AI Bots Could Beat 6max Within 2 Years

The compter scientist who designed Libratus, the AI bot which comfortably beat four of the best heads-up NLHE players in the world last month, has claimed that with some minor improvements he expects that Libratus should be able to beat humans at six-max NLHE within two years.

Carnegie Mellon University PhD Student Noam Brown, who also co-created Claudico the AI poker bot which the humans defeated in 2015, made the claim in a recent interview with Brian Pempus at CardPlayer.

In the interview, Brown reveals that he was surprised by just how big a margin (14bb/100) Libratus was able to beat this year's human team of Jason "PremiumWhey" Les, Dong "Donger Kim" Kim, Jimmy "ForTheSwaRMm" Chou, and Daniel "dougiedan678" Macaulay. Brown said that he had expected Libratus to have an edge over the humans as it had beaten Claudico by a bigger margin than the humas had - but that to beat them quite so handsomely was a little surprising.

In the end Brown put the victory down to the inability of the humans to exploit Libratus in the way they were able to exploit Claudico. He did admit that the humans did find some weaknesses (specifically how it wasn't adjusting well to certain opening sizes), but that Libratus was able to quickly plug leaks whilst constantly training during the humans downtime.

Brown also said that he saw no reason why a machine like Libratus couldn't win playing more than one player at a table saying, "Generally speaking, the techniques that went into Libratus work really well even if you have more than two players", although he did say that at this time six-max is a little bit beyond the abilities of Libratus and similar AIs. However, he noted that the annual computer poker competition is adding a six-player league and that as such you could expect massive improvements in that area and that, "I think that with some minor improvements to Libratus, you’d be able to see it beating humans at six-max within two years"

The big question with six-max is whether GTO is the best way to beat the game, or whether exploiting weak players is always going to show better results in the longs run - and that at present, humans are still better at exploitative play than AI computers.

To read more of Noam Brown's thoughts, check out the interview in its entirety here


Other Poker news

Leave a Reply

To post comments you need to Login or register your free HighstakesDB account.


No comments have been posted yet.