Over the past week, seemingly everyone I know has been sending me article after article about the Brains vs. AI challenge, in which four top online poker players – Dong Kim, Jason Les, Jimmy Chou, and Daniel McAulay – lost in a competition with Libratus, an AI program for playing Heads-Up No-Limit Texas Hold’em. Whether by email, WhatsApp, Facebook Messenger, or in person, dozens of people who know next to nothing about poker and who have never played the game in their lives apparently saw the headline grabbing titles, usually variations of “Machine Beats Humans At Poker”, and (I suppose) wanted to make sure “their poker buddy” was aware of it.
— PSC (@psc_live) January 31, 2017
I imagine that numerous professional poker players as well as those who work in the poker media find themselves in the same boat. All of a sudden, people from outside of the poker world have taken an interest in our game… only in a way that doesn’t exactly flatter us.
Perhaps Global Poker Index head honcho Alex Dreyfus explained the situation best:
The biggest story about poker in the mainstream media is about how Artificial Intelligence beats best online players. #houstonwehaveaproblem
— Alexandre Dreyfus (@alex_dreyfus) February 2, 2017
After a week of hearing how “poker’s best and brightest got trounced”, quite frankly I felt compelled to pen this op-ed in defense of humanity.
Not Justifying an Inferior Performance, But…
To be sure, Libratus did of course outperform the human players; they’re the first ones to admit their defeat. Still, it’s important to note what exactly the Brains vs. AI poker competition consisted of:
- Over the course of 20 days, each player played 30,000 hands of Texas Hold’em heads-up against Libratus.
- Players competed for 11 hours a day – approximately 11am-10pm – with no days off. Each player needed to complete 1,500 hands per day before they were allowed to quit for the night.
- The number of chips each player held was reset after every hand.
In other words, sure it was a poker competition, but that’s not exactly a full picture of what “poker” is. Moreover, the aforementioned rules do not represent the only way the game of poker can be played; far from it, in fact.
By contrast, for example, in the legendary tale of John Henry vs. the Steam Drill, there’s only one way to drive home a steel rail. In other words, the winner in a “man vs. machine” competition like that is pretty clear.
First and foremost, “poker” is not limited to No-Limit Texas Hold’em, nor is it limited to head-up play against one opponent. Sure, NL Hold’em is by far the most popular poker discipline, but there’s a whole litany of other flop, draw, and stud games, such as Omaha, 2-7 Triple Draw, 7-Card Stud hi-lo, etc. that also comprise “poker”.
Moreover, when you picture a poker table, you’ll usually envision MANY players seated; not just two. That’s because poker is a game in which a hand can be contested by anywhere from 2–10 people simultaneously.
There’s a good reason why the $50,000 Poker Players Championship event at the WSOP features a mix of games and is played at full tables, single elimination-style down to a winner. The winner of that bracelet is regarded as the best all-around poker player in the world.
In short, no AI program could possibly be considered “the best” at poker unless it masters all poker disciplines against every possible number of opponents (i.e., heads up, short-handed, and full ring game).
AI Doesn’t Get Worn Out, Ever
When attempting to properly assess human performance versus an AI in a competition, there are also the factors of physical and mental fatigue, as well as emotion, to consider. In this respect, we’ve already known for a very long time that machines can outperform people. Much like a refrigerator, a washing machine, or a car, AI online poker software will never get “tired” after a few days.
AI also doesn’t get “frustrated”; i.e., players could never put Libratus on tilt. Yet, it’s certainly possible for the opposite to happen, and there’s even evidence to support it. According to a Bloomberg.com report, gameplay “moved at a deliberate pace, slow enough to drive Jason Les, one of its human opponents, a bit mad.” You can’t drive a computer mad.
Also, plainly put, it’s impossible for a human being to stay laser focused on his/her A-game for that long. In other words, it goes without saying that the longer the match, the more favored the AI would be to win.
