Sign up for our newsletter and get the latest big data news and analysis.

Poker and AI: The Rise of Machines Against Humans

We humans like to think our emotional intelligence separates us from machines, but even in the highly skilled human game of poker we could be about to get out-bluffed by artificial intelligence (AI), according to our friends over at PokerSites.me.uk. who put together the detailed infographic below.

For many years AI has been able to defeat human world champions at games like chess because all moves and odds can be calculated – there is ‘perfect’ information. However Poker hands are hidden and bluffing is a complex human process, making the game a great base for developing more powerful and intelligent AI.

In 2015 Limit Hold-em was solved by the University of Alberta’s Cepheus program, meaning it can play the game without losing money in the long-term. Using powerful CPUs and considering over six billion hands, the computer earned its edge over humans.

Intrinsic advantages of AI over humans are numerous. AI doesn’t get tired or complacent, it isn’t influenced by the value of money, and it can map weakness patterns in humans better than humans themselves.

In 2017 DeepStack, another UA AI bot managed to defeat pros at Texas Hold’em via its artificial neural networks that had been trained in poker intuition.

That all sounds so simple, but there are 316,000,000,000,000,000 different game situations in Limit-Hold’em, which could take 10 billion years to play through. In No-limit Hold’em there are more unique situations than there are atoms in the universe!

This is where supercomputing comes in. The Pittsburgh Supercomputing Center’s Libratus has formulated a poker strategy based on 15M core hours of computation and requires 274 Terabytes of RAM. It has the ability to analyze its own play every night and tweak accordingly for the next day’s play. Against 4 top players Libratus won at a rate of $14.72 per hand. Jason Les lost the most – $880,097.

It’s not all just fun and games however; the machine learning being driven by poker bots is expected to be adopted by the cyber security industry to the tune of $96 billion by 2021!

 

 

 

Sign up for the free insideBIGDATA newsletter.

Leave a Comment

*

Resource Links: