Re: [neat] Re: Othello Competition
- Kenneth Stanley wrote:
>--- In email@example.com, Ian Badcoe <ian_badcoe@y...> wrote:Good point. The competition is against other machine discovered
>> I saw that. Given the difficulty we are having with TTT, I
>>thought Othello was a little out of our league for the moment.
>I actually don't think this is true. The difficulty we are having
>with TTT is getting it perfect, and the same problem would likely pop
>up in Othello. However, not being perfect at Othello leaves a lot of
>room for notable achievement. The problem with TTT is that it's so
>simple that we know everythign about what it should be doing. Othello
>is far more complex.
s-expressions so it's likely they'll encounter the same limitations that
we have. On the other hand if much better players (than our typical TTT
player) are found then there will be some lessons to be learned.
>In fact, have you guys seen this paper on discovering the mobilityThanks for the link.
>strategy in Othello using neuroevolution from 1995? :
>It's pre-NEAT, but it's quite interesting that the NN learned an
>advanced strategy that had only recently been discovered by humans.
>However, even with these results, I'm still skeptical about our
>ability to make close to human-level progress in board games with the
>current NE technology. So in a sense I agree with Ian at the deepest
>level. But we can still get fairly far and perhaps discover something