An AI just crushed a team of scientists in a five-day poker competition, taking home a whopping $290,000 (£230,000). Named Libratus, the poker-playing machine had beaten four world-class poker pros in January. Now a team of scientists challenged the upgraded variant of it known as "Lengpudashi" using their vast knowledge about machine learning to their game but still went home defeated.

Lengpudashi means cold poker master, and it took on World Series veteran Alan Du who won in the $5,000 buy-in, no-limit, Texas Hold'em category last year, and a group of engineers, computer scientists, and investors. They prepared for the game by applying everything they know about game theory and machine learning instead of pure poker skills used by the first defeated group. Their strategy didn't work after playing 36,000 hands against the machine at a resort on China's Hainan island.

According to  Engadget, the AI Libratus co-developer Noam Brown said that it's clear that humans and computers have very different ways of interpreting a bluff. He said that people tend to think that bluffing is very human, which as it turns out is not true. On the other hand, a computer learns from experience that if it has a weak hand and it bluffs, it can increase the chances of winning the game.

Furthermore, per BBC, poker is not like chess and Go, in which all the playable pieces are visible on the board. The game is what computer scientists call an "imperfect information game." It requires a lot of relying on complicated betting strategies, the ability to bluff, and being able to spot when opponents are bluffing.

The AI systems were developed by Tuomas Sandholm from Carnegie Mellon University in the US, and Ph.D. student Noam Brown. The prize money won by their invention will go to Strategic Machine, which is a firm founded by the duo. It's not known yet if the machine will play against other human players.