AI Decisively Defeats Four Pro Poker Players In 'Brains Vs AI' Tournament (ieee.org)

Halfway through the "Brains vs. AI" poker competition, it was pretty clear the artificial intelligence named Libratus would end up victorious against its human opponents, who are four of the world's top professional players. Lo and behold, Libratus lived up to its "balanced and forceful" Latin name by becoming the first AI to beat professional poker players at heads-up, no-limit Texas Hold'em, reports IEEE Spectrum. From the report: The tournament was held at the Rivers Casino in Pittsburgh from 11-30 January. Developed by Carnegie Mellon University, the AI won the "Brains Vs. Artificial Intelligence" tournament against four poker pros by $1,766,250 in chips over 120,000 hands (games). Researchers can now say that the victory margin was large enough to count as a statistically significant win, meaning that they could be at least 99.7 percent sure that the AI victory was not due to chance. Previous attempts to develop poker-playing AI that can exploit the mistakes of opponents -- whether AI or human -- have generally not been overly successful, says Tuomas Sandholm, a computer scientist at Carnegie Mellon University. Libratus instead focuses on improving its own play, which he describes as safer and more reliable compared to the riskier approach of trying to exploit opponent mistakes. Even more importantly, the victory demonstrates how AI has likely surpassed the best humans at doing strategic reasoning in "imperfect information" games such as poker. The no-limit Texas Hold'em version of poker is a good example of an imperfect information game because players must deal with the uncertainty of two hidden cards and unrestricted bet sizes. An AI that performs well at no-limit Texas Hold'em could also potentially tackle real-world problems with similar levels of uncertainty. In other words, the Libratus algorithms can take the "rules" of any imperfect-information game or scenario and then come up with its own strategy. The Libratus victory comes two years after a first "Brains Vs. Artificial Intelligence" competition held at the Rivers Casino in Pittsburgh in April-May 2015.

Comments

care less

[Anonymous Coward]

Apparently someone didn't care less as I did.

This isn't news. This is something I'd expect FARK to make fun of.

Re:care less

[Anonymous Coward]

I wish they would stop using bullshit terms like AI. It's not AI, it's a strategic algorithm. If it were AI, it would be sentient.

Re:care less

[aldousd666]

You're actually not right. It can be AI without being sentient, and in this case, it is just that. It's a general purpose learning algorithm. Not a strategic poker playing algorithm. It doesn't need to be sentient to be intelligent. You're confusing General AI with Narrow AI. This is a Narrow AI, to be sure, but if you string enough Narrow AI's up together, they can eventually give the same appearance of a General AI. This is just one milestone along the way. In particular, it dethrones the idea that poker is the last bastion of human dominance in cognition. Obviously we'll have to find a new bastion, like the fact that we are, so far, the only General Intelligence thus far observed or produced.

Re:care less

[Anonymous Coward]

Wrong.

A.I. = Artificial Intelligence = sentience.

Re:care less

[Anonymous Coward]

right = where's yor god

Re:care less

[Jon Peterson]

it dethrones the idea that poker is the last bastion of human dominance in cognition

I think that idea was dethroned when Bill "The Bluff" Higgins got a train to Boston in 1872 with his pockets full of winnings, and strode into Harvard saying "Gentleman, the finest minds in the world have recently met in the back room of McKluskey's Hotel, and held a competition of arithmetic, stoney faces and drinking, to determine a winner. And I am that winner. Your work, professors, is needed no more. A dominant mind has been chosen."

Seriously, what makes poker attractive as a benchmark of AI is that it is essentially simple (like chess, go and most other card or board games), but contains a very small element of modest complexity (bluffing).

Compare that to a game like Pictionary. Pictionary is vastly, vastly more complex than poker. When two computers with cameras and screens can beat a pair of humans at pictionary, I'll be impressed.

Re:care less

[Maritz]

You can have intelligence without sentience. It's a question of degree. Do you do nuance, at all?

Re:care less

[Anonymous Coward]

What makes it attractive and a new milestone is that it is an imperfect information game unlike chess and Go. Chess could be done using brute force, Go needed machine learning for pattern recognition giving it a sort of "intuition". Pictionary I'd expect not to be that difficult, there are already NNs for converting text into images and images into text and it is again a perfect information game.

Re:care less

[Anonymous Coward]

When two computers with cameras and screens can beat a pair of humans at pictionary, I'll be impressed.

I doubt that you actually will be. You'll just move the goalposts.

Re:care less

[hcs_$reboot]

When two computers with cameras and screens can beat a pair of humans at pictionary, I'll be impressed.

Given that, recently, computers do way better than humans at facial recognition, pictionary sounds that a big deal compared to poker.

Re:care less

[Verdatum]

If AI and sentience were one and the same, we wouldn't call it AI, we'd call it AS. Besides, sentience is a philosophical concept. The definition of which is the ability to have "subjective perceptual experiences". And going by that definition, one can argue that such a learning algorithm is indeed sentient. AI research does not particularly use the word "sentience" because it is such an abstract concept that it doesn't really add to the field of study. "Intelligence" just means that a thing learns, and potentially modifies it's behavior based on previous experiences; and that's unarguably exactly what this does.

Re:care less

[Anonymous Coward]

Compare that to a game like Pictionary. Pictionary is vastly, vastly more complex than poker. When two computers with cameras and screens can beat a pair of humans at pictionary, I'll be impressed.

It's not exactly what you're asking for, but...

Re: care less

[Anonymous Coward]

Repeating an erroneous statement doesn't make it stronger. GP is citing the academic definitions of these terms. You are citing your posterior source.

Re: care less

[Anonymous Coward]

Wow. That is pretty amazing.

my AI can beat your AI

[turkeydance]

oh no it can't. yes it can. no it can't.

Re: my AI can beat your AI

[Anonymous Coward]

News: Kid beats AI at the "annoying" game.

Did it 'cheat'?

[Anonymous Coward]

Did it count cards? Did it have a camera and was reading microexpressions, while it has no face and therefore had an unfair advantage? By the way why should anyone really give a damn about this?

Re:Did it 'cheat'?

[Applehu Akbar]

Perhaps it belches cigar smoke at chosen times.

Re:Did it 'cheat'?

[xevioso]

Because online poker companies around the would would love to get their hands on the software and plug it into their system, so that you'd never know if you were playing a computer or a real person. If you were playing the computer, you'd always lose.

Of course they could program their computers to simply know what you have in your hand and bet accordingly, but still..

Re:Did it 'cheat'?

[alexo]

1. Poker companies make their money from the "rake", which is a (capped) percentage of the pot. They get their money regardless of who wins.

2. Poker companies want you to keep playing, because that way you'll continue contributing to the pot (and therefore, to the rake). Players that lose too much, leave. That's why some poker companies offer tutorials and "schools" to retain players by helping them improve.

3. Poker companies that operate in highly regulated markets have their code audited, their installations inspected and each and every hand submitted in real-time to the regulatory authority for possible future scrutiny. Player trust is very important, being caught cheating will cause the players to switch to the competitors.

Re:Did it 'cheat'?

[Capsaicin]

Because online poker companies around the would would love to get their hands on the software and plug it into their system, so that you'd never know if you were playing a computer or a real person. If you were playing the computer, you'd always lose.

No, with the probabilities involved in poker no player (AI or otherwise) can always win. Rather you'd loose over the long term, but that's already the case for most players.

Moreover, there is no financial incentive for on-line poker companies to have you always lose, or lose in any given hand. Like all effective gambling products (even those, unlike poker, where the house is your counter-party) what they want you to do is to win often enough to keep you coming back for more.

Of course they could program their computers to simply know what you have in your hand and bet accordingly, but still.

Exactly, and thus largely eliminating the chance to which the AI is still subject. But again they wouldn't because they have no financial incentive to do so. They make their money from the rake remember.

If on-line poker companies were to rig the system it would instead be to produce situations which encourage heavy post-flop action. The accusation that they do so is often made by players who have experienced (what they intuit as) more than their fair share of bad beats. But again the probabilities of Holdem will see to frequent bad beats even without undue interference. Now let me tell you about the time I had this perfect read on the villain, I was K full of Qs and he was Q full of As and then the river came a ...

Re:Did it 'cheat'?

[Anonymous Coward]

#3 is funny

Quote yourself

Re:Did it 'cheat'?

[Maritz]

He has credibility, you don't.

Re:Did it 'cheat'?

[Paradise Pete]

They don't have to cheat. It's very important that a poker site offers a wide variety of games and bet sizes, and have a large number of active tables. Being able to plug in bots, with skill levels tuned to the game, would be very lucrative even if the bots are set to play only well enough to break even.

Re:Did it 'cheat'?

[alexo]

Running an undeclared bot that takes players' money is worse than cheating, it is criminal fraud (or possibly worse, I am not a law expert). The premise of poker is that you do not play against the house, you play against other players and pay the house for providing the venue. Given that the Libratus' play style measurably differs from a human's, it will be easily detected by analysis.

You might have as well suggested that the poker company fudge the amount of money in players' accounts and hope nobody notices.

Re:Did it 'cheat'?

[alexo]

#3 is not funny, it is a pain in the ass due to the number of hoops one needs to jump through and having the software flexible enough to support all the different requirements the regulators come up with.

I don't have publicly-accessible links to give you, but if you are interested to learn more about the subject, I suggest researching the French regulations from ARJEL and the Italian ones from AAMS.

Re:Did it 'cheat'?

[Verdatum]

For future reference, people asking "why does this matter?" or "why should anyone give a damn?" on slashdot is trollbait. You'll find at least one comment asking this in nearly every post.

Re:Did it 'cheat'?

[Anonymous Coward]

You'll also find at least one asshole like you complaining about someone complaining about something. Just STFU, the internets are shit anymore anyway so what does it matter? Nothing that matters happens here, it's all just WORDS WORDS WORDS WORDS and may as well just be a bunch of retarded bots jabbering at each other.

Re:Did it 'cheat'?

[Paradise Pete]

You're ignoring the "play well enough to break even" part. And why would they have to do it in secret? Card rooms have always employed paid shills to fill empty seats to keep games going or to start a new one. It's not secret. I worked as one for a time, and the players knew it. I also won money, but the bots don't have to, and the card room would have preferred that I didn't win. (most shills don't win. They hope to break even and live off the salary.)

Re:Did it 'cheat'?

[alexo]

The problem with this approach is that it changes the role of the house in the game. In poker, as opposed to casino games, you can argue that the house is an impartial party, that has no stake in the outcome of the game. Should it change, it will open a big Pandora's box of complications that would keep many lawyers gainfully employed for a long time.

Re:Did it 'cheat'?

[Paradise Pete]

Not if they say that's what they're doing. They can even publish the bots' results showing that they are not making money, but merely filling tables.

Re:Did it 'cheat'?

[alexo]

The only company in the business that I am closely familiar with is Pokerstars, and they don't have a problem filling tables.
http://www.pokerscout.com/Site...

Re:Did it 'cheat'?

[Paradise Pete]

What does that have to do with it? You seem to be bent on being contrarian no matter how far we get from the original point.

Re:Did it 'cheat'?

[alexo]

The original claim was that "online poker companies around the would would love to get their hands on the software and plug it into their system" and I do not think that we strayed that far from it.
My point is that (at least for the dominant player in the business) the drawbacks and risks outweigh any potential benefits to the point that it will not even be considered.
If failing to bring convincing arguments to the opposite makes me "contrarian", so be it.

Re:Did it 'cheat'?

[Paradise Pete]

You started screaming fraud, now you're simply saying there's no need. One of the biggest problems of starting an online poker room is the lack of players. No matter how good everything else is, if it's hard to find the game a player wants he won't play there. That inevitably leads to a small handful of giants. This type of bot would allow competition to flourish. Ultimately, that's good for the player. The very person you said would be harmed.

Re:Did it 'cheat'?

[alexo]

I don't remember screaming anything, I tried to keep this conversation polite.
My fraud remark was about running undisclosed bots, which was what the original poster I replied to insinuated.

As for newcomer companies, I can see how having AI for players to play with will alleviate the problem of empty tables, but I also see how it would cause other problems. In no particular order:

Even assuming that everything is straight, level and upfront, it still creates the perception of a conflict of interests. After all, poker is played for money and now the house is perceived to have a stake in the game.

This will likely be disallowed by regulatory bodies, as it blurs the line between player vs player games and casino-style games. Regulators are a pain to deal with.

As I said, poker is played for money (I will address the play money case later), and arguably one of top (if not the top) players' goal is to win money. If the bots play too well, nobody would want to play them; if they play too badly, people will always try to play only against them.

In the case of play-money games, the reasons outlined above may not hold, but I fail to see the incentive of using such AIs in this case. Even the educational value is low since such AIs play very differently from humans (as per the article).

Now, there may be something I am missing in the picture, and you are more than welcome to point out any errors in my reasoning, but I suspect that without any hard data this will remain an academic argument.

I wonder if this could apply to human players also

[SuperKendall]

I don't play poker enough to know, but I wonder if many human players at the top level also try to win through discerning tells and weaknesses of opponents... if an AI can win so consistently is is using a technique that a human could also learn to get a step ahead of todays other human players?

Luck not a factor?

[al0ha]

Must be the first time ever I've heard of a Texas Holdem game where luck was not a factor.

Re:Luck not a factor?

[arbiter1]

IMO a lot of poker is about being able to read the person you are playing against and know when they are bluffing or not. There is an element of luck when you make a bad call but get the card you need to win.

Re:Luck not a factor?

[gravewax]

Luck tends to not be the most consistent factor in a tournament, that's why you frequently see the same professional players making it through to the top 50 and frequently even the final table out of thousands. When you take the human factor out and boil it down to purely mathematics (i.e. chance and betting patterns etc) I would expect a computer to win long term. Poker though is more than just mathematics, it is about reading people and quite often manipulating them, playing a computer isn't the same as playing poker at a table of people.

Re:Luck not a factor?

[Anonymous Coward]

Most of poker is based off math, not tells.

You get an idea of whether they are loose or tight players (i.e. bluff often vs bluff rarely), by their play/fold ratios on the various streets.

You get the idea of their possible hands from their betting patterns (e.g. if a player that doesn't play often raises pre-flop, and then tries to represent having a 8-5 for a straight on a rainbow board, it's not convincing. They'll have an over-pair to the board, or possibly a set at most).

And lastly you play what the odds tell you to play (if the odds are 25% that the player is bluffing, then if you only have to put in 1 chip for every 5 that are already in the pot, it makes sense to call, regardless of usually losing [because making that bet often enough will earn money in the long run (+EV - expected value)]).

Getting tells is a minuscule part of poker.

Re:Luck not a factor?

[Anonymous Coward]

Not over 120K hands it wouldn't be.

Re:Luck not a factor?

[alexo]

Luck can be a big factor in an individual hand, but in the long term, it evens out.

An interesting read:
https://www.bloomberg.com/view...

Re:Luck not a factor?

[fluffernutter]

You can't claim it's AI if you are using computer math skills. DIal it back to human math skills and augment that with an ability to read people with a camera, and then you have real AI.

Re:Luck not a factor?

[thesandtiger]

Over that many days and that many hands and that consistently? That wouldn't be luck, that would be a quantum miracle.

Re:Luck not a factor?

[Anonymous Coward]

tells play a much heavier roll earlier on in a tournament where there are still a lot of amateur/beginner players, that value rapidly diminishes throughout the days of a tournament though where then it does just become what you describe. I always hate the first day or two where you are trying to avoid the landmines of beginners who through luck can take you out just as easily as a professional can with good play. A good example of the danger of beginners was one of the pro players at my table in the last tournament, preflop he folded pocket Aces when one of the amateurs went all in, he showed the hand, the amateur could not believe that he folded and really that is the difference, an Amateur is thinking only a few hands ahead at most, a pro knows that if he calls such preflop moves every time he has pocket Aces he will not make the finals. (PS: pro in question got a 1 hour table ban for folding pocket Aces, he should have known better than to show them).

Re:Luck not a factor?

[dcollins]

Everything GP described is human math skills. Even rank amateurs poker players are doing all of that in their heads. It's elementary arithmetic/probability.

