AI has beaten top poker professionals

[Hell-raiser]

So in artificial intelligence being so good at chess and now poker.
Does this mean that the house can always win?

If that is so then what is the point in playing against them online in a poker match.

Cause you know the outcome will always be the same
They will win 100% of the time and you are just wasting your money in playing an AI opponent.

That's how matters stand these days. Ultimately, in my view, there is nothing wrong with that because we're not competing with, for example, a tower crane or freight train in strength as we don't compete with a calculator in calculations. But this doesn't mean that humans cannot compete with other humans in whatever it might be. Whether some people are going to cheat is another question though. But people have always been cheating anyway.

[1713933684]

1713933684

[1713933684]

1713933684

[1713933684]

1713933684

[1713933684]

1713933684

[Seeker#9]

There comes a time when artificial intelligence will dominates human lives. The AI that used in chess and poker might be the first step in modernizing the gambling world. The gambling machines inside the house will also equipped with AI in the future.

[mOgliE]

Well it's not really surprising is it?
I know that poker has even been solved in texas holdem limit. Which means that an IA has managed to find the perfect way to play the game, no matter what the human does. But it's only on head up (two players) and with limit to the bet amount.

So next one to fall might be limitless poker :p

[marlboroza]

Interesting article and unbeatable proof that texas holdem isn't gambling.
I wonder how this will affect whole online poker industry in future.

But who are these guys? Let that machine meet STEPHEN CHIDWICK or Daniel Negreanu with all the skills these guys have!

They are good because they can read opponents. They can't read AI.

Lol.  Not sure if serious.  Negeaunu turned out to be one of the weaker poker players after a generation of online players started taking the scene during the mid 2000's.  Guys like Cole South, Brian Townsend and I guess Patrik Antonious and a whole lot of young guns.

On side note he took third place as top 10 poker players in 2017.

[DARKANGEL6415]

This should basically not be a surprise to anyone in this day in age. I remember when they had the computer go up against the chess world champion player. I remember how mad i was when i lost $500 because i told myself there was no way a computer would win. Granted this was a few years ago so computers was not as good as they are nowadays.

[icanscript]

This should basically not be a surprise to anyone in this day in age. I remember when they had the computer go up against the chess world champion player. I remember how mad i was when i lost $500 because i told myself there was no way a computer would win. Granted this was a few years ago so computers was not as good as they are nowadays.

Yes, it is very reckless not to believe in new technologies. AI does a better job than man. In computer logic there are no mistakes. Cold calculation wins the human intellect.

[Hell-raiser]

This should basically not be a surprise to anyone in this day in age. I remember when they had the computer go up against the chess world champion player. I remember how mad i was when i lost $500 because i told myself there was no way a computer would win. Granted this was a few years ago so computers was not as good as they are nowadays.

Yes, it is very reckless not to believe in new technologies. AI does a better job than man. In computer logic there are no mistakes. Cold calculation wins the human intellect.

Yes, but AI is not about cold calculations. In fact, it is not about calculations at all because it is about what is called fuzzy logic and machine learning. The former is about dealing with the concept of partial truth, which defies "cold calculations", while the latter is about adapting to a changing environment. These two factors are what makes AI so formidable up to a point of being frightening.

[Don Pedro Dinero]

Bots have been known in online poker for long, but it was easy to exploit their weaknesses using stats programs as Holdem Manager or Poker Traker (legal programs). Anyway, bots are banned, and it is difficult to see one of them in the major sites.

This is different, though, but I’m not surprised. AI will change the world as we know it, and this is just a small part of it. It’s just been tested for HU but as those machines keep learning, I think they will soon beat 6-max or 9-max tables. However, I’m not so sure about other formats, like MTTs, where there is high variance.

[yojodojo21]

In short human themselves produce an AI to defeat something that is undefeatable but in other words it's human itself who defeat the opponents since they are the creator of AI. Nothing change about that game who always win is the poker devs and admins at all. But all Ina all money has a big role to human beings that is why it is circulated into humanity. Great, Cheers!

[arisatox]

In short human themselves produce an AI to defeat something that is undefeatable but in other words it's human itself who defeat the opponents since they are the creator of AI. Nothing change about that game who always win is the poker devs and admins at all. But all Ina all money has a big role to human beings that is why it is circulated into humanity. Great, Cheers!

