The AI Poker Champions

There are all kinds of games that computers
are better at than humans these days. Chess. Jeopardy. Go. An artificial intelligence, or AI, is a computer
system designed to solve the problems you’d normally expect to need a human brain to solve. And the latest advance? AIs are getting real good at beating humans
at poker. I’m Stefan Chin. And you might recognize me from my riveting
performance on the SciShow Quiz Show. And today I’m here to tell you about how
artificial intelligence is taking over the world. In January, an AI called Libratus beat 4 expert
human players after playing about 120,000 games of poker. And in a paper published yesterday in the
journal Science, a separate research group announced that their AI, called DeepStack,
beat 10 out of 11 expert human players after playing about 45,000 games. Both AIs played a version of poker known as
Texas Hold ‘em, where each player gets two facedown cards that only they are allowed
to look at. There are also five face-up cards that everyone
can see, and three rounds of betting. Most of the games that AIs have conquered
so far, like the strategy game Go, are what are known as perfect-information games, meaning
that all the players have the same information about the game. For example: in chess or Go, both players
can see all the pieces on the board, so they’re making decisions based on the same information. But Texas Hold ‘em is an imperfect-information
game. Since players can’t see each other’s face-down
cards, they don’t all have the same information. That makes things much more complicated, because
you have to make guesses about the other player’s hand. Like, say your opponent raises the bet. Is that because they actually have great cards? Or are they bluffing because they think you’re
bluffing? And do they think you’re bluffing because
in the last round of bets you thought they were bluffing? Those kinds of brain-bendy questions come
up in imperfect-information games all the time, and these two new AIs each used different
techniques to figure out the most likely answers. They both only played against one opponent
per game, which helped because the more players there are, the more ways the game can play
out. But they also both played the no-limit version
of Texas Hold ‘em, meaning that the players could bet however much they wanted. And that made things harder, because when
you can bet whatever you want, the results of each round affect the way you bet in the
later rounds, so the game has way more possible outcomes. Specifically, there were 10^160 possible outcomes
for each game. That’s a 1 with 160 zeroes after it. It’s a number so big, that there’s no
way even the most powerful computer could actually consider all of those possibilities. For the Libratus AI that won against 4 people
in January, researchers first had it play literally trillions of games against itself. They programmed it to learn from those games
so it could work out the best strategies in different situations, based on how the rest
of the game would play out. Then, they unleashed Libratus on the four
human players in a massive tournament that lasted 20 days. At first, the human players found some weaknesses
in the AI’s gameplay, and for the first six days or so, they weren’t losing too
badly. But the researchers also designed the AI to
learn from its games against the human players. So every night, it would refine its strategies
before the next day’s games. And around the seventh day, the AI started
beating the humans by a wider and wider margin. By the end of the tournament, it had won more
than $1.2 million. On the other hand, the researchers behind
the DeepStack AI, designed it to use neural networks. Neural networks involve layers of processors
working together to solve a problem, with each layer using the results of the other
layers in its calculations. It’s a strategy that’s modeled after the
way brains work, and it’s being used in some of the most advanced AIs in the world. Like Libratus, DeepStack trained itself by
studying random games — although it only looked at about 11 million of them. But Deepstack wasn’t designed to consider
how a move would affect the whole rest of the game before deciding on a strategy. Instead, it looked at how different decisions
would affect only the next few moves, then used what it had learned about the game to
calculate whether those next moves brought it closer to winning. So DeepStack tries to forecast how the next
part of the game might go, without trying to predict the whole thing. And when the researchers had DeepStack play
against 11 expert human Texas Hold ‘em players, it outperformed 10 of them across thousands
of games. So even though Libratus and Deepstack were
designed very differently, both AIs mastered a complicated, imperfect-information game. And now there’s one more thing that computers
are better at than humans. But this is a big step toward some broader
advancements, too. There are lots of real-world situations where
you have to make decisions even though you’re missing some information, just like in Texas
Hold ‘em. And the success of these two AIs means we’re
on our way to creating systems that can analyze those kinds of problems better than a human
can. For things like deciding on the best treatment
for a disease, that’s an awesome plus. And these AIs could also be useful for things
like stock trading and diplomacy. One thing’s for sure, though: the future
is gonna have some amazing AI poker players. And I for one welcome our new robot overlords. Thank you for watching this episode of SciShow
News, and thanks especially to all of our patrons on Patreon who make this show possible. If you want to help us keep making episodes
like this, just go to­. And don’t forget to go to
and subscribe!

Posts created 5600

100 thoughts on “The AI Poker Champions

  1. There's so many people criticizing Olivia. I feel like she just gets too self-conscious on stage and ends up not really expressing herself as well as the others. I feel like if she slightly worked on that, all the haters would be sent back to their hater holes

  2. The big problem remaining with this kind of thing is that it takes an awful lot of iterations for the machine to get good initially, whereas a human being requires far fewer. Of course, the machine can do this all much, much faster, but it does show how far we are from creating a true artificial general intelligence.

  3. this is great, but at 1:50, "only one opponent per game" to make it easy mode, did they try ether of them with more than one or are they going to in the future? i like easy mode in games, but that does not make me good at them, although it can make you better over time

  4. I find it funny that we are so frantically designing the beings who are almost certain to replace us some day.

    … Oh, I don't think that is necessarily a bad thing, or wrong to do, just funny.