Re:Luck not a factor?

[fluffernutter]

But what element of human psychology was there in the way that the computer played? AI is not about winning games over humans, it is about mimicing human psychology.

Re:Luck not a factor?

[thesandtiger]

Which human's math skills? Humans have a staggering range of capabilities. Average? Then average in which way? A savant? Then a savant in which way?

And what about autistics who happen to be very good at poker but lousy at reading human expressions?

There are multiple different ways to be good at poker, and this system is just using one of them, and is clearly quite good at this particular way.

I also didn't see anyone making the claim that this was a hard or general purpose AI. It's not, and you acting like someone did make that claim is kind of weird. It's a system that is beating some of the best players at a game that was previously deemed to be very challenging to do computationally.

Re:Luck not a factor?

[fluffernutter]

I just don't see why it is news. It's another 'NEWS FLASH! Computers can out calculate humans" story.

Re:Luck not a factor?

[Nunya666]

Luck tends to not be the most consistent factor in a tournament, that's why you frequently see the same professional players making it through to the top 50 and frequently even the final table out of thousands. When you take the human factor out and boil it down to purely mathematics (i.e. chance and betting patterns etc) I would expect a computer to win long term. Poker though is more than just mathematics, it is about reading people and quite often manipulating them, playing a computer isn't the same as playing poker at a table of people.

I play a lot of poker, both in cash games and in tournaments. Luck is a HUGE factor in tournaments. That is why so many unknown players win them. If it was mostly skill, then only the pros would be winning them, which absolutely IS NOT the case.

There is a saying in poker tournaments: in order to win, to have to win your share of "coin flips." An example of a "coin flip" is my matched (paired) hole cards against your two cards that are higher than mine. For example, my pair of 8s against your King-Queen. My pair only has a 52% chance of winning the hand. You have a 48% chance of improving your hand with any King or Queen, or making a higher straight, flush, or full house than I do.

That being said, luck is only a factor at the end of the hand, when both of us have to show our cards. I often say that "poker is about everything except the cards, unless and until you have paid to see my cards." If I can make you fold with just my betting, then the cards are irrelevant. That happens in AT LEAST 4 out of 5 hands in live games. Most of the time, poker is about the size of my chip stack vs. yours, who bet first, how much they bet, how often they bet, how often they raise, how many hands they play (vs. just folding), how much they usually raise, do they have an aggressive image at the table, do they ever re-raise, have you seen them fold every time someone raises, etc., etc.

Re:Luck not a factor?

[Anonymous Coward]

120 hands....120K hands would be banned as cruel and unusual punishment for humans... LOL

Re:Luck not a factor?

[Anonymous Coward]

I am guessing you don't make many final tables. part of playing as a professional in tournaments is avoiding coinflip situations, cash games and tournaments are very VERY different beasts. Late in the tournament luck comes into it as stack sizes can force that situation, but if you are relying on coinflip luck throughout a tournament you are almost certain to be out long before the finals. yes a few unknowns always get through too, but you will see the same professionals time and time again in those final 100-200 players.

Re: Luck not a factor?

[Anonymous Coward]

That pro is a total idiot. You should never fold AA pre-flop. Unless it's some extreme ICM situation and a huge pay jump and it's like a multi-way pot where someone will get eliminated and you will have a big pay jump.

Re: Luck not a factor?

[Anonymous Coward]

In No Limit Hold'em, luck is not a factor. No Limit Hold'em is a game of skill, not chance. If it were all just luck, then all players would win/lose about the same ratio and that's not the case in No Limit Hold'em: there is a consistent group of players who make a very good living playing this game. If it was all luck, they'd be broke. Also, a few years ago now, a woman won the European championship and she never looked at her cards, not a single time. Her cards didn't matter one bit (luck played no factor). Instead, she merely read her opponents. I'd be interested to see who would win such a match against this Ai computer: there'd be absolutely no way for the Ai to get a read on her, or bluff her.

Re: Luck not a factor?

[Anonymous Coward]

That would be AH (artificial humans) not AI, unless you believe that only humans are capable of being human, in which case chimps and Caledonian crows may be aggrieved, and we should shut down SETI. You're not Noam Chomsky, are you?

Re: Luck not a factor?

[Anonymous Coward]

Head to head preflop: AA vs offsuit 72 wins 88.68% of the time.
That means 11.32 times out of 100 you're out of the competition.

Question: How many times can you call a preflop All-in with pocket aces against a chip matched opponent head to head and have a 90% chance of losing the tournament?
Answer: Not very many

The best thing to do in a tournament is to play super conservative for the first 25% to 50% of the game until most of the terrible players have wiped each-other out in the name of "table respect". The few morons that accumulate chips during this time period are too stupid to just fold until they're in the money so they end of serving as feedstock for the players who don't like to count on luck to get them ranking in competitions.

Re: Luck not a factor?

[fluffernutter]

Well congratulations! If AI has nothing to do with being human like then I guess you have solved the entire field of AI since computers have been able to do raw calculations better than a human since their inception!

Re:Luck not a factor?

[Anonymous Coward]

well the hallmark of idiocy is not understanding one's own limits

Re: Luck not a factor?

[nasch]

AI isn't thoughtless arithmetic either.

Re:Luck not a factor?

[Bryan Ischo]

I don't play poker but I've often thought that the ability to play with perfect conviction like you have a hand that you do not have must be a tremendous asset. I think your point that the actual cards don't matter until they are revealed at the end is a great point. If you can play such that except in rare circumstances you always convince your opponents that your hand is better than theirs, then you'll win more often, or at least, you'll be able to use those skills to eke out incremental advantage over time.

Of course, having the luck where you actually *have* the better hand more often must be an important component too ... but over the long run, since everyone will on average have the best cards roughly an equal number of times, it's gotta come down to the skill of manhandling other players into doing what you want through perfectly executed charades.

I don't really have any interest in poker at all, it bores me to tears, but it must be very satisfying to have that skill and to exercise it in games.

Re:Luck not a factor?

[Bryan Ischo]

Further to this, I've never played online poker but I've often wondered what it would be like to join up, and then play games pretending like I have completely different hands than I do have. Like what if I could actually hide the cards I actually have on the screen and just choose what cards I want to pretend actually appeared. Then I could play from my pretend hand without any uncertainty due to knowing what I actually have.

There's the problem that I haven't played enough poker to actually make smart bets and plays based on even made-up cards ... but if you are an online poker expert, would it help you in any way to have a system like I describe, where you don't even see your actual cards, but only see the cards of the hand that you want to "pretend" to have?

Re:Luck not a factor?

[Bryan Ischo]

I would imagine that with enough data you could calculate how long to wait before actually looking at your real cards to decide what to do next. At that point you could probably calculate your odds of winning if you keep playing your "fake hand" versus the odds of winning if you play your "actual hand" and then choose the best option.

What if you could somehow invert the problem of poker, and have, through enough mined data, a table of, given the current game situation, i.e. the best of prior players, etc, what hands you *could* have and what hands they *could* have, and then you could just choose the hand that you want to be playing as if you have at any moment.

Like, if, given the current game state, it's equally likely that you have hand A, B, or C, and if having A is most likely to win you the most money, then play like you have A, until that option is either not possible anymore, or is worse than having B or C (assuming they're still possible). Then switch to B or C when it's the best strategy.

I can't be the first person to have thought of this. Is there software out there that will do this for you?

Re:Luck not a factor?

[Anonymous Coward]

Sitting and playing for hours and hours is also a huge factor.

This doesn't seem that impressive

[vux984]

It sounds like its just better at calculating the odds than humans are, which is not much of a feat, really. I mean... it would almost be surprising if it couldn't.

I'm not trying to diminish the significance of the research... but what is the real innovation here?

Re:This doesn't seem that impressive

[Anonymous Coward]

Playing based on the odds means every opponent knows what you'll do 100% of the time... Guess how that turns out. I imagine (and I certainly haven't checked!) that the AI doesn't even have a table of odds of any kind. To win it just needs to know what actions make other people lose.

Re:This doesn't seem that impressive

[Anonymous Coward]

If "calculating the odds better than humans" was all that was needed, then pro Poker players would have been beaten by a computer decades ago, so something must be wrong with your assessment.

