How AI Conquered Poker (nytimes.com) 40
Good poker players have always known that they need to maintain a balance between bluffing and playing it straight. Now they can do so perfectly. From a report: One of the earliest and most devoted adopters of what has come to be known as "game theory optimal" poker is Seth Davies's friend and poker mentor, Jason Koon. On the second day of the three-day Super High Roller tournament, I visited Koon at his multimillion-dollar house, located in a gated community inside a larger gated community next to a Jack Nicklaus-designed golf course. On Day 1, Koon paid $250,000 to play the Super High Roller, then a second $250,000 after he was knocked out four hours in, but again he lost all his chips. "Welcome to the world of nosebleed tourneys," he texted me afterward. "Just have to play your best -- it evens out." For Koon, evening out has taken the form of more than $30 million in in-person tournament winnings (and, he says, at least as much from high-stakes cash games in Las Vegas and Macau, the Asian gambling mecca). Koon began playing poker seriously in 2006 while rehabbing an injury at West Virginia Wesleyan College, where he was a sprinter on the track team.
He made a good living from cards, but he struggled to win consistently in the highest-stakes games. "I was a pretty mediocre player pre-solver," he says, "but the second solvers came out, I just buried myself in this thing, and I started to improve like rapidly, rapidly, rapidly, rapidly." In a home office decorated mostly with trophies from poker tournaments he has won, Koon turned to his computer and pulled up a hand on PioSOLVER. After specifying the size of the players' chip stacks and the range of hands they would play from their particular seats at the table, he entered a random three-card flop that both players would see. A 13-by-13 grid illustrated all the possible hands one of the players could hold. Koon hovered his mouse over the square for an ace and queen of different suits.
The solver indicated that Koon should check 39 percent of the time; make a bet equivalent to 30 percent the size of the pot 51 percent of the time; and bet 70 percent of the pot the rest of the time. This von Neumann-esque mixed strategy would simultaneously maximize his profit and disguise the strength of his hand. Thanks to tools like PioSOLVER, Koon has remade his approach to the game, learning what size bets work best in different situations. Sometimes tiny ones, one-fifth or even one-tenth the size of the pot, are ideal; other times, giant bets two or three times the size of the pot are correct. And, while good poker players have always known that they need to maintain a balance between bluffing and playing it straight, solvers define the precise frequency with which Koon should employ one tactic or the other and identify the (sometimes surprising) best and worst hands to bluff with, depending on the cards in play.
He made a good living from cards, but he struggled to win consistently in the highest-stakes games. "I was a pretty mediocre player pre-solver," he says, "but the second solvers came out, I just buried myself in this thing, and I started to improve like rapidly, rapidly, rapidly, rapidly." In a home office decorated mostly with trophies from poker tournaments he has won, Koon turned to his computer and pulled up a hand on PioSOLVER. After specifying the size of the players' chip stacks and the range of hands they would play from their particular seats at the table, he entered a random three-card flop that both players would see. A 13-by-13 grid illustrated all the possible hands one of the players could hold. Koon hovered his mouse over the square for an ace and queen of different suits.
The solver indicated that Koon should check 39 percent of the time; make a bet equivalent to 30 percent the size of the pot 51 percent of the time; and bet 70 percent of the pot the rest of the time. This von Neumann-esque mixed strategy would simultaneously maximize his profit and disguise the strength of his hand. Thanks to tools like PioSOLVER, Koon has remade his approach to the game, learning what size bets work best in different situations. Sometimes tiny ones, one-fifth or even one-tenth the size of the pot, are ideal; other times, giant bets two or three times the size of the pot are correct. And, while good poker players have always known that they need to maintain a balance between bluffing and playing it straight, solvers define the precise frequency with which Koon should employ one tactic or the other and identify the (sometimes surprising) best and worst hands to bluff with, depending on the cards in play.
Eww (Score:4, Funny)
I visited Koon at his multimillion-dollar house, located in a gated community inside a larger gated community next to a Jack Nicklaus-designed golf course.
He lives in hell? How awful for him.
Re: Eww (Score:2)
Koon sounds like a real Koont.
Alt. Title (Score:4, Insightful)
How algorithms help players play better.
But I guess that isn't as sexy as the overused "AI" which is nothing more then a glorified table lookup. /s
Can we please now start worrying about AGI? (Score:5, Funny)
Re:Can we please now start worrying about AGI? (Score:4, Interesting)
Re: (Score:2)
Re: (Score:3)
Not really. In poker or go, the goal is to win at the expense of all other players. In government, the naïve view would be that improvement should come at no one's expense. This is fiction. Reality is that there are different demands that are mostly mutually exclusive: Politicians want to stay in power, businesses want unfettered ability to make money, common folk want to rise in wealth, and live comfortably and happily. Who's needs are you going to prioritize in this scenario? The maximum good?
Re: (Score:2)
"The plan is that by 2030 ‘You will own nothing, and you will be happy.’”
There is nothing in your link that supports that assertion.
Re:Can we please now start worrying about AGI? (Score:4, Funny)
What you really need are AI voters. Much of the electorate ends up voting against its own best interests, because they are misinformed and constantly bombarded by propaganda. If voters voted more rationally and in their own interests, the outcomes would be much better.
Or better still just abandon the two party system and move to proportional representation and coalition governments.
Re: (Score:2)
The solution to governments, which lead to corruption in nearly all cases, is simple. People need some form leadership. The removal of the need for that leadership means eliminating the people that have the need. Skynet logic 101.
Re: (Score:2)
Governments aren't interested in evidence based policy. They're interested in vote-getting based talking points.
