AI Might be closer than we think [also poker related]

[noxiousdog]

$150 of Cloud Computing beat Chris Ferguson, Greg Merson, Darren Elias, Linus Loeliger and others in 6-player no-limit hold'em. Convincingly.

Pluribus also uses new, faster self-play algorithms for games with hidden information. Combined, these advances made it possible to train Pluribus using very little processing power and memory — the equivalent of less than $150 worth of cloud computing resources. This efficiency stands in stark contrast to other recent AI milestone projects, which required the equivalent of millions of dollars’ worth of computing resources to train.


...

“Pluribus is a very hard opponent to play against,” Ferguson said after the experiment. “It’s really hard to pin him down on any kind of hand. He’s also very good at making thin value bets on the river. He’s very good at extracting value out of his good hands.”

...

Pluribus would have won an average of about $5 per hand and would have made about $1,000/hour. This result exceeds the rate at which professional players typically expect to win when playing against a mix of both professional and amateur players.

I think the scary part is that Facebook is involved.

[The Meal]

The results of the actual hands played show that Pluribus was actually a small loser against the humans, and it's hypothesized that it's how the researchers performed their variance reduction (the AIVAT) that let them declare victory. One well known Poker researcher has qualms about the CMU team (coincidentally, the same Darse Billings was the winner of my very first NNPT hockey tournament).

Poker AI, especially for non-traditional multi-way Nash Equilibria, has come a long way. But there's significant room before one could declare multi-way no-limit poker as solved.

[noxiousdog]

That makes me feel much better.

Facebook with leading AI would be terrifying.

[noxiousdog]

It seems published in Science so we'll see.