Artificial Intelligence

Enrique Pelez Ph.D.

Agenda

Definition of Artificial Intelligence Fundamentals of Artificial Intelligence History of Artificial Intelligence Why Artificial Intelligence?

Definition of Artificial Intelligence

AI is part of the Computational Sciences, focus on the creation of HW and SW with intelligent behaviour. AI refers to the techniques that allow systems to perform intelligent perception, reasoning and behaviour. AI looks for techniques to make the machines outperform tasks, that in some cases, are still better performed by humans.

Definition of Artificial Intelligence

Long term objectives

To develop systems that reach the level of intelligence similar/comparable/better than humans. It is possible to occur within the next 10-15 years
In specific tasks: To develop systems that reach levels of intelligence similar/comparable/better than humans Today there exist several systems that actually outperform humans, in some orders of magnitude.

Short term objectives

Definition of Artificial Intelligence

Within the context of the short term objectives:

AI is not looking for simulate human intelligence. But, it looks for reproducing the effects of an intelligent behaviour.

Analogy with flying

Definition of Artificial Intelligence

Applications in AI

Everyday tasks:

Perception (vision y talk)

Formal tasks:

Expert tasks:

Applications in AI

Learning:

Automatic knowledge acquisition. Knowledge based systems. Intelligent data bases. Games and Theorem provers. Natural language comprehension. Visual interpreters (Computer Vision). To perform mechanical and manual processes (Robotics). To generate, verify, depurate and automatic optimization of programs.

Reasoning:

Perception:

Locomotion and Manipulation:

Creation:

Fundamentals in AI

During the last 2000 years, Philosophy has contributed with diverse and multiple theories about reasoning and learning. During the last 400 years, Math has contributed with formal theories related to logic, probability, decision and computation theory. Math has provided the tools to manipulate certainty in logic, as well as uncertainty in probability. Math has also put the grounds for managing reasoning using algorithms. Psychology also has contributed with tools to allow current research about the human brain, and has provided the scientific language for expressing the theories. Physiologists have reinforce the idea that humans and other animals might be modelled as machines for processing information.

Fundamentals in AI

Linguistics also has offered theories about the structure and meaning of language. Linguists have demonstrated that the use of a language also follows a structure. Computer Science has provided the computational tools for making of AI a reality. Computer Engineering has also contributed to the field, providing solutions for making AI systems a reality. AI programs are generally big and will not perform without the speed, memory capacity and computational power of todays devices.

Fundamentals in AI

The sciences that contribute to the development of AI are:

Philosophy Medicine Mathematics Psychology Linguistics Computer Science and Engineering

History of Artificial Intelligence

1950-1965. Classic Period'

Harvesting [McColluck y Pitts, Shannon, Turing] 1956, AI has its impulse after the Conference in Darmouth College. McCarty named the field as AI. Minsky (founder of MITs AI lab), Shannon (Bell labs), Rodhester (IBM), Newell (president of A.A.A.I. American Association of AI), Simmon (Nobel laureate from Carnegie Mellon University) Neural Nets, robotics (Shakey) State search, Heuristcs, LISP

History of Artificial Intelligence

Turing Test

History of Artificial Intelligence

1950-1965. Classic Period

General Problem Solvers (GPS) [Newell, Simon]. Common Sense problems, which include reasoning about physical objects and their relationships, reasoning about actions and its consequences. Simple tasks only were resolved, because there were not enough knowledge about a specific domain. Main focus in implementation of games (chess, backgammon), as well as math theorems demos. Limitations in searching, combinatorial explosion.

History of Artificial Intelligence

1965-1975. Romantic Period

General representation of knowledge Semantic nets [Quillian] Prototyping (frames) [Minsky] Perceptron [Minsky y Papert] Logic [Kowalski] Block worlds [Winograd] Perception (vision y speaking), natural language comprehension, robotics.

History of Artificial Intelligence

1975-Today. Modern Period, Specific Intelligence vs. General''.

The need for collaboration with other knowledge fields has been identified. Explicit representation of very specific knowledge domains. Expert System MYCIN (Stanford University) Knowledge based Expert Systems. Neural Nets are back[Hopfield, Rumelhart, Hinton], genetic algorithms [Holland, Goldberg] Voice recognition, uncertainty (Bayesian Nets, Fuzzy Logic), planning, learning. Real applications (medicine, finance, engineering, exploration, etc.). Commercialization of AI.

Why Artificial Intelligence?

Two purposes

To use the computational resources for empowering the human thinking and development; just like we use motors to increase the strength or power. Robotics and the Expert Systems are 2 examples.
To understand how humans think. Not only to figure it out what is, but also how it is done. The cognitive sciences is an example.