Jason Les tells CalvinAyre.com’s Tatjana Pasalic what’s it like to play poker against a supercomputer.
Want to know who will win in a poker war between humans and machines? Ask Jason Les.
Last May, the poker pro and three of his friends played 80,000 hands of Heads-up No-Limit Texas Hold’em against Claudico, a computer program created by Carnegie Mellon University, over the course of two weeks.
“At first, we were staying there super late because we were taking too many breaks, like we’re there from 11 a.m. to 3 a.m., and we have to be there the next day at 10 a.m. Well, we can’t do this, so we decided we’re gonna cut out all breaks besides one meal break and we would just do it all day long, every sing day,” Les told CalvinAyre.com. “It was quite a grind to do that. It was pretty brutal.”
Les trailed the supercomputer by $80,482 in the final chip tally, CMU posted on its website. Bjorn Li had an individual chip total of $529,033, while Doug Polk had $213,671 and Dong Kim had $70,491. Those were not their actual winnings, according to the university, which said the four players will be compensated from a prize pot of $100,000 donated by Microsoft Research and Pittsburgh’s Rivers Casino.
For Les, however, it was all about the experience.
“See when I started off, it was really hard for me to disconnect from treating it like a human player cause that’s all I’ve been conditioned to do: whenever I play Heads-up, I play humans. So it was hard for me to let go of that part of the game at first, but once I really understood what was going on, I made like a great turnaround and my results at the end were really good,” he said.
CMU professor Tuomas Sandholm, who led the creation of Claudico, previously said computers have already shown they can outplay humans at a simpler came of Heads-up, but not the far more complicated no-limit version.
Les said it was weird for the team at first, especially since they’re used to playing humans.
“The computer has a strategy it uses that it believes is the best strategy, to not adjust to the way we play. So you could kind of do some pretty ridiculous things and the computer is not gonna sit there and think, ‘Oh, this Jason’s just full of it here,’ in the future so you just kind of get away with stuff from here and there. It was really weird playing someone that is not adjusting to how you play, basically,” he said.
But Les said if given another chance and if the terms are right, then he’ll face off with the AI again.
“I think they do wanna do it again, but learns how to play poker by running a supercomputer. So the more time it has, the more situations it can iterate through and better develop its strategy so it really is a matter of time before it is either better than us or truly too close to call,” he said. “We did some preliminary discussions on the challenge in the future and I think it will happen, but I need to reach some agreements and some terms first before we commit to that grind again. If the terms are right, yeah, I’ll do it again.”
