligence

Wolfgang Ertel
What is Artificial Intelligence?

The power of a machine to copy intelligent human behavior.

- ability
Intelligence  to understand and study something
- not automatic
- think and understand

Faculty of Computer Science 2

John McCarthy (1955):
• The aim of AI is to develop machines that behave as if they
were intelligent.

• Two simple Braitenberg-vehicles and their reaction to a light

source. (by very simple electrical circuits)

Encyclopedia Britannica:
• AI is the ability of a digital computer or computer-
controlled robot to solve problems that are normally
associated with the higher intellectual capabilities of
humans…
• According to this definition, every computer is an AI-system.
Faculty of Computer Science 3
Elaine Rich:
• Artificial Intelligence is the study of how to make computers do
things at which, at the moment, people are better.
• Still up-to-date in the year 2050!
• Humans are still better in many fields (e.g. understanding pictures,
learning ability)!
• Computers are already better in many fields (e.g. playing chess)!

 AI is only concerned with the pragmatic implementation of intelligent

processes. (dangerous to conclude)
 deep understanding of human reasoning and intelligent action
 Human intelligence is adaptively – adjusting to various environmental
conditions and change the behavior through learning. (Machine
learning – central subfield of AI)

Faculty of Computer Science 4

 Brain Science and Problem
Solving
• Intelligent system – try to understand how the human brain
works and model/ simulate it on the computer.
• starting from a problem and trying to find the most optimal
solution
• AI offers a broad palette of effective solutions for widely
varying applications. (no universal method)
• Cognitive science – research into human thinking at a
somewhat higher level
• Humans have consciousness. How does it come to be?
• Mind and consciousness are linked with matter, that is with the
brain.

Faculty of Computer Science 5

 Turing Test and Chatterbots
Alan Turing:

• The machine passes the test, if it can mislead Alice in 30% of the
cases.
• Chatterbots – initial responses are quite impressive. Their artificial
nature becomes apparent after a certain amount of time.
- e.g., online customer support, e-learning system
Faculty of Computer Science 6
 History of AI
• 1931 The Austrian Kurt Godel shows that in first-order predicate
logic all true statements are derivable. In logics, on the other hand,
there are true statements that are unprovable.
• 1937 Alan Turing points out the limits of intelligent machines with the
halting problem .
• 1943 McCulloch und Pitts model neural networks and make the
connection to propositional logic.
• 1950 Alan Turing defines machine intelligence with the Turing test
and writes about learning machines and genetic algorithms
• 1951 Marvin Minsky develops a neural network machine. With 3000
vacuum tubes he simulates 40 neurons.
• 1955 Arthur Samuel (IBM) builds a learning chess program that plays
better than its developer.

Faculty of Computer Science 7

• 1956 McCarthy organizes a conference in Dartmouth College. Here the
name Artificial Intelligence was first introduced. Newell and
Simon of Carnegie Mellon University (CMU) present the Logic Theorist,
the first symbol-processing computer program.
• 1958 McCarthy invents at MIT the high-level language LISP. He
writes programs that are capable of modifying themselves.
• 1959 Gelernter (IBM) builds the Geometry Theorem Prover.
• 1961 The General Problem Solver (GPS) by Newell and Simon imitates
human thought.
• 1963 McCarthy founds the AI Lab at Stanford University.
• 1965 Robinson invents the resolution calculus for predicate logic.
• 1966 Weizenbaum's program Eliza carries out dialogue with people in
natural language.
• 1969 Minsky and Papert show in their book Perceptrons that the
perceptron, a very simple neural network, can only represent linear
Faculty of Computer Science 8
functions.
• 1972 French scientist Alain Colmerauer invents the logic programming
language PROLOG.
British physician de Dombal develops an expert system for diagnosis of
acute abdominal pain. It goes unnoticed in the mainstream AI community of
the time.
• 1976 Shortlie and Buchanan develop MYCIN, an expert system for
diagnosis of infectious diseases, which is capable of dealing with
uncertainty.
• 1981 Japan begins, at great expense, the Fifth Generation Project with
the goal of building a powerful PROLOG machine.
• 1982 R1, the expert system for configuring computers, saves Digital
Equipment Corporation 40 million dollars per year .
• 1986 Renaissance of neural networks through, among others, Rumelhart,
Hinton and Sejnowski. The system Nettalk learns to read texts aloud.

Faculty of Computer Science 9

• 1990 Pearl , Cheeseman , Whittaker, Spiegelhalter bring probability theory
into AI with Bayesian networks. Multi-agent systems become popular.
• 1992 Tesauros TD-gammon program demonstrates the advantages of
reinforcement learning.
• 1993 Worldwide RoboCup initiative to build soccer-playing autonomous
robots .
• 1995 From statistical learning theory, Vapnik develops support vector
machines, which are very important today.
• 1997 First international RoboCup competition in Japan.
• 2003 The robots in RoboCup demonstrate impressively what AI and robotics
are capable of achieving.
• 2006 Service robotics becomes a major AI research area.
• 2010 Autonomous robots start learning their policies.

Faculty of Computer Science 10

 Agents
• a system that processes information and produces an output from
an input
• a program mapping from perceptions to actions
• Software agent – program that calculates a result from user input.

Faculty of Computer Science 11

• Hardware agent

• Reflex-Agent: function from the set of all inputs to the set of all outputs.
• Agent with a memory: is not a function.
• Agent capable of learning
• Distributed agents
• Markov decision process: only the current state is needed for the
determination of the optimal action.
• goal-oriented agent
Faculty of Computer Science 12
• Example: Spam filter: aims at assigning emails to their correct
classes.

Which agent is better?

cost-based agent – to minimize the long term cost (i.e. the average cost) caused
by wrong decisions. The sum of all weighted errors results in the total cost.
utility-based agent – to maximize the long-term benefit (i.e. the average
benefit) caused by correct decisions.

Faculty of Computer Science 13

Environment
• observable (chess computer) – The agent always knows the complete state
of the world.
• partially observable (robot)
• deterministic (8-puzzle) – An action always leads to the same result.
• non-deterministic (chess computer, robot)
• discrete (chess computer) – Only finitely many states and actions occur.
• continuous – infinitely many states or actions

Faculty of Computer Science 14

Knowledge-Based System

Faculty of Computer Science 15

Separation of knowledge and inference has advantages:
• Inference is application-independent (e.g. medical expert system).
• Knowledge can be stored declaratively.
Representation of knowledge with a formal language:
• Propositional logic
• First-order logic (shortly: FOL).
• Probabilistic logic
• Fuzzy logic
• Decision trees
Faculty of Computer Science 16
 Assignment
Examine the AI literature to discover whether the following tasks can
currently be solved by computers:
a. Playing a decent game of table tennis (ping-pong).

b. Driving in the center of Cairo.

c. Buying a week's worth of groceries at the market.

d. Buying a week's worth of groceries on the web.

e. Playing a decent game of bridge at a competitive level.

f. Discovering and proving new mathematical theorems.

g. Writing an intentionally funny story.

h. Giving competent legal advice in a specialized area of law.

i. Translating spoken English into spoken Swedish in real time.

j. Performing a complex surgical operation.

Faculty of Computer Science 17