AI Wins $290,000 in Chinese Poker Competition (bbc.com) 81
An AI program has beaten a team of six poker players at a series of exhibition matches in China. From a report on BBC: The AI system, called Lengpudashi, won a landslide victory and $290,000 in the five-day competition. It is the second time this year that an AI program has beaten competitive poker players. An earlier version of the program, known as Libratus, beat four of the world's best poker pros during a 20-day game in January.
useless article (Score:2)
Re:useless article (Score:4, Informative)
It is heads up, no limit.
Re: (Score:2)
Who cares, real question is what is the AI going to spend the winnings on.
Details? (Score:2)
Go, Poker, Chess, Jeopardy... (Score:5, Funny)
Who'se your daddy?
I don't know that this is the official moment when AI becomes smarter than us,
but I do suspect strongly that current AI could handily beat Donald Trump at the task of rationally governing the world's most powerful and dangerous nation, and I for one know which one I would vote for.
grammaro (Score:2)
"who's your daddy?"
Re: (Score:2)
Sadly, if it were Trump or Siri, I'd reluctantly have to go with Trump.
Re: (Score:2)
I don't know that this is the official moment when AI becomes smarter than us,
It's not, this is just weak AI. Unless you also think a calculator is smarter than us, which, from a certain perspective, it is.
Re: (Score:2)
In other words, you are looking at it wrong: Strong AI is not a
Re: (Score:2)
Re: (Score:2)
Roughly speaking, what it was programmed to do, was "learn". So this whole argument line that "it did exactly what it was programmed to do" is specious.
I know there are more specifics that get put in by humans still these days, and that human guidance in selecting training data and in tagging data etc is still required, but the extent of human intervention in "learn" is decreasing considerably, and fairly soon "learn" will be completely general.
Re: (Score:2)
Roughly speaking, what it was programmed to do, was "learn".
Very roughly. That's how we explain it to people who are more comfortable with anthropomorphization than with the actual workings of neural networks. It is more appropriately described as a curve fitting problem, with a heuristic in place because the domain is too large to actually fit in computers.
but the extent of human intervention in "learn" is decreasing considerably, and fairly soon "learn" will be completely general.
This is the sort of comment that makes me think you've never writ
Re: (Score:2)
AlphaGo taught itself Go by playing millions of time with itself.
Re: (Score:2)
or can do logic and arithmetic without going through manual algorithms like long division
How would you even do that?
No Human Element? (Score:2, Interesting)
I thought Poker was a game of understanding your opponents not only based on past actions with cards but also by looking at facial expressions, body language and determining whether or not they have a good hand. Along with that, a big part is developing subtle gestures to throw your opponents off.
Without this information, isn't this win somewhat random or "lucky" and not really indicative of how the AI can play against other humans?
It's interesting that the AI can develop a database on other player's style
Re: (Score:1)
See, your mistake is assuming that those things (facial expressions and past actions) matter in a game where the winner is very clearly almost always the person with the best chances at winning based on the cards they hold. The facial expression and past action thing is just another way to be wrong, whereas going by statistical likelihood is almost always the correct bet.
Are they bluffing? The look on their face says they're bluffing! But wait, what if they made that face to make me think they're bluffing..
Re: (Score:3)
See, your mistake is assuming that those things (facial expressions and past actions) matter in a game where the winner is very clearly almost always the person with the best chances at winning based on the cards they hold.
This ignores the fact that a great many no-limit hands do not go to showdown. The worse hand wins frequently.
Re: (Score:2)
Re: No Human Element? (Score:1)
Not really. Laying down the best hand when you are relatively weak and facing a large bet, waiting for a better spot is very frequently the best play.
When you've got a pair of nines on the turn or river with overcards on the board, you're gonna fold to Ace high frequently because in that spot, you are going to lose often enough such that you don't have odds to call in no limit, particularly in a tournament.
Players aren't mind readers, you play the player, and you play the odds, and even highly experienced
Re: (Score:3)
If you bet solely on stats, you might do OK against a bunch of amateurs, but pros will wipe the floor with you. Because they not only know the stats backwards, but can also read your tells, and know when and how to bluff you.
I'm curious as to how a computer does that, but I don't for a moment doubt that it's possible.
Look at it the other way... a machine has no tells (apart from maybe calculation time) so it removes an ability from the opponent, in effect lowering the opponent from pro to amateur playing the stats, and humans might know the stats to within a couple percent but the computer will know them exactly.
"Bring them down to your level and beat them with experience"
Re: (Score:2)
See, your mistake is assuming that those things (facial expressions and past actions) matter in a game where the winner is very clearly almost always the person with the best chances at winning based on the cards they hold. The facial expression and past action thing is just another way to be wrong, whereas going by statistical likelihood is almost always the correct bet.
Are they bluffing? The look on their face says they're bluffing! But wait, what if they made that face to make me think they're bluffing...
You should bet on stats, not on feels. AI is good with stats.
I believe you are partially correct in this case.
