201
Living Room / Re: Interesting "stuff"
« Last post by TaoPhoenix on September 18, 2015, 12:06 AM »Deep Learning Machine Teaches Itself Chess in 72 Hours, Plays at International Master Level MIT Technology Review-Arizona Hot (September 15, 2015, 02:11 PM)
Original paper (masters dissertation) here: http://arxiv.org/abs/1509.01549-mouser (September 15, 2015, 04:56 PM)
I remember when I was a kid I could beat the chess computer. At the time I never had any fear of AI. Both of those things have changed.-Renegade (September 17, 2015, 09:11 PM)
I downloaded the paper and might read it in more detail when I have time. For now, I'd be interested in how it does in the traditional "anti-computer but legit lines" like defending the White side of a King's Indian Defense because it has those long nebulous buildups before the attack looks like it makes sense.
And also I don't exactly know where it picks up its knowledge, but even "simple learning" can survey say the existing GM games and say "let's do that!" So using for example my crude tool here at DC "PGN Extractor", you go "Select games of Hikaru Nakamura+Black Plays King's Indian+White Wins" and it studies whatever White did there. Before modern web realtime-y game knowledge came along, that's how I was studying a decade ago.

Recent Posts