The big difference here is that people get tired and make mistakes, while computers will always do exactly what they are told. Someone who has a lot of poker skill can program it to do better than they could because of going on tilt, being tired, being rushed, distracted, and other things.

[MadZ]

Perhaps I missed it in the article, but these hands were played online, not live, right?

Also, I think it's pretty interesting that it wasn't until the 30k hand mark that the AI started to become consistently profitable, and it wasn't until 80k hands played that this profitably was statistically greater than 0. I think this suggests that the AI's ability to win isn't from some sort fundamentally better strategy against all opppnents, rather, its improved ability to learn from opponents and exploit them over a large sample size. This bot can't just sit down at a random online table and start crushing, which is a good thing.

[Hell-raiser]

Perhaps I missed it in the article, but these hands were played online, not live, right?

Also, I think it's pretty interesting that it wasn't until the 30k hand mark that the AI started to become consistently profitable, and it wasn't until 80k hands played that this profitably was statistically greater than 0. I think this suggests that the AI's ability to win isn't from some sort fundamentally better strategy against all opppnents, rather, its improved ability to learn from opponents and exploit them over a large sample size. This bot can't just sit down at a random online table and start crushing, which is a good thing.

Yeah, it may well be so for the first time, but after a few months of continuous playing the AI should be able to squash all competition at once, at the first game. It is basically the same with chess, though AI doesn't seem to be used there so much. Chess bots use, for example, opening and endgame databases so they don't need to play every game from a cold start, so to speak. This roughly corresponds to the level of AI "expertise" in playing poker acquired after a number of games.

This is what we humans call experience.

[arisatox]

Perhaps I missed it in the article, but these hands were played online, not live, right?

Also, I think it's pretty interesting that it wasn't until the 30k hand mark that the AI started to become consistently profitable, and it wasn't until 80k hands played that this profitably was statistically greater than 0. I think this suggests that the AI's ability to win isn't from some sort fundamentally better strategy against all opppnents, rather, its improved ability to learn from opponents and exploit them over a large sample size. This bot can't just sit down at a random online table and start crushing, which is a good thing.

Yeah, it may well be so for the first time, but after a few months of continuous playing the AI should be able to squash all competition at once, at the first game. It is basically the same with chess, though AI doesn't seem to be used there so much. Chess bots use, for example, opening and endgame databases so they don't need to play every game from a cold start, so to speak. This roughly corresponds to the level of AI "expertise" in playing poker acquired after a number of games.

This is what we humans call experience.

It can do well as a tight player setup but people bluff and it has to learn how each individual plays before doing well against them. Chess is all open and public, while poker is mostly hidden and is about tells, so they're two different things.

[MadZ]

Perhaps I missed it in the article, but these hands were played online, not live, right?

Also, I think it's pretty interesting that it wasn't until the 30k hand mark that the AI started to become consistently profitable, and it wasn't until 80k hands played that this profitably was statistically greater than 0. I think this suggests that the AI's ability to win isn't from some sort fundamentally better strategy against all opppnents, rather, its improved ability to learn from opponents and exploit them over a large sample size. This bot can't just sit down at a random online table and start crushing, which is a good thing.

Yeah, it may well be so for the first time, but after a few months of continuous playing the AI should be able to squash all competition at once, at the first game. It is basically the same with chess, though AI doesn't seem to be used there so much. Chess bots use, for example, opening and endgame databases so they don't need to play every game from a cold start, so to speak. This roughly corresponds to the level of AI "expertise" in playing poker acquired after a number of games.

This is what we humans call experience.

True, it seems from the article that the AI also improved in performance against other AI over time. One thing I will say is that I'm sure the players' knowledge that they were playing against an AI probably had some influence on their playstyle. Whether that means playing tighter or more aggressively, I'm not sure, but I know I'd play differently against an AI than a human opponent (probably more aggressively, since you have much less fold equity against bad players who will call with marginal hands, and I doubt an AI would do this). Regardless, I wonder if the results would be the same if players were unaware their opponent wasn't human.

[BTCevo]

Perhaps I missed it in the article, but these hands were played online, not live, right?

