OPA: An Othello Playing Algorithm

Abstract

Human intelligence seems to be closely related to flexibility and generalization, in agreement with current opinion characterizing human thought as heuristic and mechanical models as largely algorithmic. The Othello program developed in this SIP can and does win against human beings; but that same human being can then play hangman and recognize a phrase with only a very few letters revealed; to watch such a performance is to marvel at the unknown heuristics of the mind and, simultaneously, to despair of modeling it. To actually catch a glimpse down this philosophical crevasse, one needs only to do what I did: write a program that plays a game, understanding the algorithm fully, and yet be beaten regularly by it. The paradox is of course a false one, albeit at times very convincing. Had I written a program that calculated cubic splines or found the first 10,000 prime numbers, I would not expect my performance to be superior to it, any more than an airplane designer expects to be able to fly, but I must admit to feeling a bit of magic at such times.