Lectio Libratus

Over the last few days I have received multiple links from poker and non-poker friends on different social media platforms concerning Libratus, the AI that just busted up four human poker players. I’m grateful for being kept in the loop since this is exactly the sort of thing I might overlook while dealing with being on the wrong financial side of the Falcons-Patriots fiasco and trying to be on time for a dinner date.

It’s possible that, due to my age, I’m not great at dealing with such social media bombardments. Specifically, I felt pressured to provide a response to the implication that this pretty much wraps it up for poker, which due to lack of time was invariably both terse and glib. I thus rapidly ran out of responses such as “Let’s see it play PLO,” and “pfffft, NLHE HU!?” and was compelled to think about what, if anything, Libratus’ victory actually means.

And in so doing I came across Robbie Strazynski’s piece In Defense of Humanity.

“Beans,” I murmured. “This is more serious than I thought.”

Strazynski’s engaging piece explains lucidly and in more detail than my Facebook reactions to this alleged calamity a central flaw in the dominant argument; poker is a game with multiple variants that is mostly played many-handed. Thus this contest only probes a tiny and esoteric corner of The Holy Game. But as I read on I realized that I didn’t feel a need to provide a defense of humanity arising from the narrative of “machine-crushes-humankind.”

The people lost.

So what?

Machines are better at the assembly-line manufacture of automobiles. They can lift far heavier hunks of steel, and spot-weld at a pace impossible for people. Stronger, faster, better. And while there are clearly social impacts of this reality that have been compounding since the industrial revolution, I’ve never been on the side of the loom-smashers. I see no point in defending humanity against the charge that it is less good than machines at carrying out monotonous, repetitive tasks that completely fail to engage our greatest asset: our intellect.

But maybe that is the point? Poker requires intellect, thus to lose to a machine is hitting us below the belt, or more accurately above the neck.

Here’s a little exercise for those of you who are comfortable interacting with others. Ask your friends, or if you have no friends your opponents at the poker table, to name a popular game that requires… oh I dunno… “intellect”? “Brain thinking”? A game played by two in which it is really helpful if you are clever. A game that requires concentration, study, knowledge. A game that, in this intellectual sense, requires skill.

Here’s another way of ambling around to the point that I will eventually make. When Norwegian television station NRK wanted to demonstrate what a smart guy world chess champion Magnus Carlsen is, they set up a chess game between him and Bill Gates. The latter is, presumably, by virtue of amassing vast wealth and being clever with computers, the epitome of smart.

Now it turns out that Gates is pretty bad at chess and Carlsen beat him in nine moves, but what is really illustrated here is that there is universal acceptance that being very good at chess is difficult because chess is a game of intellectual skill.

No human being has been able to beat the best chess engines in match-play for two decades.

In 2017 a poker AI beat humans for the first time in a format ideally suited for a machine.

Apart from anything else, kittens, I see the Libratus match as a tremendous opportunity to promote our game in defiance of the dark forces which insist poker is not a game of skill and thus should be subjected to the same neolithic legislation that blights the broader gaming industry.

Poker is a game of skill! It’s really effing hard to play well! It took a dedicated research team at Carnegie freakin’ Mellon to produce a computer that could, on favorable terms, come out ahead!

But there’s another interesting angle advanced by Strazynski that gets into a deeper element of this humanity business and why it might need defending. He writes:

“Bots demonstrate no joy at winning a big pot, or even hubris upon decimating an opponent. Bots don’t cry after getting felted. Bots don’t demonstrate guile, guts, or intestinal fortitude in running a dangerous squeeze play or a risky bluff. There’s no finesse, no creativity, and no passion.”

Ah. The joy of victory. The agony of defeat. Those interminable, cunningly-crafted, emotionally-manipulative ESPN montages accompanied by shit music and people crying or jumping around like gorillas because… “Norm! It’s a JACK!!!”

With the exception of a handful of times when friends made final tables I haven’t watched WSOP coverage since the last time I played the Main Event in 2006. It gives me hives.

When I win a big pot I go to some lengths not to demonstrate joy. It would be rude. And my opponent has to be an A-1 dickhead for me to exhibit anything approaching hubris upon “decimating” him or her, partly because I am a linguistic pedant and taking only one tenth of his or her stack is really nothing to crow about.

There’s another important positive I feel poker players can take away from Libratus’ victory which again draws on the last couple of decades of the history of chess.

While the IBM Deep Blue project and it’s eventual defeat of then world chess champion Garry Kasparov was likely motivated by a desire to boost IBM’s stock, it also heralded a new age for chess. On the PC on which I am writing this article, I have a chess engine that, at classical time controls, is rated considerably higher than Magnus Carlsen. Hundreds of thousands of these engines are installed on computers all over the world. Neither Magnus Carlsen in particular nor humanity in general is worried about this.

Quite the contrary. Easy access to what is, in part, an incredibly powerful analysis tool has seen a renewed interest in chess. It is difficult to tell what specific role chess engines are playing in the expansion of chess, just as in poker it’s hard to separate the Moneymaker effect from the advent of hole cameras and online poker. However, it strikes me as at least plausible that growing sophistication in poker AIs will expand rather than contract interest and participation in our game.

I agree with Strazynski that some silliness has been written about the Libratus victory, but I don’t feel the need to defend humanity, nor do I fret about the consequences for the game of poker. It seems to me the very existence of the Libratus project adds a helpful gravity and seriousness to our game. It highlights that poker is based on the most human of attributes: intellect.

Most of you know the frustration of telling someone you’re a poker player and being confronted with a derisive snort. “Oh right. You’re a gambler. How’s that working out?”

Libratus allows us a new response:

“Back off, man. I’m a scientist.”

~ Kat Martin, February 2017.
Twitter: @TheGameKat