theweaselking: (Default)
[personal profile] theweaselking
Hey, remember when people were all "sure, CHESS is solved, but Go is a game where a human will always be better than a computer"?

Google Deepmind is playing Lee Sedol, a 9th dan professional Go player widely considered one of the two best players in the world. And it won the first game.

Worth noting, remember that program that finds fucked up faces of dogs in everything? Also Deepmind.

(no subject)

Date: 2016-03-09 08:06 pm (UTC)
From: [identity profile] ironphoenix.livejournal.com
Go strategy is largely based on dog faces. The Chinese word for dog is even pronounced "go".

(no subject)

Date: 2016-03-10 01:19 am (UTC)
From: [identity profile] franklanguage.livejournal.com
"God…dog…god…dog…"

(no subject)

Date: 2016-03-15 07:22 pm (UTC)
From: [identity profile] falconwarrior.livejournal.com
what if someone uploads stuff to your brain that you'd rather not know?

(no subject)

Date: 2016-03-09 09:56 pm (UTC)
From: [identity profile] nojay.livejournal.com
Analysis of the first game suggests Sedol blew it by overthinking the opening, pulling a trick move out of his hat that gave the Google program an early advantage. After that it see-sawed a bit with both "players" making mistakes and recovering from them but Sedol slipped up at the end of the midgame and as a result he was toast as the computer wasn't going to lose the endgame where there are many fewer possible moves to evaluate in look-forward.

AlphaGo will only get better even if it loses against Sedol and other top players. Anyone who claimed master-level Go was always going to be the preserve of humans was talking out their ass. The really neat thing is the approach the AlphaGo people have taken in developing their engine which has few similarities with previous computer Go programs, some of which are based on algorithms developed from successful chess programs such as Monte Carlo and alpha-beta tree pruning.

(no subject)

Date: 2016-03-10 12:33 am (UTC)
From: [identity profile] skington.livejournal.com
"The really neat thing": do you have any handy links? That sounds far more interesting than Moore's Law + Google > J Random Go Player.

(no subject)

Date: 2016-03-10 11:30 am (UTC)
From: [identity profile] theweaselking.livejournal.com
This involves a lot of Go jargon but explains it pretty well regardless. (https://gogameguru.com/alphago-defeats-lee-sedol-game-1/)

(no subject)

Date: 2016-03-10 01:45 pm (UTC)
jerril: A cartoon head with caucasian skin, brown hair, and glasses. (Default)
From: [personal profile] jerril
Something I'd heard on the brief CBC radio coverage was that Sedol was showing signs of being nervous, and flustered/intimidated by his opponent, right from the outset. That's not a game-winning mindset, for sure. I've heard of this being actually one of the advantages the chess computers had too - the computer always presents a blank slate, and a player who's nervous and not used to machine players can anthropomorphize that as "calmly self-confident". Pretty intimidating.

Me, I've played videogames for so long I'm willing to ascribe both brainless emotionless mechanical action and the complete gamut of emotions and motivations to a machine player. Probably screws me up regularly, too :D

(no subject)

Date: 2016-03-10 02:50 pm (UTC)
From: [identity profile] theweaselking.livejournal.com
AlphaGo won game 2, too.

(no subject)

Date: 2016-03-11 11:15 pm (UTC)
From: [identity profile] chrisrw109.livejournal.com
And it played the second game extremely well.

That is some very, very, scary stuff... it gives me the goosies (in a good way)

Profile

theweaselking: (Default)theweaselking
Page generated Aug. 8th, 2025 10:52 am