Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI Beats Four Top Poker Players

bbc.co.uk

Post reply on HN
Print view View original

131–140 of 234 posts

Re: AI Beats Four Top Poker Players

#131

Post by afastow »

ndh2 wrote: post #116
afastow wrote: post #111

Sounds like the end of online poker is very near. This version required a supercomputer and took a long time to decide its actions but those type of things tend to be quickly improved given enough motivation. Even if poker sites could somehow perfectly detect automated players(which they can't of course), highly skilled poker is profitable enough that some people would be willing to manually execute the actions thems…

Not really. 1v1 is just a small part of poker. Ring games are a whole different beast.

They are more complex for sure but not fundamentally different. I'm amazed the AI managed to win but now that it's happened it's bound to proliferate quickly because there's so much money on the line.

Re: AI Beats Four Top Poker Players

#132

Post by gnaritas »

brilliantcode wrote: post #8
feral wrote: post #7

Earlier quoted context omitted.

> "AI run hedge fund blows up due to black swan event" Knight Capital, 2012: https://www.bloomberg.com/news/articles/2012-08-02/knight-sh...

How many of these hedge funds running on AI are out there in America? I'd think one would make a lot of money shorting such outfits.

Only if you can time when they're going to fail. It's not enough to be right, you have to be right at the right time.

Re: AI Beats Four Top Poker Players

#133

Post by jzwinck »

brilliantcode wrote: post #112
deepnotderp wrote: post #101

Earlier quoted context omitted.

Okay, here's a million dollars a year budget, who do you hire ? If you've read papers and know about deep reinforcement learning , you'll know who to hire.

I hire the guy that read the papers and know who to hire :) In an orchestra, the conductor is not replaceable but the instrument players are. Similarly, someone who allocates his or other's capital cannot be easily replaced very much like the worker's. Your engineer leaves there are 100 others ready to take the place. Business owner leaves you'd have to find a buyer to keep everyone going. This is different from a CE…

> Your top sales guy makes $9 for every $1 spent on him (+$8) where as your top engineer costs $3 for every $0 he generates (-$3)

In this case you should immediately fire all your engineers, your CFO, and your COO. Because why would you want to employ a team of engineers who are costing you money but bringing you no value?

You should also fire your cleaning staff, who have never closed a deal in their lives, plus the plumber who most days dosen't even come in.

Re: AI Beats Four Top Poker Players

#134

Post by andrewprock »

mormegil wrote: post #109
leeleelee wrote: post #74

Earlier quoted context omitted.

This was my thought too. Poker is quite "solvable" meaning, whenever you're confronted with a decisions -- there is always a "correct" answer which does not have to depend on the other players' behavior or style. And you can find that answer by simulation, or game trees, and other methods. It's also important to keep in mind that the best AI can still lose, and the worst AI can still win (and everything in between).…

That depends on what you mean by "correct". Sure, you could theoretically find Nash equilibrium of poker and by playing the equilibrium strategy, you can ensure you won't lose. But that does not mean this is the best strategy to use at a given table against the given opponents, who (being imperfect humans) almost certainly do not play the equilibrium strategy themselves. And, by playing a proper nonequilibrium strate…

The usual way these games are solved is to create an "abstract" game which is tractable, find the Nash equilibrium, and map state in the real game back to the "abstract" game. In the limit, the solutions for a well designed "abstract" game will converge to that of the real game.

Re: AI Beats Four Top Poker Players

#135

Post by brilliantcode »

xapata wrote: post #120
brilliantcode wrote: post #112

Earlier quoted context omitted.

I hire the guy that read the papers and know who to hire :) In an orchestra, the conductor is not replaceable but the instrument players are. Similarly, someone who allocates his or other's capital cannot be easily replaced very much like the worker's. Your engineer leaves there are 100 others ready to take the place. Business owner leaves you'd have to find a buyer to keep everyone going. This is different from a CE…

Funny, I know more people who are asking about machine learning than people who are experts. So who is in high supply?

Machine learning experts are in higher supply, your anecdotal experience offers little weight. High demand creates a high supply which drives wages to zero.

Re: AI Beats Four Top Poker Players

#136

Post by brilliantcode »

xapata wrote: post #121
brilliantcode wrote: post #79

Earlier quoted context omitted.

probably not going to get any details out of you but worth trying: can you give any pointers to implement such strategies and the minimum capital involved?

A Google search result for "machine learning stock trading": https://www.udacity.com/course/machine-learning-for-trading-...

well to be honest I never bothered googling that exact term but I was more interested in the details around his deployment of smaller and successful neural networks as part of a portfolio.

