(from https://acoarecovery.wordpress.com/tag/anger/)
Coping w/ AlphaGo Development
-
Bonobo
- Oza
- Posts: 2223
- Joined: Fri Dec 23, 2011 6:39 pm
- Rank: OGS 9k
- GD Posts: 0
- OGS: trohde
- Universal go server handle: trohde
- Location: Germany
- Has thanked: 8262 times
- Been thanked: 924 times
- Contact:
Coping w/ AlphaGo Development
When, here and there, I watch some Go players’ reactions to the latest development re: AlphaGo this chart comes to mind …
(from https://acoarecovery.wordpress.com/tag/anger/)
(from https://acoarecovery.wordpress.com/tag/anger/)
“The only difference between me and a madman is that I’m not mad.” — Salvador Dali ★ Play a slooooow correspondence game with me on OGS? 
-
HKA
- Lives with ko
- Posts: 180
- Joined: Fri Apr 23, 2010 6:02 am
- Rank: Declining
- GD Posts: 2428
- Location: Usually the third line
- Has thanked: 66 times
- Been thanked: 341 times
Re: Coping w/ AlphaGo Development
the glittering sword
dances on the trembling peak
as we wash our necks
dances on the trembling peak
as we wash our necks
My days of not taking you seriously are certainly coming to a middle
-
Bill Spight
- Honinbo
- Posts: 10905
- Joined: Wed Apr 21, 2010 1:24 pm
- Has thanked: 3651 times
- Been thanked: 3373 times
Re: Coping w/ AlphaGo Development
Glist'ning neck,
Just cut it off!
Glist'ning neck.
Just cut it off!
Glist'ning neck.
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
-
belikewater
- Dies in gote
- Posts: 29
- Joined: Fri Nov 27, 2015 9:05 pm
- GD Posts: 0
- Has thanked: 7 times
- Been thanked: 14 times
Re: Coping w/ AlphaGo Development
Do not go gentle
into that third game. Rage, rage
against the machine
into that third game. Rage, rage
against the machine
-
belikewater
- Dies in gote
- Posts: 29
- Joined: Fri Nov 27, 2015 9:05 pm
- GD Posts: 0
- Has thanked: 7 times
- Been thanked: 14 times
-
Babelardus
- Lives with ko
- Posts: 161
- Joined: Thu May 26, 2016 4:09 pm
- GD Posts: 0
- Has thanked: 2 times
- Been thanked: 25 times
Re: Coping w/ AlphaGo Development
Well, it was to be expected.
Since the 90's, when people started saying that "Go masters would never be beaten by computers", I've always said: "They will. Computers originated mostly in the West, and Chess is Western. It's logical that development of thinking software started there. As soon as Chess is definitively mastered, research will fully turn to Go, and 'the' way to program a computer for it will be developed."
Go is not impossible to play well for a computer, as we have seen now. The only thing is that Chess programming doesn't work for Go. The other way around also doesn't work; there have been Monte Carlo experiments in Chess engines, and it doesn't work well.
The only thing I'm afraid of is that we will hit the same snag as with chess engines. They are so strong, that it's almost impossible to defeat them without giving them a thinking time of a second or less. Even then it's very hard: on my computer, Stockfish calculates at least 10-12 ply in the first second. You basically get a chess program that plays instantly all the time. That's no fun.
Therefore, people have been looking at making an engine play naturally when not set at full strength. When set at ELO 1800, it will use all of its time, but it will make mistakes an 1800 player would make. That sort of stuff. Engines getting better and better at that, so you feel as if you are playing against an 1800 player, instead of a severely limited 3200+ ELO player.
I hope something similar happens with Go as well; train a neural network to play at different levels, and play as a player of that level, from 9d down to 20k. It's no fun having an instantly playing program, or a program you have to battle with 9 handicap stones all the time to stand a chance.
Since the 90's, when people started saying that "Go masters would never be beaten by computers", I've always said: "They will. Computers originated mostly in the West, and Chess is Western. It's logical that development of thinking software started there. As soon as Chess is definitively mastered, research will fully turn to Go, and 'the' way to program a computer for it will be developed."
Go is not impossible to play well for a computer, as we have seen now. The only thing is that Chess programming doesn't work for Go. The other way around also doesn't work; there have been Monte Carlo experiments in Chess engines, and it doesn't work well.
The only thing I'm afraid of is that we will hit the same snag as with chess engines. They are so strong, that it's almost impossible to defeat them without giving them a thinking time of a second or less. Even then it's very hard: on my computer, Stockfish calculates at least 10-12 ply in the first second. You basically get a chess program that plays instantly all the time. That's no fun.
Therefore, people have been looking at making an engine play naturally when not set at full strength. When set at ELO 1800, it will use all of its time, but it will make mistakes an 1800 player would make. That sort of stuff. Engines getting better and better at that, so you feel as if you are playing against an 1800 player, instead of a severely limited 3200+ ELO player.
I hope something similar happens with Go as well; train a neural network to play at different levels, and play as a player of that level, from 9d down to 20k. It's no fun having an instantly playing program, or a program you have to battle with 9 handicap stones all the time to stand a chance.