Monday, April 17, 2017

[jyketcyj] Humans improving a chess engine

Given a game in which a chess engine loses (probably against another engine), ask unassisted human players to identify the losing move (perhaps allowing probabilities to express degree of certainty).  (The answers can be verified with longer computation.)  Moves which human players can quickly and accurately identify as bad are candidates for areas of improvement for the losing engine, perhaps even automated improvement using machine learning.  (This assumes anything a human can compute quickly, a computer should be able to, too.)

Previously, on ways a human can help improve a chess program.

No comments :