next up previous
Next: Main references
Up: Conventional AI: Evans' Analogy Program
Previous: Preamble

Evans' program

Marr's model illustrates the advantages of having more than one representation, and how different representations make different kinds of information explicit. But his model is only a hypothesis about the brain's workings. The point can be made more forcefully by looking at a man-made construct (a computer program) whose author has chosen the representations and documented the reasons for each.

So to Evans' program. This was one of the first programs (though not the first) to simulate intelligent behaviour. Evans wrote it partly to investigate the use of ``high-level'' representations of simple pictures. This was an idea new to computing and AI at the time (1965-ish) --- now it's old hat. So the Analogy program is worth knowing because it demonstrates why different representations are useful in a particular task. By trying to simulate its behaviour, you can get a feel for this.

In its use of digital computers to solve ``intelligent'' problems (and not, e.g. motion control or implicit learning) it is typical of the symbolic approach to AI. Being fairly simple, it's a good point to introduce this topic. And you are quite likely to get a question on it, and/or on representations, in Finals.

So could you please write the following essay: ``Why might minds, or computers, need to represent information in different ways? Illustrate by describing the workings of Evans' Analogy program with reference to a worked example. How would you extend the program to solve 3-D problems?''

Jocelyn Paine
Tue Jun 3 11:33:37 BST 1997