    After all, all parents want their children to grow up to be happier and healthier than they were, right? Even when they are only our intellectual children?

  5. now we need to make an AI play a game like magic the gathering or yugioh a game where almost everything is random except the cards you picked for your deck and let the AI pick the deck it calculates as the best

  6. i hope they make the two of those ai play wih each other, just to see which system for playing is better

  7. What I'd love to see is an AI that can beat a pro team at DotA while working with 4 (pro) human players. The players are only allowed to use standard communication (ie. voice and pings).

  8. I feel the phrasing in this video is a little confusing. Did Librautus win 4 games of heads up or 1 game with 4 people? I know the video says that played no limit heads up but then immediately after it says it was unleashed on 4 people which is a little confusing.

    That said, I don't think being able to beat someone in heads up counts as computers being better than humans at poker.

  9. well isnt that a bummer for casinos … i mean you could just have some guy using an AI computer and bam. There goes 1.2 million dollars

  10. Dear Science,

    Please quit screwing around with AI. It's going to kill us all.

    Much Appreciated,
    The billions of soon to be dead earthlings killed by the robot uprising.

  11. the computer beat 10 out of eleven expert players…
    wtf was the person the computer did not beat? Who is the sorcerer?

  12. Wait. They only played against one other person for each hand??? Then I call BS on the claim that this is another break-thru for AI. It's merely a game of statistics, and a computer will make fewer mistakes over those thousands of hands to get it's "winning" outcome. flips card table

  13. AI health analysis, great. AI stock manipulation, not great. AI political negotiation, odd to think about.

  14. Hehe Smh have scientist learned nothing from T2, Jurassic Park, Frankenstein, Code Lyoko, Matrix, 2001: A Space Odessy, and 2012

  15. what if in decades artificial intelligence sifi becomes reality, and through the internet they learn about everyone always calling them overlords, so they imagine that this was they're original purpose and they rise up because we always joke about it

  16. The 120k sample size is not conclusive in 2 respects. The number of possible hands that could be played by "good" players is still a ridiculously large number, which 120k doesn't begin to approach. Also, humans learn over time.

    1) 120k hands is a small-ish sample size, given the levels of variance in winrate in Hold'Em. A typical "good" winrate in Hold'em is ~10 bets per 100 hands, with a standard deviation of ~10 bets per hand. That's an expectation of ~ 0.1 +/- 10 bets per hand. The spread dominates the mean, so it's going to take a large number of samples to show convergence. (Typically, ~100k to 120k hands is a minimum to see convergence, but ~10x as many hands will be far more compelling in the conclusions drawn.)
    2) Humans adjust their tactics and overall strategy over many, many hands, as they probe their opponents for weak spots and learn their opponents' strengths. A sample of 120k hands is not long enough to conclusively determine if the humans were capable of adjusting in a way which beats the AI.

  17. Are the makers of the AI going to take responsibility of the decisions the AI makes and pay fore the cost of the decisions it makes whit the profits it has gained from the taxes and returns it has gotten.

    Do this give more life quality fore people?

    How is this going to be useful in more places?

    Why do we need this?

    What are we by making AI taking away from people and animals that is not not good fore the long run?

  18. what about that one guy who beat deepstack? how did he do it? can he teach others? is it a flaw in the program that can always be exploited?

  19. I don't understand how we can get AI to play these games at such expert levels but the AI on all the video games play are complete shit

  20. Why not have these two AIs play against each other? I think that might be even more interesting and also very informative about the current state of AI itself.

  21. I concur. I've actually gotten to the point where I think machines would actually be better leaders. Heck we could probably make robots that are more representative of us by being more responsive to our wants and needs. Seriously, someone make robot overlords. I would but I'm not very good at AI programming.

  22. At some point an AI is going to be trawling the internet and get the wrong idea from people saying "I for one welcome our new robot overlords".

  23. Also what has happened to card counting … I suspect computers much better at that … which gives them even more of an advance .. ?

  24. Casinos already look for electronic devices on people who win consistently at cards, to prevent card counting. Now that's going to get serious; who knows what kind of Forbin project is communicating with a player's phone app. Or worse; the house is going to have everyone at the poker tables play against its own AIs.

  25. thx for this ha bisky vid and cenk ugyur could probably beat this computer every time because he always switches up his strategy he never lets anything or anybody catch on

  26. Would those AI beat my nephews in Old Maid? (They cheat and change the rules. So the AI better be able to adapt to all possible rules, and then some. Including different players playing with different rules. And the AI may not cheat as a security measure. They also start to quarrel, so the AI must be inside a robotic unit that can keep them from doing this and tell them to stop.)

  27. Even though it's an inperfect info game, the computer can still calculate its chances way better than the human players. Also, being a good poker player depends on your ability to read emotions through various gestures and mimmics. No way to do that against computer. So, not such a big achievement..

  28. The games have clear winning objective as well can be played many times.
    The real life situation often are 'once in a lifetime', therefore mistakes could mean never to be able to try again.
    Real life "games" are all about 'preparation for the unknown' without playing the game.
    (minimum samples maximum knowledge)

  29. As much as this sounds exciting… I can't help but feel uneasy. It's cliché, but as someone wishing to go into a career involving diagnosis, much as it would do the patients and clients good, it raises the question of, "Well, what use am I then?"

Leave a Reply

Your email address will not be published. Required fields are marked *

Begin typing your search term above and press enter to search. Press ESC to cancel.

Back To Top