next up previous
Next: Planning and story-understanding
Up: AI and Problem Solving: Planning
Previous: What else is needed?

Unconventional approaches

Some researchers say that conventional AI has had such little success that we need a radically new approach. Their attitude, if one can generalise across the whole work, is that most of what's needed in a natural intelligence was placed there by evolution long before the human brain evolved (consider the two billion years life has existed with the history of the human race). Conventional AI has concentrated on rational intelligence, printing the postage stamp on top of the Empire State Building without worrying about the building itself. What we should be doing is to work our way up the evolutionary tree, making a complete creature at each step and then building the next on top of the last.

This work is new, and we can't yet say where it will lead. But it seems to have some tempting successes, and anyone who's going to take AI seriously should know of it. Start by finding the first issue of Artificial Intelligence for 1991 (numbers 1--3 in one volume), and read the article by Brooks and its criticism by Kirsh (incidentally, president of the AI Society here, 1982--83).

For more details, look at AI at MIT, volume 2 edited by Winston and Shellard, pages 3--26; Minimalist Mobile Robots by Connell, chapter 1; and Intelligence as Adaptive Behaviour chapter 1. All these are now in the Psychology library.

Then note that not everybody agrees: read the contrary view in the Forward to Readings in Planning.

Jocelyn Paine
Tue Jun 3 10:55:12 BST 1997