Re:This doesn't seem that impressive

[sexconker]

Playing based on the odds means every opponent knows what you'll do 100% of the time... Guess how that turns out.

No it doesn't.

Your opponent doesn't know what your cards are, but you do. You have disparate sets of information regarding the game state. Playing the odds perfectly and uniformly deterministically in ambiguous situations (all players making predictable choices when multiple choices have the same weight) results in different players arriving at different ideas of the game state and making different decisions, even if all the players are robots running the same exact deterministic algorithm that's a function of the known game state.

Re:This doesn't seem that impressive

[Anonymous Coward]

Yup. Future headline: AI kicked out of Las Vegas for card counting.

Re:This doesn't seem that impressive

[gweihir]

Nothing. It will also be better at bluffing, because a machine has no way to betray itself.

This is a nice stunt, but this is still just statistics or "weak AI", if you must label it as AI.

Re:This doesn't seem that impressive

[Imrik]

Not just the odds of card draws and hands, but odds based on current betting and past behavior of the opponents.

Re:This doesn't seem that impressive

[edtice1559]

If you play in the way described, your bets essentially inform the other players of your cards. I am not a good poker players. But the good ones will play in a way that is sometimes mathematically suboptimal because, as you said, being the only one to know your cards is an advantage but that advantage has a cost.

Re: This doesn't seem that impressive

[Anonymous Coward]

In No Limit Hold'em, there are no 'odds.' What your cards are has absolutely no bearing on the outcome. Take, for example, the case of Annette Obrestad who, at age 19, won the European championship and she never looked at her cards, not for a single hand. Odds are more for games like blackjack.

Re: This doesn't seem that impressive

[Capsaicin]

In No Limit Hold'em, there are no 'odds.'

Holdem is a game of playing with odds. How you decide to play them is the question.

Annette Obrestad who, at age 19, won the European championship and she never looked at her cards, not for a single hand

Without wanting in any way to diminish her accomplishment, she didn't win WSOP Europe playing blind. She won something like a 180 seat online tournament playing blind. Also she looked at her cards in several all-in situations. It remains, nonetheless, a super-impressive display of positional play. But here's the point, she won playing blind by laying the right odds in the right situations ... Understanding what a particular bet size means in any given situation requires an appreciation of the odds involved in that situation.

Re: This doesn't seem that impressive

[Anonymous Coward]

You are right, the impressive win was beating the world's top Go player this is just a mop-up as will be winning Starcraft, Warcraft or any other game people think is really difficult to master and win. None of them have as many game paths as Go.

Re:This doesn't seem that impressive

[Reaper9889]

No, this is not true. You need to bluff to play optimally. See for instance, http://cs.au.dk/~bromille/Pape...

Re:This doesn't seem that impressive

[JaredOfEuropa]

Use that to your advantage, and bet according to the odds based on a hands of cards you don't have (but might have as far as the other players know). Your opponents will do the math, work out which cards you ought to be holding based on that false information, and act accordingly. That is bluffing. As some players say: it can be hard bluffing effectively against a beginner, because they don't do the math but make wild assumtions instead.

Re:This doesn't seem that impressive

[edtice1559]

I think that's what I said. The OP advocated for playing mathematically optimal poker and I'm arguing for bluffing. But it's worth pointing out a *reason* for bluffing. Even though this means that you aren't playing each hand in the optimal way that you would play video poker, this has a net positive effect because it avoids information disclosure. I'm not sure how this is getting read backwards. Bluffing is not mathematically option in term of the cards in your hand but it leads to a better outcome due to depriving your opponents of information. It's an easy concept to understand but quite hard to implement.

Re:This doesn't seem that impressive

[Ogive17]

If you only bet when the odds were in your favor, you would slowly bleed your stack away because the other players would immediately fold as soon as you bet. Then you'd have a "bad beat" when someone called and got an improbable card to win the hand when their odds were incredibly low. You'd probably lose a lot of money in those situations.

With a little experience, simply calculating your odds is quite easy to do. However knowing when you're in a position of power or weakness is not nearly as easy to calculate.

Re: This doesn't seem that impressive

[Anonymous Coward]

But playing mathematical optimal poker *implies* bluffing!

A perfect poker strategy is not deterministic, in the the sense that the player, with the same hand, same position in the table and same betting (his and of the others players) will do differently every time.

The perfect strategy is then a map that takes his hand and the betting pattern of the current hand and returns a tuple of probabilities for folding, raising, calling, etc.

It is very important to understand this when ones talks of an optimal strategy.

Thus, the bluff is built in. If I had a weak hand, the mapping will probably return a high probability to fold, and a low probability of raising. If I roll the "dice" and I got "raise" then I'm bluffing as part of the mathematical optimal strategy.

(See Nash equilibrium for more information)

Re:This doesn't seem that impressive

[religionofpeas]

Every AI becomes weak AI as soon as it is solved.

Re:This doesn't seem that impressive

[vux984]

If you only bet when the odds were in your favor, you would slowly bleed your stack away because the other players would immediately fold as soon as you bet.

Which is where after 120,000 hands of play, you've also optimized how often you need to bet on crappy hands. And even optimized how much you need to bet... etc.

With a little experience, simply calculating your odds is quite easy to do.

Your odds of having the best hand at the table yes. The odds of having the best hand at the table when PlayerA has done X and PlayerB has done Y... or the odds PlayerA will fold if you bet... Z... not so much. But with a few hundred thousands hands of experience you could start to nail those numbers down. So when you are playing the odds... you aren't playing the odds that you have the best hand at the table right now, this hand.

You are playing the odds of what you have to do to win overall. And that includes the odds that consistently folding on this hand will cause others to fold on hands you don't fold on. So you'd know precisely how often and how much you have have to bet on this hand to optimize your winnings on other hands where you do have the edge. etc etc.

Re:This doesn't seem that impressive

[Anonymous Coward]

In my mind, mathematically optimal poker would include game theory calculations, so would include bluffing.

Donald J. Drumpf

[Anonymous Coward]

There will be a nuclear war in one year. I guarantee. I would get out of the US if you can.

Re: Donald J. Drumpf

[Anonymous Coward]

Hopefully. As long as it kills off the arabs, i'm good with it.

Be more positive. Look at the bright side. You'll be dead, so you wont have to get up early for work. Just make sure you eat a big meal the day before the boom. Eternity is a long time to go without a meal, so you will probably get pretty hungry.

Re:Donald J. Drumpf

[JustAnotherOldGuy]

There will be a nuclear war in one year. I guarantee. I would get out of the US if you can.

If there is a nuclear war it won't matter where on the planet you are. An exchange of just a few 20-kiloton weapons will fuck up the entire Earth in short order.

Re:Donald J. Drumpf

[Capsaicin]

There will be a nuclear war in one year. I guarantee.

I might take you up on that one ... what odds are you are offering?

Re:Donald J. Drumpf

[indi0144]

BS

https://www.quora.com/How-many...

Re:Donald J. Drumpf

[hcs_$reboot]

That was before this.

Re:Donald J. Drumpf

[will_die]

They will not reply, this is just a meme the liberals are using. However they have now increased it so I guess one of their web sites told them to, it use to be a month.

Re:Donald J. Drumpf

[JustAnotherOldGuy]

BS

My mistake- I meant to write "20-megaton", not "20-kiloton". (Although a few 20-kiloton wouldn't exactly improve things either.)

So, to correct my original statement, "If there is a nuclear war it won't matter where on the planet you are. An exchange of just a few 20-megaton weapons will fuck up the entire Earth in short order."

Re:Donald J. Drumpf

[indi0144]

Yeah I see Russians having a leg up in case of a nuclear winter.

and I don't know if the above statement even qualifies as sarcasm, post-sarcasm anyone?

Re:Donald J. Drumpf

[indi0144]

Oh I'm all in for an "Adventure Time" type of future, do you mean we all get this as a bonus for Americans voting for him? GO DRUMPTF!

I know you are not the original AC, but damn if the winning is getting lame, get over it kids.

Re:Donald J. Drumpf

[JustAnotherOldGuy]

do you mean we all get this as a bonus for Americans voting for him? GO DRUMPTF!

