2CS24 - TOPICS IN INFORMATION PROCESSING

ARTIFICIAL INTELLIGENCE

Contents:

  1. Definition.
  2. Techniques.
  3. Development.

See also Trevor Bench-Capon's 2cs64 KNOWLEDGE BASED SYSTEMS Course materials.




ARTIFICIAL INTELLIGENCE (AI)?

ARTIFICIAL - Man made, embodied in a machine (e.g. a computer).

INTELLIGENCE - Must relate to tasks involving "higher mental processes", not:

"AI is the study of ideas that enable computers to be intelligent" - Patrick Winston.

The term is somewhat controversial - and is used differently by different practitioners with different motivation.


"HARD" OR "STRONG" AI

The aim here is to build a machine which really does think:

"The ultimate goal of AI research ... is to build a person, or more humbly an animal" - Charniak and McDermott.

Perhaps more common in the US, this approach raises some deep philosophical questions. Criteria for success are difficult to determine, although proposals have been made (eg. the Turing Test, the behaviour of the system cannot be distinguished from that of a human being).


"SOFT" OR "WEAK" AI

Aim is to enable different kinds of applications to be effectively performed by computers.

"AI is the study of how to make computers do things, at which, at the moment, people are better" - Elaine Rich.

AI is concerned with "computer systems with the characteristics we associate with human behaviour - understanding language, reasoning, solving problems, and so on" - Barr and Feigenbaum.

It is the tasks that are important - the philosophical questions are avoided.


COGNITIVE SCIENCE

Aim here is to develop, explore and evaluate theories of how the mind works through the use of computational models:

"AI is the study of mental faculties through the use of computational models" - Charniak and McDermott.

Not what is done but how it is done that it is important - intelligent behaviour is not enough, the program must operate in an intelligent manner.

Chess programs are successful, but tell us little about the ways humans play chess




AI TECHNIQUES


1. DESCRIBE AND MATCH


2. GOAL REDUCTION

3. CONSTRAINT SATISFACTION TECHNIQUES


4. TREE SEARCHING

Depth-firstHill climbing
Breadth-firstBeam
Best-first

DEPTH-FIRST SEARCH

HILL CLIMBING

BREADTH-FIRST SEARCH

BEAM SEARCH

BEST-FIRST SEARCH


5. GENERATE AND TEST


6. RULE-BASED SYSTEMS




DEVELOPMENT

PROBLEM SOLVING AND UNDERSTANDING KNOWLEDGE

The engineering of AI systems requires a complete understanding of the knowledge underpinning a problem. Unlike more main stream software applications (e.g. data processing, number crunching etc.) the nature of this knowledge is often not obvious. Questions that must be answered include:

AI software engineering processes therefore require the inclusion of knowledge acquisition/elicitation and analysis phases.


REPRESENTATION


IMPLEMENTATION




SEMINAR TITLE

"APPLICATIONS OF AI"

Some example applications include:




Created and maintained by Frans Coenen. Last updated 09 October 2002