Big thanks to you, Lex, for bringing some of the smartest researchers and practitioners to the table and sharing these great interviews with the world.
I share the opinion that applying game theoretical tools to computing applications is challenging. Some 15-20 years ago there was a game-theory hype and some interesting results appeared. Around that time, I was working on those kinds of models. In my experience, applying game theoretical solutions concepts necessitates some kind of overhead work. For instance, in my case I applied Nash bargaining theory for multi robot coordination. The robots had to formulate their private utility profiles at the coordination point, share the profiles, and calculate the full game equilibria (with the help of the other robot). Since the game in the general case had multiple equilibria, the initiating robot also shared a random number with the other robot and both used the number to choose the same joint solution, and then finally implement the individual part.
19 minutes of watching and this turn out to be most interested conversation i had listen so far.. he makes a point here.. other then deep reinforcement learning "to be precise learning only' there are non learning methods .. if we can somehow introduce learning in such methods then that will be something new
Why wasn't each poker pro given an HUD to real-time track Libratus' betting stats? This is something all online pros use (and such pros would never agree to play high stakes online without one). Libratus certainly had access to the equivalent of an HUD, right? I presume this created significant imbalance between Libratus and the humans.
They had access to hand history - are you sure that they were prohibited from using HUDs? Do you happen to know how many hours a day they played each day - it would also be useful to know if the competing players were made aware of their collective standings in the competition whilst it was on & indeed if the players could interact daily? I'm also interested to know by what margin the AI won - If the players were aware of their positions, this may well have factored into their in-game decisions. Shame there isn't a paper linked.
The computer is playing game theory optimal poker. It's the equivalent of a bot in rock-paper-scissors that's perfectly random each turn. It won't win by the biggest margin, but it will win.
I'd love to know more about this contest. For example, did the computer know after the fact how the player played certain hands? In a real heads up tournament you may never see what cards your opponent had. I suspect the computer was given certain information to help it establish trends and other information about how the players played. Top players are always mixing up their play to avoid having anyone get a take on how they react in certain situations. Bluffing plays an enormous role in poker and most of the the time if you bluff, your opponent will never know it and vice versa.
Yes. Even if we can't calculate it analytically, we can do it very well for tens of bodies in practice. I am not thrilled by this poker playing with only one other player. The benchmark of playing with 11 others is still there. Much, much more difficult one. I would bet on some "Deep Zero" NN system.
Can anyone knowledgeable about ai and autonomy recommend a reading list for the average layman to gain better insight on such systems. Thanks for this video lex.
watch a video on how on a hand writing number recognition system works. It's the hello world of neural networks and deep learning. But this is not the method used for incomplete information games.
As a pokerplayer i find this conversation really interesting. Not much comes out of the highstakeworld online but i can gve some information. In both )texas holdem and pot limit omaha the topplayers use alot of tools to study and play GTO. However its not humanly possible to play GTO you can spend 100s of hours for spots that if you play professionally never will occur during your lifetime. Therefore topplayers focus hard to study the low hanging fruit commonly occuring situtaions. As i said all toppplayers study GTO but the player i think is the best is Berri Sweet whos playing superexploitive. For sure he knows and studies GTO but his playstyle is exploit
It seems like people have a hard time predicting things with exponential growth. So humans will probably systematically underestimate the performance of AI for a while yet. At some point it will switch and people will assume AI to be better at everything.
Yes libratus was created to play in Heads Up NLHE. Heads up means two players. The game tree for multiplayer NLHE is even bigger and much harder to solve with these GTO approaches.
One of the best interviews. Bring the old Lex back, the podcasts of the last 1-2 years are hard to listen to, with Lex really out of it. Perhaps too much of the high life.
Games that go on forever ... Pi calculation? I know it's not the same and I haven't looked up if this has been done but it just occurred to me: I wonder how well a learning algorithm would do at predicting Pi digits.
Probably not very well. The digits of pi are Borel normal -- statistically indistinguishable from total randomness -- while most learning methods are statistical in nature.
This content is a tour de force. I read a comparable book that left an indelible mark. "Game Theory and the Pursuit of Algorithmic Fairness" by Jack Frostwell
It would be interesting to see AI secretly "tell" a person how to play poker against another human. I wonder if there would be any diffrence. Or if people thaught that they are playing another human being.
This professor with the breakthrough that he made in AI deserved a better interview platform than this... PR game is weak, maybe AI can help you with that Professor LOL????
