5 ms·

Wonder when we will see AI tutors appear? Seems like there is a massive potential to train an AI super player then have it teach you. Does anyone know of such t

Wonder when we will see AI tutors appear? Seems like there is a massive potential to train an AI super player then have it teach you. Does anyone know of such tools yet? Guess there are probably many AI augmented things we use all the time (Google Search) but just don't know it. Fascinating to think.

burkaman 10y ago
There was a recent post here about a new LiChess feature that identifies mistakes in your (or someone else's) games, prompts you to look for a better move, and tells you the best move if you can't figure it out. I think this is a form of AI tutoring. https://news.ycombinator.com/item?id=13241669 https://news.ycombinator.com/item?id=13241669
- WestCoastJustin 10y ago
  Very cool. Thanks!
- mormegil 10y ago
  The same feature has been in GNU Backgammon (and other programs, I guess) for a long time. Also, it allows you to request you or the AI get good or bad dice rolls, which is an interesting training option.
Deimorz 10y ago
Some go programs already have features going in that direction, where they can look at a particular board state and show you which moves they think are best from that position. People have been starting to use them to review their games to see how an AI stronger than them would have played differently. For example Leela (a free program) does this, and can display it with a "heatmap" style: https://sjeng.org/leela.html https://sjeng.org/leela.html Crazy Stone Deep Learning ($80 USD) has some similar analysis/hint features: http://www.unbalance.co.jp/igo/eng/ http://www.unbalance.co.jp/igo/eng/
Splines 10y ago
I'm not an expert, but I imagine these AIs attack the game using methods that just don't work with humans - predicting long deep branches of the game that are impossible to keep in your (well, at least my) head. The way I recall building simple game AIs is to create a board evaluator, then build a game tree, prune as necessary, and min-max as system constraints allow. I'm sure "real" AIs are much more sophisticated but I don't know how much of what it does can be taught to a human.

burkaman 10y ago

There was a recent post here about a new LiChess feature that identifies mistakes in your (or someone else's) games, prompts you to look for a better move, and tells you the best move if you can't figure it out. I think this is a form of AI tutoring. https://news.ycombinator.com/item?id=13241669 https://news.ycombinator.com/item?id=13241669

WestCoastJustin 10y ago
Very cool. Thanks!
mormegil 10y ago
The same feature has been in GNU Backgammon (and other programs, I guess) for a long time. Also, it allows you to request you or the AI get good or bad dice rolls, which is an interesting training option.

Deimorz 10y ago

Some go programs already have features going in that direction, where they can look at a particular board state and show you which moves they think are best from that position. People have been starting to use them to review their games to see how an AI stronger than them would have played differently. For example Leela (a free program) does this, and can display it with a "heatmap" style: https://sjeng.org/leela.html https://sjeng.org/leela.html Crazy Stone Deep Learning ($80 USD) has some similar analysis/hint features: http://www.unbalance.co.jp/igo/eng/ http://www.unbalance.co.jp/igo/eng/

Splines 10y ago

I'm not an expert, but I imagine these AIs attack the game using methods that just don't work with humans - predicting long deep branches of the game that are impossible to keep in your (well, at least my) head. The way I recall building simple game AIs is to create a board evaluator, then build a game tree, prune as necessary, and min-max as system constraints allow. I'm sure "real" AIs are much more sophisticated but I don't know how much of what it does can be taught to a human.