Mind Matters Natural and Artificial Intelligence News and Analysis
Flush royal in poker player hand. Lucky winner.

AI Smash Hits #2 To Win AI-Run Poker Humans Must Find Blind Spots

Each pro separately played 5,000 hands of poker against five copies of Pluribus and Pluribus won. Now this, in itself, was an astonishing result.

Our Walter Bradley Center director Robert J. Marks is back with the second instalment of 2020 smash hits in AI. Readers may recall that we offered a fun series during the holidays about the oopses and ums and ers in the discipline (typically hyped by uncritical sources). This time, Dr. Marks talks with Eric Holloway about AI programs that can beat humans at poker.

Our story begins at 16:36. Here’s a partial transcript. (Show Notes and Additional Resources follow, along with a link to the complete transcript.)

Robert J. Marks: Carnegie Mellon and Facebook AI beats professionals in six player poker. This result astonished me. I have always heard poker players say that poker is not a game of rules. It is a game of bluffing, of psychology. But apparently not. The AI that was developed is called Pluribus. And in the game of Texas Hold’em, Pluribus was able to beat one-on-one professional Texas Hold’em players. And it is interesting that this world series of poker that the same people show up year after year and Pluribus bus beat Darren Elias. I don’t know these people. I don’t watch the world series of poker, but he holds the record for the most world poker tour titles. And Chris Ferguson. He was the winner of six world series of poker events.

Each pro separately played 5,000 hands of poker against five copies of Pluribus and Pluribus won. Now this, in itself, was an astonishing result. What was more astonishing is the Pluribus won in a game with five different pros at the same time. So there were six players. One of them was Pluribus and they played a total of 10,000 hands. And again, Pluribus emerged victorious. This to me was an astonishing result and lets me know that winning at poker is very highly algorithmic. There’s some randomness that goes in here for sure, but the fact that artificial intelligence can win at Texas Hold’em, to me is astonishing.

One of the things they used, which I hadn’t seen used before was something from game theory called the Nash equilibrium. Possibly you saw the movie A Beautiful Mind (2001) starring Russell Crowe. It is the story of John Nash who had some mental problems but was a mathematician of genius. He came up with the idea of a Nash equilibrium. And he won a Nobel prize for the Nash equilibrium as a game theory concept. It was applied to winning this Texas Hold’em and was a technique used by Pluribus.

Eric Holloway: I think I saw previous results where AI had beaten two player games, but I haven’t seen this one with five players before. It’s very interesting. And I’m also very curious about these, all of these results where AI beat humans. Because you you’re only ever told the end result that the AI beat the human and you’re given some kind of insight into how their play style differs from the human, but you don’t really know much about what exactly is going on under the hood.

Robert J. Marks: That’s very interesting because we saw nothing about Pluribus playing Pluribus, right? What would happen then?

Eric Holloway: Yeah. Well, I think that’s actually part of its training. It plays itself a number of times to… It’s like the AlphaGo Zero that plays itself a whole bunch in order to develop its strategies.

It’s just very interesting to think about, because like we were talking about with the AI hacking, when you have these really complex AI models, they have these blind spots where if you know where the blind spots are, you can poke those spots and make them do what you want. So I’d be curious if down the road with all these game-playing AIs, if people start finding out these blind spots in the AIs and figuring out how to control the game AIs.

Note: There are many limits to artificial intelligence. Here are six: #4 “Computers don’t actually experience things (qualia), which limits actual understanding.” So yes, there are certainly blind spots.

You may also enjoy: Does AI really “get” poker? Why that matters. Science journalist Maria Konnikova, also a professional poker player, explores the human side of poker and efforts to automate it

Here are the AI 2020 Smash Hits to date:

3 AI Smash Hits 2020: AI can help paralyzed people move again. The human brain can interface directly with electronics. An “exoskeleton walking device could get many paralyzed people out of their wheelchairs.

4 AI Smash Hits 2020 AI helps detect dreaded White Eye disease. The first step in treatment is correct diagnosis. Baylor University profs developed an app that enables eyeball disease in small children to be detected easily.

5 AI 2020 Smash hit: Deepfakes—What they can and can’t do. Deepfakes? Our minds often actually fill in a lot of our background for us when we are not even aware of it. One way of thinking about deepfakes is that they are liked instant mashed potatoes. The water is removed and re-added later.

6 AI Smash Hit: AI defeats fighter pilot hands down. The future of warfare may involve more machine waste but less human carnage. Eric Holloway: It’s going to be a more a hybrid approach where you have the fighter pilot and then a bunch of robot wingman that he can control.

7 AI Smash Hit: Why AI can’t do your thinking for you. Robert J. Marks: you change a pixel or two in an image and the deep convolutional neural network is totally wrong. Eric Holloway: The machine’s confidence in its result is complete certainty and it’s absolutely certain about the wrong result.

8 AI 2020 Smash Hit: Big gains in practical self-driving cars. The people who have been pursuing Level Five self-driving are nowhere but Level Four is working well. Jonathan Bartlett: You can think of Level Four self-driving as an engineering project and Level Five as a philosophy project

9 AI Success: Smarter cars for non-millionaires If your car is a recent model, an affordable aftermarket kit might transform it into a much smarter car. One possible risk is that a hacker could take over your car but, no matter what we do with AI, we must deal with security issues.

10 Smash Hit: #10 AI Success!: Translation gets faster and better. Machine translation, properly used, can help us communicate better. What’s made AI tech translation work so well is not that it’s perfect, but we’re going to have a second pass.

Show Notes

  • 00:31 | Introducing Jonathan Bartlett
  • 00:40 | Introducing Dr. Eric Holloway
  • 02:47 | #5: Deepfaking for Entertainment
  • 10:12 | #3: Paralyzed Man Moves in Mind-Reading Exoskeleton
  • 14:32 | #4: Deep Learning for leukocoria, or “white eye”
  • 16:36 | #2: AI Beats Professionals in Six Player Poker
  • 20:23 | #1: AI Cracks Protein Folding

Additional Resources

Podcast Transcript Download

Mind Matters News

Breaking and noteworthy news from the exciting world of natural and artificial intelligence at MindMatters.ai.

AI Smash Hits #2 To Win AI-Run Poker Humans Must Find Blind Spots