By January 13, 2015

Scientists Use Poker to Improve Robot’s Decision Making

A couple of scientists working in their laboratory.Whether they’re serving us drinks on a cruise ship or parking our cars for us, robots are learning new tasks at what feels like an extraordinarily rapid pace. Driving that point home is a new study out of the University of Alberta in Canada that demonstrates the incredible decision-making capabilities of one particular robot.

According to NPR, researchers at the university developed a program that can beat any human opponent at a variant of Texas Hold’em poker. It’s known as Heads-up Limit Texas Hold’em, a version of the game that only features two players and capped betting. Despite being more limited than other variants, the research team still found there are “10 trillion different decision points in the game.” And one of those can apparently be used to win no matter what the other players is doing, which is interesting considering the act of bluffing is crucial to any version of poker.

But lead researcher Michael Bowling said that no matter how good the human player is, the computer will win. It took him and his team many years and the power of some 5,000 computers to figure it all out, with one researcher claiming it took nearly “a thousand years of computation” to reach these results.

Despite this program’s obviously strong skill set when it comes to this poker variant, Bowling said that online players won’t have to worry about it dominating virtual poker rooms. For one, it’s playing a variant that isn’t terribly popular among poker players at large. That being said, Bowling noted that there are indeed bots out there that players should look out for—just not his. For Bowling, it all comes down to improving the robot’s ability to make decisions when facing uncertainty, not clearing out the bank accounts of rival players.

Of course, if they did want to make the robot’s decision-making processing even that much more incredible, the scientists could look to a slew of other poker games to test it out on. According to online platform betfair poker, there are a number of variants out there including one close to what the robot was taught to play. That’s the multi-player Texas Hold’em, perhaps the most popular version of the card game thanks to its boom in the late 1990s and early 2000s. That all being written, these more complicated games could prove exceedingly difficult and likely impossible in terms of programming. Can you imagine the number of variables at play here? Again, this is especially true when noting the very “human” factor of bluffing, a major component to play any form of poker.

For those interested in reading more about this poker-playing robot study, head to the Cepheus Poker Project. With this type of decision-making programming in place, it’ll certainly be interesting to see where other, similar studies can make advancements—and how they can be applied to real-life situations.

About the Author:

More than 20 years in the IT industry. Blogging with a passion and thirst for new technology since 2005.
Loading Facebook Comments ...

Post a Comment

No Trackbacks.

https://www.ukmeds.co.uk/surgical-face-masks