Maria Konnikova (left), a science journalist who quit a good gig to become a poker player, learned a a good deal about the human side of the game and about AI programmers’ efforts at automating it. Along the way, she won money and wrote a book, The Biggest Bluff: How I Learned to Pay Attention, Master Myself, and Win (June 2020).

In an excerpt at Wired, she reflects on the fact that computer pioneer John von Neumann (1903–1957) was a poker player and “Not just a poker player, but someone for whom poker inspired brilliant insights into human decisionmaking, someone who considered it the ultimate game for approximating the strategic challenges of life.”

Poker is a game of skill but it features a comparatively large element of chance and uncertainty. As Konnikova puts it,

In poker, you can win with the worst hand and you can lose with the best hand. In every other game in a casino—and in games of perfect information like chess and Go—you simply must have the best of it to win. No other way is possible. And that, in a nutshell, is why poker is a skilled endeavor rather than a gambling one.

Indeed, when the economist Ingo Fiedler analyzed hundreds of thousands of hands played on several online poker sites over a six-month period, he found that the actual best hand won, on average, only 12 percent of the time, and less than a third of hands went to showdown (meaning that players were skillful enough to persuade others to let go of their cards prior to the end of the hand).

Maria Konnikova, “Poker and the Psychology of Uncertainty” at Wired (June 23, 2020)

At Nature, Liz Boeree tells us, while reviewing The Biggest Bluff, that she too, as an astrophysics graduate, spent ten years playing poker on the professional circuit. As she sees it,

The game sits in a Goldilocks zone between the crisp, perfect information of chess (no hidden knowledge; best player almost always wins) and the mindless gamble of a roulette wheel. It involves just enough luck and just enough skill to resemble the messiness of reality.

Liv Boeree, “What the world needs now: lessons from a poker player” at Nature

So, given all that, how easy is it to teach an AI to play poker? Konnikova decided to find out from a pro, Carnegie Mellon University computer scientist Tuomas Sandholm, who hopes to use poker to teach AI applications to handle uncertainty:

“Poker is the main benchmark and challenge program for games of imperfect information,” Sandholm told me on a warm spring afternoon in 2018, when we met in his offices in Pittsburgh. The game, it turns out, has become the gold standard for developing artificial intelligence…

The goal isn’t to solve poker, as such, but to create algorithms whose decision making prowess in poker’s world of imperfect information and stochastic situations — situations that are randomly determined and unable to be predicted — can then be applied to other stochastic realms, like the military, business, government, cybersecurity, even health care.

Maria Konnikova, “The Deck Is Not Rigged: Poker and the Limits of AI” at Undark (July 17, 2020)

So far three programs have been designed to compete with human players: Claudicus (which flopped), Libratus (which succeeded) now Pluribus, which can play five players at once. Modules that added to the success were a strategy mimicking regret (“Regret for a computer simply means realizing that an action that wasn’t chosen would have yielded a better outcome than one that was”), a subgame solver and a self-improver. Konnikova comments,

Of course, no algorithm is perfect. When Libratus is playing poker, it’s essentially working in a zero-sum environment. It wins, the opponent loses. The opponent wins, it loses. But while some real-life interactions really are zero-sum — cyber warfare comes to mind — many others are not nearly as straightforward: My win does not necessarily mean your loss. The pie is not fixed, and our interactions may be more positive-sum than not.

What’s more, real-life applications have to contend with something that a poker algorithm does not: the weights that are assigned to different elements of a decision. In poker, this is a simple value-maximizing process. But what is value in the human realm? Sandholm had to contend with this before, when he helped craft the world’s first kidney exchange. Do you want to be more efficient, giving the maximum number of kidneys as quickly as possible — or more fair, which may come at a cost to efficiency? Do you want as many lives as possible saved — or do some take priority at the cost of reaching more? Is there a preference for the length of the wait until a transplant? Do kids get preference? And on and on. It’s essential, Sandholm says, to separate means and the ends. To figure out the ends, a human has to decide what the goal is.

Maria Konnikova, “The Deck Is Not Rigged: Poker and the Limits of AI” at Undark (July 17, 2020)

Like so many AI professionals, Sandholm has a sunny view of what is happening today: he believes that “The world will ultimately become a lot safer with the help of algorithms like Libratus”:

Logic is good, and the AI is much better at strategic reasoning than humans can ever be,” he explained. “It’s taking out irrationality, emotionality. And it’s fairer. If you have an AI on your side, it can lift non-experts to the level of experts. Naïve negotiators will suddenly have a better weapon. We can start to close off the digital divide.”

Maria Konnikova, “The Deck Is Not Rigged: Poker and the Limits of AI” at Undark (July 17, 2020)

The way things are going, a one-word response comes to mind: China.

Konnikova learned several things of interest from her visit to Sandholm’s lab. The main system, Bridges, ran Libratus using over two and a half petabytes (2.5 million gigabytes) but only succeeded in heads-up poker under “limited circumstances”:

Yet despite the breathtaking computing power at its disposal, Libratus is still severely limited. Yes, it beat its opponents where Claudico failed. But the poker professionals weren’t allowed to use many of the tools of their trade, including the opponent analysis software that they depend on in actual online games. And humans tire. Libratus can churn for a two-week marathon, where the human mind falters.

But there’s still much it can’t do: play more opponents, play live, or win every time. There’s more humanity in poker than Libratus has yet conquered. “There’s this belief that it’s all about statistics and correlations. And we actually don’t believe that,” Nystrom explained as we left Bridges behind. “Once in a while correlations are good, but in general, they can also be really misleading.”

Maria Konnikova, “The Deck Is Not Rigged: Poker and the Limits of AI” at Undark (July 17, 2020)

Pluribus did much better but Konnikova asks, can AI make the unpredictable predictable when the deck is not rigged?:

There are more things in life than are yet written in the algorithms. We have no reliable lie detection software — whether in the face, the skin, or the brain. In a recent test of bluffing in poker, computer face recognition failed miserably. We can get at discomfort, but we can’t get at the reasons for that discomfort: lying, fatigue, stress — they all look much the same. And humans, of course, can also mimic stress where none exists, complicating the picture even further.

Pluribus may turn out to be powerful, but von Neumann’s challenge still stands: The true nature of games, the most human of the human, remains to be conquered.

Maria Konnikova, “The Deck Is Not Rigged: Poker and the Limits of AI” at Undark (July 17, 2020)

Kenny Rogers (1938–2020) would certainly have agreed (as the iconic “Gambler”): “Son, I’ve made my life out of readin’ people’s faces, And knowin’ what their cards were by the way they held their eyes”:

The takeaway point seems to be that a powerful enough computer will win at any game that depends on calculating all possible odds, dealing with uncertainty, and making the best decision for a limited, defined goal. But so much of life isn’t like that. The information is often sparse and very imperfect or misleading. More, it is hard to decide, in many cases, what the “best” outcome is. Things change and people move on. Those who hope that an algorithm can come along and solve the basic problem of uncertainty in life are kidding themselves.

You may also enjoy:

Can the AI poker champ improve real-world decisions?
That’s the claim aired at Nature for Pluribus, the new Texas hold ‘em champ. Bradley Center fellows are sceptical. Eric Holloway: “Modern AI is like a whiz student that can pass every test but does so by using a cheat sheet. It can get all the right answers in the restricted domain of the test environment.”