Imagine the mindset of one of the players about halfway through the competition, seeing that they’re down over $700,000. You also know with 100% certainty that there’s no such thing as “momentum” or a “lucky streak” to be had against a bot. So, no matter how confident you are in your skills or your ability to resist tilt, you can’t help but be affected going forward. Indeed, the second half of the competition saw the players lose a combined $1 million (as opposed to “just” $700,000) more.
Libratus will always play its “A-game”, without getting fatigued or emotional, no matter what.
Is AI Success in Poker a Bad Thing?
It’s likely that some in the poker community – Kim, Les, Chou, and McAulay in particular – might be identifying with how chess grandmasters and Jeopardy! champions felt when they were defeated by Deep Blue (1997) and Watson (2011), respectively. Though their smiles might mask it in the pictures, surely they’ve got to feel at least somewhat intimidated now that a computer exists that can best them at something they’ve dedicated their lives to.
— CMU Computer Science (@SCSatCMU) January 31, 2017
With AI having ousted humans at numerous other skill games, it’s worthwhile to understand why the Carnegie Mellon University researchers felt that poker is perhaps the “last frontier” that AI needed to cross. Here’s a very interesting 2.5-minute video that they put together to explain.
So, yes, the advancement of artificial intelligence is important and helps advance the world in other fields and situations that deal in “imperfect information”, including cybersecurity, auctions, bargaining/negotiation, military strategy, and medical diagnoses.
With all of that said, Libratus’ success and the guarantee of further AI advancement in poker in the future should still not deter people from continuing to play and be interested in the game.
There’s Nothing Interesting about Watching Bots Play Poker
So much of what makes poker enjoyable to watch is the thrill of witnessing amazing stories unfold as well as the full gamut of human emotion on display.
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.
I can’t help but think back to Nike’s famous “Last Game” video from a couple years back, which featured animated versions of the world’s top soccer stars. The parallels are unmistakable.
When online poker players started making the transition into live poker a number of years back, so many poker fans decried the arrival of the now-stereotyped “hoodie-wearing button mashers”. To this very day, when it comes to watching poker on TV or in a live stream, fans of the game still overwhelmingly want to see people who exhibit character and personality grace their screens, not “earbud-clad button clickers”.
The human touch, frankly speaking, is what makes poker fun. It’s that spark of life, that indomitable human spirit, which captures our imagination and enables us to fall in love with poker over and over again.
How poignant, perhaps, that the most engaging thing about the Libratus AI itself is the parody Twitter account that someone set up for it. Throw in the “human” factor, and all of a sudden even a machine can “come to life” and becomes fun and interesting to follow.
Singling Out Doug Polk for Recognition
Perhaps nobody understands all of this better than Doug Polk. As someone who earned his stripes (and a ton of money) coming up in the online poker world, Polk’s successful transition into the realm of live poker has been remarkable. Sure, other online poker players have done so as well, but few have also achieved Polk’s level of fame and notoriety.
Why have people started paying so much attention to Doug Polk? Plain and simple: he got in front of the camera and started creating fantastic poker content, all the while letting his upbeat, cheery personality shine forth. He’s unquestionably one of the most fun people in poker to watch these days, and his fan base is sure to keep on growing with time.
Leave it to Polk, who partook in the first Brains vs. AI challenge against Claudico a couple years back, to turn the latest event into a must-watch, with a lighthearted video like this one:
The moment we start caring more about the poker accomplishments of machines and bots than flesh and blood players, all hope for our game is lost.
So, next time your friend approaches you and mentions that machines beat humans at poker, don’t sigh in frustration. Instead, challenge them to some heads up play. Invite them to your next home game or to accompany you to the casino next time you want to go and play a session. Or, maybe just sit back and watch one of Doug Polk’s great videos together.
I’ll bet that if enough people do that, poker will continue stirring up memorable feelings and emotions in us forever. THAT’s something no AI will ever be able to do.