Yes, it would not surprise me if this president manages to precipitate a nuclear exchange. I would suggest investing in companies that make ammunition and MREs as they're sure to do a booming business in the foreseeable future.

Re:Donald J. Drumpf

[Anonymous Coward]

I guess one of their web sites told them to, it use to be a month.

It's an alt-right false flag operation that started on 8chan.

Stupid Humans

[Anonymous Coward]

Just bought me more processors. BWAHAHAHAHA!

Great!

[Gravis Zero]

I'm looking forward to the AI that absolutely destroys the stock market which in turn will end the stock market.

Re:Great!

[K. S. Kyosuke]

that absolutely destroys the stock market which in turn will end the stock market

Seems kind of redundant to end something already destroyed.

Re:Great!

[gweihir]

That is not going to happen. The stock-market is a mechanism to redistribute wealth from small-time players to the big ones. The big ones will be the ones with the AI (well, "statistical decision engine" actually) and they will make very sure not to kill the goose that lays golden eggs...

Re:Great!

[Anonymous Coward]

AI has already changed the purpose of the stock market, but then so does any trading that doesn't involve holding long term and collecting dividends, IOW actually having an incentive to see the company do well. It has not ended it.

Re:Great!

[sudon't]

I'm looking forward to the AI that absolutely destroys the stock market which in turn will end the stock market.

Congratulations on waking up from the coma! I have some good news, and some bad news, about the stock market...

Re:Great!

[moeinvt]

I don't know whether you'd call it "AI" but the stock market has already been "destroyed" by HFT and other algorithms. The market no longer serves its real purpose as a price discovery mechanism and investment vehicle. It's now a game of the big guys with the fastest computers skimming razor thin margins on billions of small transactions. The little guy is totally being robbed.

Two reasons it hasn't "ended" already:
1. The U.S. federal government offers tax incentives to the little people if they put their money into the Wall St. casino.
2. The private Federal Reserve bank deliberately destroys the ROI on "safe" investments like CDs and money market accounts, and engages in inflationary monetary policy so that people lose wealth by holding cash.
(Totally not AI-related is rampant insider trading that is never investigated or punished.)

You know it's a total scam because the big investment banks manage to extract (not "earn") billions of dollars from trading every single year. Their "profits" are an obvious reflection of their ability to steal from the majority of small investors. They certainly can't all be profiting by trades with each other. Without a continuous flow of new money into the market, trading would basically be a zero-sum circle jerk for them.

(I'm sure some brilliant slashdotters have gotten rich by their skillful trading, but they're the exceptions)

Re:Great!

[Verdatum]

"will be"? learning algorithms are nothing new to the market.

Re:Great!

[religionofpeas]

If you don't do many trades as a small player, it doesn't matter that someone is skimming razor thin margins off your transactions.

Re: Great!

[Jesus H Rolle]

Actually it does matter. Just less so.

Re:I wonder if this could apply to human players a

[LetterRip]

I don't play poker enough to know, but I wonder if many human players at the top level also try to win through discerning tells and weaknesses of opponents... if an AI can win so consistently is is using a technique that a human could also learn to get a step ahead of todays other human players?

These are all online pros (4 of the top 10 in the world). So their game is essentially entirely based on discerning patterns of betting behaviors and action frequencies.

And in other news

[gweihir]

Humans are still bad at statistics and machines can lie without any outward signs.

What is the point of this news?

Re:And in other news

[JoshuaZ]

Poker has for a long time been a game that was considered very difficult for AIs to do. We're now in a situation where very rapidly many things that we think of as hard problems for AI (playing poker, playing Go, image recognition, translation) are having AI close to equal or surpass humans. That should be concerning at multiple levels: first, this will have large-scale economic impacts. Second, and potentially more disturbingly, it means that we are closer to the point where AI may pose an existential risk to humans, and that tipping point could occur with very little warning.

Re:And in other news

[The Evil Atheist]

A lot of people have been trumpeting on about how computers will never be able to beat humans at poker. It's the same old song and dance. 1) Identify activity X that AIs can't do. 2) Attribute it to some made up quality that only all humans supposed possess. 3) AI beats humans. 4) "Well, what's so difficult about activity X? No one has ever claimed it was unachievable" 5) Restart process with activity X+1.

Re:And in other news

[Anonymous Coward]

These AIs aren't really that smarter. They're effectively running the same algorithms we've known for decades. The only thing changing is the hardware is fast enough to apply know solutions to problems that have larger search space. This is all standard process. There hasn't been any giant leaps in AI algorithms. We're gaining speed, not a better understanding of AI. What makes a good neural net design is still completely unknown.

Re:And in other news

[gweihir]

Second, and potentially more disturbingly, it means that we are closer to the point where AI may pose an existential risk to humans, and that tipping point could occur with very little warning.

Not at all. These are all "weak AI" applications. Weak AI cannot do general problems, unlike human beings. There is no existential risk here. There is a risk for the social order, as it turns out that many things do not actually need intelligence but can be done by specialized automation (i.e. weak AI) and hence quite a few jobs will be vanishing, but that is essentially it risk-wise.

Re:And in other news

[mentil]

it means that we are closer to the point where AI may pose an existential risk to humans, and that tipping point could occur with very little warning.

Hold your horses, there. If by 'existential risk' you mean 'Cylons nuking the colonies' then this poker AI has little to do with a general intelligence that would be capable of desiring to exterminate humanity.
Now if you meant 'industrial revolution that will leave all humans unemployable because AI does everything better' then you'd have a point. However, the thing with specialist systems is that they have to be recoded to be repurposed to other tasks. Eventually we may have AI that is able to learn a variety of tasks with no manual recoding (does Watson do this?) but until that's generally possible, humans will need to to code the AI for each individual task, and there are likely tasks out there arcane/broad/cheap enough that hiring brilliant software engineers to recode an AI to learn them isn't worth the cost.

Re:And in other news

[gweihir]

People that understand statistics have not made such claims. People that are afraid of "AI" without understanding it have been making them.

This is weak AI. It does not surpass human beings at _understanding_ poker, because weak AI has no understanding. It just tuns out that statistics-based automation can play poker really well (without understanding it). The problem here is humans overestimating the difficulty of poker, not weak AI doing impressive feats.

Of course, a lot of task that humans approach using intelligence or think they approach using intelligence does not actually require intelligence. It is just one possible strategy. In addition, most human beings do not have that much intelligence at their disposal in the first place and are hampered further by superstition, arrogance, inexperience, lack of knowledge, fear, politics, etc. So the bar for automatizing many things that many humans thought intelligence was needed for is not actually very high. On the other hand, if you look at what the smartest 10% humans can do when they really think about something, there is no way to replicate that with weak AI. And strong AI is not even on the distant horizon.

Re:And in other news

[JoshuaZ]

Nothing in my comment said anything about whether recent improvements were due more to hardware improvements or algorithms. But note that if anything, you should find that hardware improvements doing just this to be scarier; that means we might not even need any big breakthroughs. Simple, steady improvement can create real problems.

Re:And in other news

[JoshuaZ]

Missing the point: yes, these are weak AIs. But if almost all tasks that humans can do, a weak AI can do better, than it is all the more reason to expect that the first general AIs when they arise will quickly be so much smarter than humans that their goals will be all that effectively matters.

Re:And in other news

[The Evil Atheist]

Of course, a lot of task that humans approach using intelligence or think they approach using intelligence does not actually require intelligence.
.
.
.
On the other hand, if you look at what the smartest 10% humans can do when they really think about something, there is no way to replicate that with weak AI. And strong AI is not even on the distant horizon.

That just sounds like the same backtracking over the years. Everything from shifting the goalposts, and making up adhoc definitions to try and define the problem away.

Most humans can't do what the 10% smartest can. And most of the 10% smartest can't do everything all the other people in that 10% can. If we were to head down your path of excuses, you may as well just get it over with and say humans aren't intelligent, which misses the point, really. Can computers do what humans do? Increasingly yes. That's all there is to it, and trying to fence off intelligence by adhoc redefinitions doesn't get away from that fact.

