Top

Facebook AI Beats Professional Poker Players In Major Artificial Intelligence Breakthrough

What Is Pluribus?

A team of researchers from Facebook have recently created a revolutionary poker bot, which has successfully beaten some of the world’s best poker players.

The bot was created by the AI team at Facebook, along with researchers from Carnegie Mellon University, and has been named Pluribus.

poker.png


Pluribus was the first AI bot which is capable of beating as many as six players in a game of no-limit Texas Hold’em poker.

 Over 12 days the AI system played 12 world-renowned poker players in over 10,000 hands of Texas Hold’em. Pluribus won an average of $5 per hand, with an impressive hourly winning of around $1,000.

 

A Huge Milestone In AI Research

The creators of the bot have recently published an article in the journal Science, in which they state that this is a huge milestone in AI research.

 

AI has already reached superhuman levels in board games like chess and Go, and computer games like Dota, six person no-limit Texas Hold’em represents a significantly higher benchmark of difficulty.

 

This is in part due to the fact that the game involves ‘hidden’ information- meaning that the bot has no way of analysing the other players’ cards.

 

“It Can Bluff Better Than Any Human”

Perhaps even more impressive still, the bot is capable of bluffing and detect when other players might be doing the same.

“We think of bluffing as this very human trait” explained Noam Brown, the lead researcher from Facebook’s AI team, in an interview with BBC News.

“But what we see is that bluffing is actually mathematical behaviour.”

“The AI doesn’t see bluffing as deceptive. It just sees the decision that will make the most money in that particular situation.”

“We can show that AI can bluff, and it can bluff better than any human.”

 

Practice Makes Perfect

The researchers at Facebook developed the bot by programming the rules of no-limit Texas Hold’em  and then having the bot play hands against itself trillions of times, a technique known as reinforcement learning. In other words, practice makes perfect.

 

Pluribus Became A Poker Pro In Just 21 Hours

Another notable development with this piece of intelligence is the incredibly low amount of computing power needed to train Pluribus.

 

It is reported  that the programme was able to run from just $150-worth of cloud computing resources. To put this achievement into context, similar bots created by Deepmind, Google’s AI research shop, have relied on supercomputers with over 5,000 specialist processors, costing millions of dollars.

Furthermore, just 20 hours of learning was needed for Pluribus to train itself up to the ability of a world-beating poker professional.

Top Poker Players Gave Pluribus Their Seal Of Approval...

Pluribus has won the seal of approval from several poker champions who had been invited to play against the AI.

“Pluribus is a very hard opponent to play against,” said Chris Ferguson, a World Series of Poker champion. “It’s really hard to pin him down on any kind of hand”.

“Whenever playing the bot, I feel like I pick up something new to incorporate into my game,” says poker pro Jimmy Chou.

...But Pluribus Won’t Be Entering Casinos Any Time Soon

However, Noam Brown has made it clear that neither he nor Facebook had any plans to use the AI in real poker games.

The firm has said that it is not publicly disclosing the code used to programme Pluribus, for fear of having a negative impact on the poker community.

Pluribus Beyond Poker

However, the researchers still hope that the AI technology will have practical uses beyond Texas Hold’em. It is thought that the bot could prove useful in everything from cyber-security, to fraud-detection, to navigate traffic with self-driving cars.

For more Facebook news check out our blog