next up previous
Next: The original source
Up: Main references
Previous: How Evans' program works

Some history

Artificial Intelligence by Minsky, in Scientific American September 1966. This is in the Psychology library and the RSL (Gen Per 36).

Read the whole article for historical background, to see why anyone should want to write such a program. Don't follow the account of how the program works (unless you're keen) because it's confusingly different from Winston's simplified account.

As well as Evans' program, the Scientific American article describes three other programs of the time: Samuel's checker (i.e. draughts) player; Bobrow's Student; and Roberts' scene analyser. The whole lot is well worth reading for a general view of AI at that time (Minsky is one of the greats of the subject). As you read it, note these points, all relevant to AI:

  1. The historical context.

  2. Brute-force exhaustive search is impossibly long for most problems (w.r.t. Samuel's program).

  3. So we need specific methods: heuristics.

  4. Breaking down a big problem into subproblems.

  5. ``Evans' program is certainly the most complex ever written'' - now, we have a version in Pop-11, running quite happily on the VAX. So does technology advance.

  6. General comments on analogy.

  7. The need for plans and strategies.

  8. Optimism in the conclusion! Reports from those early days often seem more exciting to read than what's written today.

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