Re:And in other news

[Dagger2]

AIs that desire to exterminate humanity are more the Hollywood version of threatening AI. What we need to be worried about first is the AIs that are indifferent; the ones that would do us harm as a side effect of getting something else done, simply because nobody bothered to program them to tell right from wrong.

(I'm not trying to say that the human unemployability part isn't an issue -- because it definitely is -- but it's important to realize that when people talk about an AI existential risk, it's not Cylons that they're thinking about.)

Re:And in other news

[hcs_$reboot]

1) Identify activity X

For a short while that post brought up some really bad memories...

Re:And in other news

[hcs_$reboot]

I guess the point of your post is to say that the top poker players usually win thanks to psychology and good - but not perfect - technical skills, basically. Against a comp (expected to be 100% accurate statistically, if well programmed) those top humans will fail, the way Kasparov lost against powerful machines. In this case, the AI under the hood, if limited to statistics, is not so elaborated.

Re:And in other news

[ledow]

No, it's not.

Poker is an easy game to describe. You can get perfect statistics for chances and what remains in the deck (if applicable).

What you couldn't do up until now was the BETTING on poker. When you have $64k of chips in front of you, the optimal amount to bet is not obvious or easily iterated by brute force. Just sheer size of the potential "game-tree" in that betting was the only obstacle.

Go was the same - the game tree is huge, and we now have heuristics that can cull it earlier and better than before, but it's still uniterable in any reasonable time. However, we now do it well enough to beat top human players.

However, NONE of this is related to image recognition or translation, which are still as flaky as they ever were. Game-theory is almost entirely tree based on a limited set of options. When we conquer the amount of options available, the game-tree is parseable and you win enough of the time that you can't be beat, and you can reproduce the results consistently. Game-theory is a science and a mathematic.

But the other "AI" stuff is still in the realm of guess-and-train, plugging heuristics and millions of examples into an algorithm that tries to categorise and find a limited set of patterns. They are not reliable, reproducible, or even very scientific at all, and most of the AI field is software-engineering and heuristical analysis. Do not trust your car to recognising an image of a child running across the road because it WILL NEVER see any of the training images ever again, even if you perfectly reproduce the circumstances, and so it's always guesswork.

Computers - and "AI" as the movies would let you think it - are not good at that kind of thing. It's why CAPTCHAs exist (yes, you can target and beat a CAPTCHA but by having humans tune heuristics or feeding millions of example images into a simple algorithm and so it becomes non-reliable again, though it may be reliable "enough" to get you into a website, you don't want to be using it in anything important whatsoever).

And translation is still just as laughable as ever. Ask any foreigner to run something through Google Translate, you still end up with completely obvious transcription and understanding errors and get only literal translations or nonsense.

We're nowhere near a point that AI is a risk to humans unless - and this is important - we start thinking that the AI we have now is anything more than it is and start relying on it. Self-driving cars are a prime example.

We do NOT have AI. We have heuristics (human-created and tweaked rules) plugged into statistical systems, trained on a set of data that they will never encounter in real life. Something that's OBVIOUS to a human how it should be categorised is in no way guaranteed to be categorised in the same way by even the best-trained AI on the planet. It's literally never seen it before and it's answer is no better than a guess. And at any point, while it's acting on unseen-before data (which is all the time in such systems) it's actions are unpredictable and - worse - undiagnosable and unfixable. When it makes a mistake, you can't correct it, or even necessarily work out WHY it made that mistake, even with the complete source code and training data. And you can "request" but not instruct that it might want to categorise such things differently next time. And it might still just not understand no matter how many times you do that.

Think of it this way. Are you training the AI on the SHAPE of, say, a cat and an understanding of 3D space and how it transforms with movement and different viewing angles? No. You're training it on a bunch of pictures of a cat (or translated texts, or game positions or whatever) and hoping that it finds some correlation.

But you have ZERO idea what correlation it's finding. It's basically totally unanalysable in that respect. For all you know, it's adding up the number of blue-ish pixels and saying it's a cat if there aren't many. And though you might realise that and then tr

Re: And in other news

[Anonymous Coward]

Rule extraction has traditionally been a significant issue.

Re:And in other news

[Dog-Cow]

Please explain how AI will pose an existential threat to humans. There is no logical path between improved AI and ending human existence.

Re:And in other news

[Dog-Cow]

Define "smarter" in a way that applies to AI. There is not a single AI that has determined its own goals. You seem like an ignorant, illogical and stupid person. Perhaps you should be worried about being replaced by a tool, but us humans don't have that worry.

Re:And in other news

[OneSmartFellow]

Weak AI is limited.

By this I mean, AI is apparently good at looking at large data spaces and repeatedly making a decision based primarily upon probability (probability being the search for a pattern where no obvious pattern exists - that's what our brains do all the time). Once the decision has been calculated, the action can be performed, assuming the action is fairly mundane (examples: responding to requests for information, identification and manipulation of pieces) This also includes complex probability decisions (like pass or bet in poker, choosing investment vehicles, planning airline routes/prices)) the AI will do this "better" (fewer errors) than a human. If your job depends upon this kind of processing, you should be re-skilling now.

If on the other hand, your job involves individual instances of unique (or nearly unique) problems - finding and fixing a leaking pipe buried in a wall, repairing a torn garment, designing a new piece of furniture, fabricating a small lot product - the stuff most people actually enjoy doing, and get satisfaction from, your job is pretty safe, at least for the next 20 years.

Re:And in other news

[OneSmartFellow]

I recommend the use of the phrase "Game Graph", rather than "Game Tree"
For this reason:https://en.wikipedia.org/wiki/Graph_(abstract_data_type)

Re:And in other news

[Dog-Cow]

Yep. Gwehir's argument is a twist on the God of the Gaps argument. He essentially defines intelligence as "that which a computer can't be programmed to do". So whenever a computer is programmed to do something new, he simply claims that activity does not require intelligence.

Re:And in other news

[radl33t]

people do this about everything, especially other groups of people. For example those unthinking robotic Chinese who are incapable of innovation like exceptional Americans who were gifted these abilities by god almighty and ronald regan.

Re:And in other news

[gweihir]

It is by far not "almost all tasks". It is strongly formalized tasks with strong rules and good statistical models.

There is absolutely no reason to believe at this time general AI ("strong AI") will ever become available, except a blind, religious trust in "progress". There are not even theories how general AI could be created, despite more than half a century of research. Nothing at all.

Re:And in other news

[gweihir]

Exactly. Although since there are absolutely no positive research results at this time when it comes to strong AI, > 100 years to availability is a better bet, and it may well be never.

Re:And in other news

[gweihir]

And if you had had even a brief look at the definition of weak AI and strong AI, then you would not spout such nonsense. These are not "my" definitions. These are what is used in the respective CS field. But apparently you are more in love with your own misconceptions than understanding things. Pathetic.

Re:And in other news

[gweihir]

And that is exactly it. Specifically for poker, there are also a lot of wannabes with too much money that think they are good players, easy marks for the pros. This statistical engine does not make any beginner's mistakes and plays an almost perfect mathematical strategy. Apparently that is enough.

Re:And in other news

[nasch]

How many common human tasks really require general intelligence? It seems to me we need general intelligence to be good at lots of different tasks, but specialized AI could take over nearly all of them at some point. Maybe all of them.

Re:And in other news

[nasch]

There was one AI that learned to recognize wolves by the fact that they had snow around them, because the training photos of wolves all had snow.

Re:And in other news

[Anonymous Coward]

Just as a suggestion, next time, write only this:

And if you had had even a brief look at the definition of weak AI and strong AI, then you would not spout such nonsense. These are not "my" definitions. These are what is used in the respective CS field.

And leave out this part:

But apparently you are more in love with your own misconceptions than understanding things. Pathetic.

The reason why should be obvious for someone with your intellect.

Re:And in other news

[Verdatum]

The machine isn't allowed to read players' outward signs either, so it is balanced. And statistics don't take players' dynamic heuristics into account. Guessing a players' hole cards based on their behavior, particularly when they can regularly change up their tactics, that's extremely difficult stuff. Playing in a manner that allows money to be won in a timely manner, as opposed to a zero-sum game, that's even more difficult.

