Machine learning is when computers use experience and the data they gather to learn by themselves, similar to humans. In the world of artificial intelligence (AI), poker bots are now able to use this and their algorithms to defeat professional players.

That’s right – computers are out-bluffing humans to the tune of millions of dollars!

The complexity involved with beating such a human game with imperfect data is pushing AI to a new level, and investment is ramping up in industries that can now see the application of similar learning strategies on the horizon.

This is particularly true of cyber-security, government defense, banking, healthcare and medicine, and certain aspects of business.

Libratus from the Pittsburgh Supercomputing Center utilizes 274 terabytes of RAM to apply a learning algorithm rather than a fixed built-in strategy. It learned the rules of poker itself via trial and error and every night it analyzes its success and recalibrates for its next game.

The more it plays the more hands and situations it encounters and the more it learns. Its strategy is currently based on 15M core hours of computation.

In January 2017, Libratus dealt the biggest ‘AI vs. Human Brains’ blow so far, defeating four professional poker players in a 20-day tournament.

Dong Kim was down by $85,649, Daniel McAuley $277,657, Jimmy Chou $522,857, and poor Jason Les lost $880,087!

On average the bot won $14.72 per hand for 120,000 hands – Not bad considering it had been playing for less than two years.

The exciting thing about Libratus is that its algorithm is not specific to poker and can be applied to any situation that has ‘imperfect information’ like hidden hands and bluffing. Machine learning in cyber-security is expected to boost big data intelligence and analytics spending to $96 billion by 2021!

Facts and data about poker vs AI and future AI trends can be found in the new infographic by