Re: AI Beats Four Top Poker Players

#137

Post by whyileft »

davidivadavid wrote: post #106
brilliantcode wrote: post #92

Earlier quoted context omitted.

Because the arbitrage opportunity isn't present in the academic papers but at a higher level involving business problems. I say this is an arbitrage opportunity because AI is essentially commoditizing human intelligence. It didn't take much brains for people to realize cars could do away with horses. You don't need to understand how a car works down to the formula. It's enough to view it as a black box and still know…

And yet the people who made fortunes in the auto industry were all car makers? What an odd coincidence. But let's indulge your point of view for the sake of the argument. So you've got a black box AI that does something. You know what inputs it needs and what outputs it produces. How do you arbitrage it? What creates the price discrepancy that allows you to arbitrage at all, given that your incapacity to understand t…

The arbitrage is on the labor. Find someone who can build x for y and then sell it for z. The arbitrage opportunity comes from creating or obtaining a market you control. That market can be everything from a platform(Apple) to a collection of sales channels(Oracle) to a monopoly IP position(Microsoft). You do not need find people because they come to you because you control the market.

> In that case, why not trade any random commodity?

Its much easier to gain control of markets that are new as opposed to established ones.

Re: AI Beats Four Top Poker Players

#138

Post by splonk »

limbicsystem wrote: post #93

Is 6-handed a significantly harder game or is it just (from the computer's point of view) just like playing 5 single games of heads-up?

It's a significantly harder game. Computing a Nash equilibrium in a 2-player game is relatively straightforward in technique (albeit very computationally expensive in the case of poker). There's several factors that make even 3-player games much more complicated than 2-players.

An additional player requires some sort of modeling of the interaction between players, which may not be feasible. For example, Player A may have multiple co-optimal strategies with respect to his own EV, but that affects the EVs for players B and C differently. Player A can choose arbitrarily between these strategies at any frequency, but players B and C can't predict this choice at all.

On top of that, Nash equilibrium requires that each player acts independently in their best interest. Teams can be modeled as a single player if necessary, but shifting alliances over the course of a series of hands can't be easily.

And on the pure computational complexity side, the state space explodes when you can have more than one opponent in a hand simultaneously. Combinations of players in a hand scales as n!, not n.

Re: AI Beats Four Top Poker Players

#139

Post by brilliantcode »

gnaritas wrote: post #132
brilliantcode wrote: post #8

Earlier quoted context omitted.

How many of these hedge funds running on AI are out there in America? I'd think one would make a lot of money shorting such outfits.

Only if you can time when they're going to fail. It's not enough to be right, you have to be right at the right time.

I know with high confidence that these AI run hedge fund outfits reverts to a zero sum mean. Doesn't mean that there won't be one or two that beats the market for sometime as a whataboutism. The timing is a major issue, it's tough to at what year these AI hedge funds fail.

The other capital that blew up from the comments suggest it was more of a human error that led to deployment failures which is exactly why AI hedge funds is such a paradox. At the end of the day, it's humans that's deploying and building the model to be approximately right with high degree of accuracy but in the day trading environment it is a poor model for success-not only do you have to be right your monetary exposure must be right...."but proper money management and you will be fine" said everyone who took part in this zero sum game.

I don't know, prove me wrong, I'm sure those ivy league engineers on wall street aren't getting paid dimes for what they do.

Re: AI Beats Four Top Poker Players

#140

Post by __s »

xapata wrote: post #117
andrewprock wrote: post #56

Earlier quoted context omitted.

I'm surprised that Andrew Ng made this claim. The strategy that was built for Libratus' predecessor did not do sophisticated modeling of the opponents, or use new algorithmic principles. Poker is solved using a very large game tree, just as with the other games. The structure of the tree is modified to support the notion of hidden state, but beyond that it is essentially the same as the other games. The structure for…

> Poker is solved using a very large game tree You mean Libratus' strategy used a very large game tree. That is not the only strategy. Take a look at research from the University of Alberta [0]. Also, I'm not certain Libratus' strategy can be simplified to "very large game tree" as I haven't seen the paper, yet. While finding a Nash equilibrium means no other player can beat you, it doesn't mean you're going to make…

Key example of that Nash point is Rock Paper Scissors. You can't exploit anybody playing 100% random, yet that's the Nash Equilibrium. RPSAI competitions tend to have 100% random players perform quite poorly
Post reply on HN