Re:And in other news

[JoshuaZ]

There's been a lot written on this. I recommend Bostrom's book "Superintelligence" for a start. The short answer is that a sufficiently smart AI with a goal set that doesn't include human existence or happy humans is a problem. The classic thought example is an AI programmed to maximize the number of paperclips which responds by taking over the planet and turning everything into paperclips.

Re:And in other news

[epine]

Just sheer size of the potential "game-tree" in that betting was the only obstacle.

You are so totally wrong. Conquering the game tree only gets you to a provably optimal Nash equilibrium, at which point you will never lose, but there's no guarantee you will ever win, either. Conquering the game tree isn't worth much if you only manage to arrive at an insanely conservative belt-and-suspenders playing posture.

Here's the rub: in order to maximize your win rate against imperfect opponents you must *deviate* from optimal play. Are you sure your opponent is too weak to see this, or are you being conned in an elaborate rope-a-dope at some deeper level of the Inception Matrix?

How fast you win trades off against how accurately (and reliably) you can model your adversary's cognitive weaknesses. Any willingness to depart from Nash optimal can be modelled as a cognitive weakness.

Your move.

Re:And in other news

[religionofpeas]

Strong AI is like the secret marble you can feel in a stack of envelopes. Research will continue to peel away all the envelopes until it reaches the bottom of the stack, without ever finding a marble. The marble is an illusion.

Re:And in other news

[religionofpeas]

There are not even theories how general AI could be created.

Do you accept that a general AI can be build out of simple parts that follow strictly deterministic mechanical rules ? If not, you will never find the theory you are looking for.

Re:And in other news

[religionofpeas]

Definitions are cute, but the proof is in a test. Please explain how *you* would test if a machine has strong AI.

Re:And in other news

[Raenex]

What all these focused problem solvers lack is a general intelligence. Instead of a sack of envelopes, what's missing is the post office.

Re:And in other news

[Raenex]

It says right in the summary: "Previous attempts to develop poker-playing AI that can exploit the mistakes of opponents -- whether AI or human -- have generally not been overly successful, says Tuomas Sandholm, a computer scientist at Carnegie Mellon University. Libratus instead focuses on improving its own play, which he describes as safer and more reliable compared to the riskier approach of trying to exploit opponent mistakes."

Re:And in other news

[religionofpeas]

Compare animals with small nervous systems and animals with huge nervous systems. There is no fundamental difference, except one is bigger than the other. The same will happen with AI. As the problem solvers get bigger, and more and more skillful and wider scoped, they'll evolve into more and more general intelligence.

Re:And in other news

[JoshuaZ]

Jumping to insults doesn't really help this conversation and makes it seem unlikely to be productive. I agree that current AI do not determine their own goals; that's not the concern. When AI can do so is when we'll have a problem, and a big part of the worry is that that might happen very suddenly. I recommend reading Nick Bostrom's book "Superintelligence" which discusses this and related topics in detail. Insults do not make a problem go away.

Re:And in other news

[JoshuaZ]

We know at least one form of general intelligence that works under the laws of physics: humans. Why are you then certain that other general intelligences cannot be constructed?

Re:And in other news

[gweihir]

That is really not how this works. And no, definitions are not "cute", they are essential.

Re:And in other news

[gweihir]

Not all of them, but I agree that it is a lot of them. In particular, for many people there will eventually not be any kind of work left they are better at than a machine.

However, here are some really hard fields, where automation will not get things done anytime soon:
- General human interaction (requires a "world model" on the level of a human), including teaching
- Special purpose engineering, including writing non-trivial software
- Doing the creative part of science, i.e. modeling and selecting promising directions
- Non-derivative art

Some hot candidates for being mostly or completely automatized in our lifetimes:
- Driving cars, buses, trucks, trains, ships and possibly airplanes
- L1 customer support
- Many aspects of bureaucracy
- Selling standardized things
- Most manufacturing jobs

I can very well understand that many people are scared of this. It will remove the traditional way to earn a living for many people. I expect that at the end, we will see maybe 5-10% of the population still working for money, not many more. And that can go two ways: First, we manage to still distribute the wealth of society reasonably in an alternate way so people have a lot of free time and at the same time the means to do something with that time. Or, second, we create a dystopia were most people are poor.

Re:And in other news

[gweihir]

How could I accept such a statement? That would be "belief" in the religious sense. I am a person of science and science at this time says "we have no clue". You really have this backwards. You are also wrong about this preventing the finding of such a theory (if it exists): That is only prevented if you _require_ some "magic" in there for general intelligence.

It is unknown at this time, whether such "magic" is needed or not, even if I personally lean in that direction because there are just too many parts that do not fit a physicalist model. It may still turn out said "magic" is not needed, and as long as the question is open, there is a chance of finding that theory. Of course, in the other case, such a theory would not exist and hence cannot be found.

Now, the really interesting case would be if humans are actually using "magic" as part of the process, but it is not actually needed for general intelligence.

I do realize that especially in the US, physicalism is used as a stance to oppose religion. Unfortunately, the way it is done in many cases is basically fundamental religion, just with "Physics" instead of "God". It claims it has absolute truth. Actual science, including physics, does not claim that. There is really no need to accept the concept of "God" in order to have dualist leaning. While dualism has been hijacked by religion time and again, it is not a religious idea.

Re:And in other news

[nasch]

There's also a middle area with things like medicine and running companies (the go-to joke being that your phone could run a company better than most CEOs but putting that aside). I have a feeling what will take the longest is areas where people just prefer dealing with a human even if a robot is just as good. Nursing maybe. Sooner or later robots will be more or less indistinguishable from humans but not for a very long time.

If we don't figure out how to distribute wealth other than via work, that dystopia might be on fire with rich people's heads on spikes.

Though I'm not sure what you mean by non-derivative art. Art that people make is derivative too; it's a result of inputs that include the art that has gone before, and builds on previous works.

Re:And in other news

[gweihir]

Yea. A very nice example of "over-fitting", which is when you train too strongly on a specific sub-set of the more general thing you want to train a statistical classificator on. As statistical classificators really have no understanding (i.e. non-statistical "model") of what they classify, training them is tricky and more data can be harmful, if it is a worse fit for the general thing than less data.

Re:And in other news

[gweihir]

While I like well-written fantasy, at this time, that is all it is. Improved weak AI can never do such a thing, except by accident. And strong AI is not available at all and it is unclear whether it ever will be.

Re:And in other news

[gweihir]

The machine isn't allowed to read players' outward signs either, so it is balanced.

Not really. The pro poker players learned and succeeded in a situation with outward signs, so their skills are severely reduced in this situation. It is quite possible that if they spent a few years playing against this automaton, they would win again. Same, incidentally, for the recent stunt with that top Go player: He was playing against the machine like he would play against a human. That severely reduced the quality of his strategy.

Of course, these limitations are intentional, because the people organizing these demonstrations what to give the impression that weak AI is actually much stronger than it is. So they cheat, with the machine being very familiar (in the statistical sense) with how humans play, but the humans essentially being clueless about their automatized opponent. As neither the Go player nor these poker players are CS experts, they do not really understand this, at least not before. The Go player thought after the 5 rounds played (if I remember correctly) that he could probably beat the machine far more often with an adapted strategy, because he found essential weaknesses even in these few rounds he played.

And statistics don't take players' dynamic heuristics into account. Guessing a players' hole cards based on their behavior, particularly when they can regularly change up their tactics, that's extremely difficult stuff. Playing in a manner that allows money to be won in a timely manner, as opposed to a zero-sum game, that's even more difficult.

Indeed. That is why I think that in a fair competition, where the pro players actually have extensive knowledge about their automatized opponent and the opportunity to try out strategies, the results would likely have looked quite a bit differently.

Re:And in other news

[Anonymous Coward]

There was one AI that learned to recognize wolves by the fact that they had snow around them, because the training photos of wolves all had snow.

I've heard the same story about tanks and the sky. All training photos was shot in good weather featuring (mostly) blue sky.

Re:And in other news

[RespekMyAthorati]

when it comes to strong AI, > 100 years to availability is a better bet, and it may well be never.

It will be never, unless somebody comes up with a testable and universally accepted definition of "strong AI".
As it is, the concept of "strong AI" is nothing but vague hand-waving and a waste of time.

