5 ms·
I think poker would also be a great challenge. It is a special game because as well as the math, you also have to understand your opponent and adjust accordingl
by hmate9 11y ago
I think poker would also be a great challenge. It is a special game because as well as the math, you also have to understand your opponent and adjust accordingly. It'd be interesting to see weather DeepMind can beat professional poker players who have different playing styles.
- CyberDildonics 11y agoThere has been lots of research into many areas of AI poker.
- jhspaybar 11y agoMy understanding is that limit Texas Holdem poker is essentially solved. No limit is a different thing though, one mistake and you could be done.
- bradleyjg 11y agoI think that's heads up limit texas holdem, not limit holdem in general. e: ref--http://ai.cs.unibas.ch/_files/teaching/fs15/ki/material/ki02-poker.pdf http://ai.cs.unibas.ch/_files/teaching/fs15/ki/material/ki02...
- im3w1l 11y agoWouldn't that depend on what you want to achieve? To maximize $/h you need to understand your opponents weaknesses and exploit them. But if you just to beat the best humans, and don't care about the margin with which you beat them, then playing a mathematically perfect game should be enough. And before you ask, a mathematically perfect game can incorporate occasional bluffs, and occasional calling of bluffs, this would be known as a Mixed Strategy. Or am I missing something?
- eru 11y agoThat's about it for a two player game. Anything with more than two people has more interesting / complicated properties.
- im3w1l 11y agoOne very real issue I'd imagine is that two players could cooperate out of ignorance, i.e. either player could do better by deceiving the other, but they are not good enough players to realize that.
- gergoerdi 11y ago> you also have to understand your opponent This could be a dangerous idea. http://lesswrong.com/lw/x4/nonperson_predicates/ http://lesswrong.com/lw/x4/nonperson_predicates/