Also, I think it's pretty interesting that it wasn't until the 30k hand mark that the AI started to become consistently profitable, and it wasn't until 80k hands played that this profitably was statistically greater than 0. I think this suggests that the AI's ability to win isn't from some sort fundamentally better strategy against all opppnents, rather, its improved ability to learn from opponents and exploit them over a large sample size. This bot can't just sit down at a random online table and start crushing, which is a good thing.

Yeah, it may well be so for the first time, but after a few months of continuous playing the AI should be able to squash all competition at once, at the first game. It is basically the same with chess, though AI doesn't seem to be used there so much. Chess bots use, for example, opening and endgame databases so they don't need to play every game from a cold start, so to speak. This roughly corresponds to the level of AI "expertise" in playing poker acquired after a number of games.

This is what we humans call experience.

True, it seems from the article that the AI also improved in performance against other AI over time. One thing I will say is that I'm sure the players' knowledge that they were playing against an AI probably had some influence on their playstyle. Whether that means playing tighter or more aggressively, I'm not sure, but I know I'd play differently against an AI than a human opponent (probably more aggressively, since you have much less fold equity against bad players who will call with marginal hands, and I doubt an AI would do this). Regardless, I wonder if the results would be the same if players were unaware their opponent wasn't human.

As long as you can predict move on AI, I do not think that it will be difficult thing to do. But the problem here is how are we going to predict it if many inputs for the AI including face detector, how to react if A using this kind of face, what move should the AI takes. I do not think that it is really easy to know that it is AI or not if the bot is pro like libratus

[Hell-raiser]

Perhaps I missed it in the article, but these hands were played online, not live, right?

Also, I think it's pretty interesting that it wasn't until the 30k hand mark that the AI started to become consistently profitable, and it wasn't until 80k hands played that this profitably was statistically greater than 0. I think this suggests that the AI's ability to win isn't from some sort fundamentally better strategy against all opppnents, rather, its improved ability to learn from opponents and exploit them over a large sample size. This bot can't just sit down at a random online table and start crushing, which is a good thing.

Yeah, it may well be so for the first time, but after a few months of continuous playing the AI should be able to squash all competition at once, at the first game. It is basically the same with chess, though AI doesn't seem to be used there so much. Chess bots use, for example, opening and endgame databases so they don't need to play every game from a cold start, so to speak. This roughly corresponds to the level of AI "expertise" in playing poker acquired after a number of games.

This is what we humans call experience.

It can do well as a tight player setup but people bluff and it has to learn how each individual plays before doing well against them. Chess is all open and public, while poker is mostly hidden and is about tells, so they're two different things.

It may be so in a face-to-face game, where physical presence is required, but poker is played mostly online these days, as many other games, for that matter. So how are you going to bluff in such conditions? Further, it actually cuts both ways. Just as you can say that players can get advantage by reading their opponents' facial expressions, there is no "face" to read in AI. While the AI can still do that.

Are you sure you will be able to deceive it?

[aslamnur28]

Yeah , AI slowly beats people in everything . Will they conquer the world too ?

[Cryptchain]

Yeah , AI slowly beats people in everything . Will they conquer the world too ?

The online world, yeah. Maybe then, we'll go back to a simple life haha!

[Cryptchain]

As long as you can predict move on AI, I do not think that it will be difficult thing to do. But the problem here is how are we going to predict it if many inputs for the AI including face detector, how to react if A using this kind of face, what move should the AI takes. I do not think that it is really easy to know that it is AI or not if the bot is pro like libratus

I think tells are overrated on poker. IA don't care about tells. IA care about position, stack, ranges, odds, blockers, sizings, etc.

[Hell-raiser]

Yeah , AI slowly beats people in everything . Will they conquer the world too ?

The online world, yeah. Maybe then, we'll go back to a simple life haha!

AI will undoubtedly follow you no matter where you go or what you are going to do until it gets you. Yeah, it may take some time until it reads looks and uses its own visual tricks and facial expressions (don't ask me what it would look like), but we will get there eventually, or rather, it will get there. In fact, it may require a lot of time and major break-throughs in the field but we will be there even if we have to build a copycat of entire human brain. Or more than that.