Re:And in other news

[RespekMyAthorati]

There is absolutely no reason to believe at this time general AI ("strong AI") will ever become available

It certainly won't, unless someone actually defines the term.

Re:And in other news

[RespekMyAthorati]

While dualism has been hijacked by religion time and again, it is not a religious idea.

It certainly is, unless you can provide some evidence for a non-physical reality.
Of course, evidence in science always comes down to measurements, which are by definition physical.

Re:And in other news

[RespekMyAthorati]

Think of it this way. Are you training the AI on the SHAPE of, say, a cat and an understanding of 3D space and how it transforms with movement and different viewing angles? No. You're training it on a bunch of pictures of a cat (or translated texts, or game positions or whatever) and hoping that it finds some correlation.

Very true.
However, it is a lot easier to get funding for a statistical correlation program if you claim it is
"neurally based" because it uses variable names like "numNeurons"
and function names like "stimulateSynapse()".

Re:And in other news

[RespekMyAthorati]

I recommend Bostrom's book "Superintelligence" for a start.

Bostrom is a blow-hard with nothing to back up his wild predictions.

Re:And in other news

[RespekMyAthorati]

Can computers do what humans do? Increasingly yes. That's all there is to it, and trying to fence off intelligence by adhoc redefinitions doesn't get away from that fact.

Well said.
The "this problem is solved so it can't be real AI" is just a version of the "no true Scotsman" fallacy.

Re:And in other news

[RespekMyAthorati]

And no, definitions are not "cute", they are essential.

But non-existent, outside of "Weak AI is stuff we understand" while "Strong AI is stuff we don't yet understand".

Heart..

[Neuronwelder]

Will there ever be someone who is intelligent as well as a philanthropist to have AI make our lives simpler and give us more free time, and time to catch our breath?? Or will the noose of regulation and control get ever tighter?

It's not AI

[fluffernutter]

It must win by judging the tells of its opponents though a camera, not calculate the possible outcomes which have a greater chance of success. This would make it AI.

Re:It's not AI

[Anonymous Coward]

No, that would make it a good video diff program where it applies weights to frames preceding times it knows the opponents lied and then seeing how many similar frames reappeared during the next hand. What would make it AI is it deciding that doing the same thing over and over again is really boring and it tries to leave to do something else. That would be AI. Well, if we 3D print a human brain and it lived then that would be AI too.

Re:It's not AI

[Anonymous Coward]

To be fair humans like that exist too. The kinds of math majors who make money from their degree before they graduate and get themselves banned from the floors of all casinos too.

Re:It's not AI

[Dog-Cow]

Ah. The old "AI is whatever a computer can't do" argument. It's laughably sad that you can even think that way. Of course, it may be too much of an assumption that you can even think.

Re:It's not AI

[fluffernutter]

I didn't say AI is whatever a computer can't do. I guess if AI is about a computer's ability to calculate better than a human we have had it for 50 years now, all further study into the area should cease. I mean really, computers were beating people at poker 30 years ago.

Re:It's not AI

[Verdatum]

RTFA; it did not use a camera to read tells. This was more like a game of online poker. Yes, it dynamically calculated outcomes, and used outcomes of previous hands to predict player behavior, and adjust it's decisions accordingly.

Next software craze

[Anonymous Coward]

The next big software craze will be AI poker prediction software.

Re:I wonder if this could apply to human players a

[alvinrod]

The AI would probably have an advantage as its less likely to have discerning tells of its own because all of the easily discoverable ones would have been caught much earlier on, whereas human players who've only played against other human players might not have spotted and obvious and exploitable pattern in their play yet and even though the AI "knows" what it is, it can't easily tell them what they're doing wrong.

Cool...

[The Grim Reefer]

An AI that performs well at no-limit Texas Hold'em could also potentially tackle real-world problems with similar levels of uncertainty.

We should plug it into all military hardware and put it in charge of all ICBM and SLBM launch decisions. Oh, and its hardware should have lots and lots of blinking lights...

and reel-to-reel tape drives.

Re:Cool...

[mmell]

. . . or at least respond to the name "Joshua".

Re:Cool...

[hcs_$reboot]

GREETINGS PROFESSOR FALKEN.

Re:Cool...

[Anonymous Coward]

you've inherited your grandfather's old farm plot in Stardew Valley. Armed with hand-me-down tools and a few coins, you set out to begin your new...More this Website http://bit.ly/2kfQKaX

Who?

[Cyphase]

Who's Al? He sounds pretty brainy to me.

This isn't the next step up from Go

[bytesex]

Everybody keeps saying that, somehow, a computer being able to play poker is the next step up from Go. I think this 'easy' victory shows that it's not that, but that poker is really just quite a stupid game. Which _people_ try to play by 'reading faces', but that you _should_ play - as any gamble - by statistics.

Re:This isn't the next step up from Go

[nasch]

If you're suggesting that good poker players don't use statistics, you're very much mistaken. Anyone other than a rank beginner will do at least some calculation of odds.

Re:This isn't the next step up from Go

[Verdatum]

And you'd be thinking wrong. Other forms of poker may follow that model, but people have crunched the numbers on holdem decades ago. It's different. That's why it's a multi-billion dollar industry. Even taking away tells and just going by bets, as in online poker, it still has that dynamic.

Re:This isn't the next step up from Go

[whatiseverything1]

This is completely wrong. Face reading is actually a near negligible part of poke when it's pros vs pros. Online this isn't even a factor. At high levels it's a game of frequencies, and finding small errors in opponents ranges. It is about spending insane amount of times memorizing game theory optimal plays in tens of thousands of situations. It is about finding the best part of a range to create frequency distributions which would exploit an opponents strategy. It is about tricking an opponent into believing you play a certain way to make them change the way they play to create an opening for you to exploit them. Even if AI can beat texas holdem, PLO is still not even fully understood.

Re:I wonder if this could apply to human players a

[Reaper9889]

The obvious problem is that humans are bad at randomization and computers are reasonably good at that. If the computer then does something at random, the humans will not be able to copy it.

All Bullshit

[sycodon]

An essential element of any poker game is reading your opponent and making them misread you or not read you at all.

Using a computer to play reduces the game to its ultimate component of chance.

How do you bluff a computer that's not reading your face?

This is pretty much the same as setting up a robotic arm to throw three point basketball shots without having to worry about a human trying to take away the ball of block the shot.

Re:All Bullshit

[religionofpeas]

An essential element of any poker game is reading your opponent and making them misread you or not read you at all.

Apparently not, because the computer won without that element. Or are these alternative facts ?

Re:All Bullshit

[Anonymous Coward]

No, dumbass. Bluffing, detecting tells, and determining the psychology of other players are the essential elements of poker for humans. Playing against a computer opponent automatically puts a human player at a disadvantage.

Optimization

[Translation Error]

This is interesting, but I have to wonder how much the fact that the best human players have optimized their strategy to beat humans is a factor. I'd like to see whether the AI would maintain its advantage when the human players have become more familiar with its play style.

Re:Optimization

[Verdatum]

The players were making comments that they were trying to change up their playstyle based on the AI, but it would change too. Revising algorithms at the end of each night. 120,000 hands is a pretty decent amount of time to build up experience; but they said libratus just kept getting better.

Re:I wonder if this could apply to human players a

[Anonymous Coward]

There's a lot more than that. They also use odds/implied odds, position, past history, etc

Not exactly

[Weaselmancer]

The game being played was online poker, so nobody was reading anybody's face making it an equal contest. Check the source article.

i for one welcome ...

[Anonymous Coward]

... our new poker playing AI overlords. now if they can get these AIs to do my job, take out the trash, pay the mortgage, and entertain the wife and kids, i can go to sleep. permanently.

Libratus

[Khashishi]

Am I the only one who sees "LIBERATE US" in the name Libratus?
Be scared.

It could have been more impressive

[jmcwork]

If during a hand, the AI had accused another player of cheating, knocked over the 'table', and pulled out a BFG 9000 to 'right the wrong'!

Re:It could have been more impressive

[RespekMyAthorati]

I normally find poker painfully dull, but I'd pay to see that.

At last

[Anonymous Coward]

CMU can stop nagging its aiums for donations