AI research throws cold water on AlphaZero chess news

[ClaytonB]

I read the Alpha Zero paper and I've watched some of these AlphaZero chess videos:



An AI researcher throws some cold water on the Alpha Zero news here (Medium). The author's points are methodological and I shared many of the same concerns after reading Alpha Zero's paper. The 1-minute-per-move rule was especially concerning because time-management is one of the key skills of advanced chess play which chess engines have mastered with difficulty. Stockfish's creator Tord Romstad is quoted in the article, "a lot of effort has been put into making Stockfish identify critical points in the game and decide when to spend some extra time on a move."

Anyway, Alpha Zero is definitely a novel development in AI and Deep Mind deserves props for their work. But this is definitely not the last chapter of the story.

Erratum: Title should read "AI researcher throws cold water on AlphaZero chess news"