AI Can Now Beat Pro Poker Players: But Why Does It Matter?

To some it’s fascinating, to others it’s unnerving, but in a reality where artificial intelligence can defeat professional poker players at their own game – why does it actually matter? It’s not like robots can cash chips.

Wait, bots are out-bluffing humans? Yes, and they’re only going to get better.

In 2015 Limit Hold’em was solved by Cepheus from the University of Alberta, Canada. That same year DeepStack bested 33 pros from the International Federation of Poker, with a 492 mbb/g margin over 44,852 hands!

Libratus from the Pittsburgh Supercomputing Center took $1.76 million in chips from poker stars Jason Les, Jimmy Chou, Daniel McAuley, and Dong Kim, earlier this year.

All of these impressive wins are broken down in a new facts sheet from PokerSites:

But what’s the purpose of AI beating humans? Why spend billions on developing such bots? Believe it or not, it’s not just fun and games.

Games (especially poker) involve a lot of complexity and once they are mastered, that level of skill can have important real-world applications. The cybersecurity industry alone is expected to spend upwards of $96 billion on machine learning by 2021.

Machine learning is the process by which AI learns and improves on mistakes without extra programming. Once it’s started, the bots begin self-learning and advancing. You might then see why a complex game like poker, full of imperfect information and human emotion is a valuable tool in developing this process.

Libratus analyzes its own results every play and every night and computes its next strategy without human interference. It makes use of 274 terabytes of RAM and is 30,000 times faster than the average PC!

This process can be applied to other high stakes situations, where decision making and continual learning is necessary (which if you think about it is a lot of situations). For example, coming out ahead in a business negotiation, any type of planning or strategizing (business, military, and government), the financial markets, and medical treatments – the list is endless.

Bluffing is also a key ability in negotiating. Imagine for example an AI bot that could negotiate the best prices for real estate or your cell phone provider? It will probably just be an app on your phone.

But don’t fret if you’re one of those that fear the arrival of ‘the machines.’ The likes of Libratus are still extremely expensive to run and despite being theoretically applicable to tasks outside of poker, we’re a long way off bot deciding how to run the United States government.


charts the history and current landscape of AI, poker and machine learning.

created by: PokerSites-Poker AI

Image URL:

Shopping Cart
Scroll to Top