Re: Can we please now start worrying about AGI? (Score:1)
Correction: politicians arenâ(TM)t interested in evidence based policy. There are plenty of government workers that are interested in evidence based policy, but are hampered but budget constraints and arbitrary rules set by politicians.
Re: (Score:2)
Re: (Score:2)
I think we already have that. The blockchain. /troll
Re: (Score:2)
Re: (Score:2)
So are we allowed to be concerned now, please?
I'm more concerned by the fact that someone is dictating to you what feelings you are and are not allowed to feel. Are you okay? Should we send a rescue team over?
Re: (Score:2)
Neither Go, nor any Poker playing AI, or any other AI for that matter has General Intelligence. It's not just a matter of putting more Neural Networks in another larger hierarchy and adding more and more parameters. The Go playing system was a large deep learning neural network that played many games to figure out the patterns, but it's not clear that that information can be transferred into another neural network system not designed for the one specific GO purpose. The size of current deep learning systems
Re: (Score:1)
A board/card game is easy to emulate such that billions of learning sessions can be ran against the AI training engine. But real life interacting with humans can't be done near that volume because it's more expensive to interface with actual people (who are serious).
Re: (Score:2)
AI Go had been making progress continually for a long time [imgur.com]. If you thought it would never be solved, it's because you weren't paying attention. Google finished it in part by using a modified algorithm, and in part by throwing way more hardware at it than anyone else.
Can we please now start worrying about AGI?
AGI will require new algorithms that we haven't discovered yet. It could theoretically happen tomorrow, if someone invents the algorithm. I don't know why that would worry you, but go ahead and worry if you enjoy that.
Re: (Score:2)
If you thought it would never be solved, it's because you weren't paying attention.
I didn't think it wouldn't be solved (although there were certainly people making that argument who I agree weren't really paying attention or were motivated by a belief in the specialness of humans) but it did seem like it happened much faster than anyone was expecting.
AGI will require new algorithms that we haven't discovered yet. It could theoretically happen tomorrow, if someone invents the algorithm. I don't know why that would worry you, but go ahead and worry if you enjoy that.
AGI if we don't get it right poses an existential risk. And yes, it will require new algorithms, but these results show that we are getting a better and better idea of understanding how thinking processes work or at least discovering algorit
Re: (Score:2)
Artificial General Intelligence (AGI) is not the type of AI which needs to be feared in 2022. It will almost certainly be developed eventually, but today's AI (often criticized as just being algorithms) has plenty of society altering possibilities all on its own. Even if incremental improvements to today's AI techniques can only accomplish 10% of what humans can do it is impactful. Just going from 5% unemployment to 15% is enough to cause major upheaval.
Anyone worrying about AGI with the current state of te
Re: (Score:2)
I bailed. (Score:2)
I played on line poker from time to time for a few years. I was "up" overall... but the lowest paid coffee slinger probably got paid a comparable hourly rate. I never felt the need to move into higher stakes. My home game was for that. But I gave up when demonstrably reasonable bots began to be available. I could see the writing was on the wall. I tried one out, and it was... pretty good. That meant it would be vastly better, and soon.
And so I stick to my reliable home game, populated by gems and assholes i
Re: (Score:2)
But I gave up when demonstrably reasonable bots began to be available.
Bots that could beat all but the worlds top professionals have been playing online since before 2011. Bots capable of beating most top pros were available in 2018.
Pokibot was developed in 2002, and it could beat most non professionals.
Re: (Score:2)
> Pokibot was developed in 2002, and it could beat most non professionals.
This thing (https://www.robot-advance.com/EN/art-pokibot-robot-ycoo-random-color-1884.htm) absolutely sucks at poker.
I for one welcome our new overlords (Score:5, Insightful)
When enough players use AIs to help them play, the AIs will practically play against each other using humans as useful idiots in between. There won't be Skynet. The AI's will battle against each other by trying to make their opponents' human avatars lives as miserable as possible by crippling their cars, overdrafting their credit cards, locking them out of their homes, inserting photos of their significant others with strange people on their phones and pushing notifications on cheap liquor nearby.
Re: I for one welcome our new overlords (Score:4, Funny)
Re: (Score:2)
When enough players use AIs to help them play
Players aren't using the AI/solver software during the game (at least the majority aren't). They are using it to learn and understand patterns of play - the type of hands that work best as a check-raise for a particular type of flop; types of hands that work well for bluffing three streets; etc.
Natural Selection -- the Fittest AIs will live (Score:2)
and the others die.
That is a critical point that defines the future in general. AI's moral values will not be determined by us any more than our own moral values are. They will be determined by natural selection.
https://computersthink.com/ [computersthink.com]
There are TWO distinct forms of Poker AI. The first is simply calculating the odds. OK, so not simple, but algorithmic.
The second actually plays the games against other human (or AI) players. Now there is a whole new dynamic involved, which includes evaluating how oth
Ball spitting machine would count as AI? (Score:1)
All this guy is doing is practicing his game off line with a machine ...
Bluffing (Score:5, Funny)
But the computer has a tell. It's CPU fan kicks on when it gets a good hand.
Cheating (Score:2)
This is cheating right? He uses this with online poker and no actual casino would let him through the doors. Or am I missing something?
Re: (Score:2)
Re: (Score:2)
After analyzing thousands of hands you will eventually get an innate understanding of how to play which hands under different circumstances. He doesn't need to bring the computer into the casino to analyze every hand before playing. He has studied enough hands (with the computer software showing him optimum play) that he can now play those hands from "muscle memory".
It would be rather hard for a casino to kick him out for "cheating" and in poker they probably don't care much since they are making their mone
High Noonbot (Score:1)
This RAM ain't big enough for the both of us!...