Bluffing doesn't only mean facial expression or posture that required vision sensors. For humans, we usually have reaction to our surrounding, so facial expression and posture are quite obvious to others. Because of the reaction, our decision becomes dynamic and could be changed depending on what's going on around us. That's what bluffing is about. The AI, in this case, eliminates the obvious part of bluffing, but may still analyze certain behaviors of oppone
Re: (Score:1)
Nowadays, most of the best poker players in the world learnt online. They rely more on reads from betting patterns than anything else - working out what hand their opponent has based on their bets in the previous rounds.
World-class human players don't tend to give much away, although they would probably be able to get tells from you and your buddies playing a game for fun.
TL;DR Tells aren't a big thing in high-level poker anymore.
Re: (Score:1)
I thought Poker was a game of understanding your opponents not only based on past actions with cards but also by looking at facial expressions, body language and determining whether or not they have a good hand. Along with that, a big part is developing subtle gestures to throw your opponents off.
Without this information, isn't this win somewhat random or "lucky" and not really indicative of how the AI can play against other humans?
Nope. The game is pure statistics based on the known cards, which computers are very good at. Collecting a database on your opponent's habits is also far more useful than looking at their face could ever be. After all, their body language is only an inevitably misleading stay stream of data, controlled entirely by your opponent, whose entire purpose is to throw you off. You're much better off completely ignoring that distraction.
Also, what better "poker face" is there than an emotionless calculating mac
Re: (Score:2)
Re: (Score:2)
Now consider that the computer realizes your strategy, and starts calling your bluffs. That's when you have a game.
Re: (Score:2)
"People think that bluffing is very human," Mr Brown told Bloomberg, "It turns out that's not true." "A computer can learn from experience that if it has a weak hand and it bluffs, it can make more money."
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Seriously though, zen
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
I thought Poker was a game of understanding your opponents not only based on past actions with cards but also by looking at facial expressions, body language and determining whether or not they have a good hand. Along with that, a big part is developing subtle gestures to throw your opponents off.
Reading body language only works with really bad players.
What modern poker AIs do is play 'game theoretically optimal' (GTO). If both players play GTO, then it is a draw, but if your opponent deviates from GTO then your GTO strategy will beat their strategy. If you are playing 'bad' players you can deviate from GTO and win even more than the GTO win rate (this is called exploitation), but it is better to simply play GTO until you know for sure that your opponent has weaknesses to exploit that you can safe
Re: (Score:2)
Hollywood much?
Also, cops fight crime mainly by experience tragic science field trips in early childhood.
Here's a pro tip. If the actors are loving it, it has probably had the living Snopes kicked out of it, suppo
“You can never tell what’s in a Chinam (Score:2)
Re: (Score:2)
Re: (Score:2)
No.
In no-limit hold'em, you see a maximum of seven cards before showdown, the five board cards, and your two hole cards.
Now can we hear about AI horribly losing? (Score:2)
It's one thing to hear about AI's "wins", but it's another to see it handily (and happily) losing to humans where it should have won.
Re: (Score:2)
Who gets the prize? (Score:4, Interesting)
the prize money will go to Strategic Machine, a firm founded by the duo.
That seems a little unfair. If I had won, the prize money would not have been given directly to my parents. If a machine wins, it should receive the prize. If it cannot actually spend it, then that would appear to be a rather basic limitation to its AI-ness. But it wouldn't be a problem for the competition or whoever awarded the prize.
You would also hope that the authorities would keep an eye on the money to ensure that whoever had access to the AI didn't defraud it of its winnings. Maybe it is time for machines to have property rights. And if they are going to be awarded assets, maybe they should be taxed on them, too.
Re: (Score:2)
the prize money will go to Strategic Machine, a firm founded by the duo.
That seems a little unfair. If I had won, the prize money would not have been given directly to my parents. If a machine wins, it should receive the prize. If it cannot actually spend it, then that would appear to be a rather basic limitation to its AI-ness. But it wouldn't be a problem for the competition or whoever awarded the prize.
You would also hope that the authorities would keep an eye on the money to ensure that whoever had access to the AI didn't defraud it of its winnings. Maybe it is time for machines to have property rights.
So you're saying the poker playing computer needs to have a module added that will let it shitpost "Muh Freedums!" on Twitter? And order vast quantities of alcohol online, of course. 'cause that's its fuel. I saw it on a documentary [blogspot.com] so it must be true.
Re: (Score:2)
That seems a little unfair. If I had won, the prize money would not have been given directly to my parents. If a machine wins, it should receive the prize. If it cannot actually spend it, then that would appear to be a rather basic limitation to its AI-ness.
Actually, if a human is incapable of making a decision, a spouse or parent can have power of attorney and make choices on their behalf. So if an AI is incapable of making a decision, why should it be any different?
Re: (Score:1)
I think his point, made in jest, is that it's not an AI. It's a piece of poker software. Following the money makes that clear.
Who? (Score:2)
Please be more exact in the headlines... I thought Ai Wei Wei now switched to Poker...
Can't wait till they put AI in a robot (Score:1)
Chinese poker is not Poker; there's no bluffing (Score:1)
In other news, hydraulic crane outlifts a human. (Score:2)
At what point will it cease to be considered news when computers beat humans at some game, especially when the game has a large computational element?
Machines beat us at all sorts of tasks. They're stronger, faster, more precise. Of course they will drive better than humans, play chess / checkers / go / poker better than humans, etc.
Re: (Score:2)