On: Chess@home: Building the largest Chess AI ever (in...

[judofyr]

> When the AI is strong enough, we’ll invite an actual french Chess Grand Master along with a Guinness official for the most epic Chess game, ever!

Ehm, what makes this so epic? Computers have been known to beat humans at chess for quite a while now.

[oldstrangers → judofyr]

Its gotten so bad (good) that cell phones are now playing at GM levels. Kind of depressing, but inevitable.

http://en.wikipedia.org/wiki/Human-computer_chess_matches#Po...

[sylvinus → oldstrangers]

Yes, it's astounding (if true) that Pocket Fritz only searches 20k nodes per second, compared to Deep Blue's 200M. There must be some serious positional intelligence involved there!

[gcp → sylvinus]

Simple improvements in the tree search algorithms to reduce the effective branching factor are enough. Those have exponential effects.

Improving the evaluation function is "merely" a constant improvement.

[TylerE → sylvinus]

Your average GM is also not Gary Kasparov. The difference between a typical GM and one of the handful of elites is pretty astounding.

[starwed → TylerE]

For a graphical grasp of this concept, take a bell curve, and then plot only the top 0.1% of the bell curve. That's the distribution of abilities among people who are any good at all at chess, and it is no longer anything close to a bell curve.

[sylvinus → judofyr]

Well, enthusiasm was clearly speaking here but you can consider it epic in that one player will be thrown against potentially "all" the computers of the world. Kind of a "Man vs. Grid" thing.

But yes, in terms of pure strength, nothing epic for now :)