Poker AI beats humans in NLHE

It looks like heads-up NLHE has been cracked. It's not much of a leap to n-player games, I'd think, once you have the basic bet evaluation code working. I wouldn't play online for significant money any longer, not without anti-bot safeguards...

https://www.engadget.com/2017/01/10/two-research-teams-taught-their-ais-to-beat-pros-at-poker/

I'm a little skeptical that a bot could beat "high level" players by that much. My understanding, is that the best HU NLHE players were already somewhat close to being perfect, and separated by a BB here and there.

"and AI researcher Noam Brown said last March there's "no chance" no-limit hold'em will be solved in our lifetime."

I also have heard several people say something to this effect. I know that the authors do not claim to have solved the game, but they do suggest that they are far ahead of any human.

42,000 hands is way too small of a sample. I do believe within a few years the bots will consistently beat the top pros.

Bots will never beat the top pro's... There is an article that we're going to discuss on GWAE. I've played with bots since 2006, and being a programmer I at least have an idea of how they're being programmed (that doesn't give me an edge alone it just enables me to discuss it much better).

Anyways, I'll get in to a lot more detail on the show, but humans will always find a way to beat the bots. 1 time and 42k hands is variance.

Quote: Romes

1 time and 42k hands is variance.

And the Bot knows that! Bots are programmed to sometimes bluff and sometimes not bluff... teh bot knows that an always bluff and always be 'tight' are losing strategies.
The bots algorithms are not always Greedy Algorithms. If you are a programmer you may soon have a profitable 'bot' playing while you sleep just as some programmers have a profitable 'bot' making BitCoins while they sleep.

Quote: Romes

Bots will never beat the top pro's... There is an article that we're going to discuss on GWAE. I've played with bots since 2006, and being a programmer I at least have an idea of how they're being programmed (that doesn't give me an edge alone it just enables me to discuss it much better).

Anyways, I'll get in to a lot more detail on the show, but humans will always find a way to beat the bots.

I disagree. I think the distinction between limit HE and NLHE is just one of scale, not of quality, and limit HE is a solved problem. In other words, there is already software out there that, for heads-up play in limit hold'em, the best human in the world cannot beat.
http://science.sciencemag.org/content/347/6218/145.full
http://poker.srv.ualberta.ca/about

Quote: MathExtremist

It looks like heads-up NLHE has been cracked. It's not much of a leap to n-player games, I'd think, once you have the basic bet evaluation code working. I wouldn't play online for significant money any longer, not without anti-bot safeguards...

https://www.engadget.com/2017/01/10/two-research-teams-taught-their-ais-to-beat-pros-at-poker/

It is scary to see how technology and Robots / AI have developed so rapidly in such a short period of time over the past decade. The human race is in serious jeopardy in the future. Future generations will have limited work, limited privacy, and ultimately limited control over their lives. It is happening. If you have children, make sure they study computer science!

Regarding the AI poker... Im assuming it was played online? I dont trust any games online to start so I it is hard to ration what a bot would do in a live game with humans.... but I dont think we have to worry about them playing in casinos...just yet!

Just adding, Chess AI didn't need to be "solved" to beat humans. Just needs to play better than the best.

Quote: rxwine

Just adding, Chess AI didn't need to be "solved" to beat humans. Just needs to play better than the best.

Exactly!! And we know that chess can eventually be solved but I don't know that NLH can be solved.

Quote: MathExtremist

I disagree. I think the distinction between limit HE and NLHE is just one of scale, not of quality, and limit HE is a solved problem. In other words, there is already software out there that, for heads-up play in limit hold'em, the best human in the world cannot beat.
http://science.sciencemag.org/content/347/6218/145.full
http://poker.srv.ualberta.ca/about

I do actually agree about fixed games... As most bot computations become much more exploitable in NL =).

Quote: rxwine

Just adding, Chess AI didn't need to be "solved" to beat humans. Just needs to play better than the best.

There was a time when the consensus was the a computer could never beat a human at chess. It couldn't have been possible because it would always be humans programming the computer, they thought. Biggest difference I see is that a better chess player almost always wins, what's the winning percentage for a good poker player against a fair one?