Deep Blue was one hell of a chess player.
It was February 1996 and the machine developed by IBM was locked in battle with Gary Kasparov. Chess was big news as the computer system project originally begun in 1985 at Carnegie Mellon University attempted to do something other chess-playing devices had been unable to do – beat a reigning world champion.
Even those with only a passing interest in chess like myself were intrigued by the matchup. Deep Blue’s designer said the machine could evaluate 200 million positions per second, and at the time, it was the fastest computer to match up with a world chess champion. Reports on the day’s progress were published in newspapers all across the globe.
Ultimately, the first match of six games was a victory for humanity with Kasparov notching a 4-2 victory. However, in May the following year, and after some additional re-engineering, it was Deep Blue coming out on top.
The Deep Blue phenomenon has been in my head for the last couple weeks as four top poker players (Jason Les, Daniel McAulay, Jimmy Chou and Dong Kim) squared off against artificial intelligence software at the Rivers Casino in Pittsburgh.
This time the AI came out on top.
As Reuters noted, “Libratus [Latin for balance], an AI built by Carnegie Mellon University racked up over $1.7 million worth of chips against four of the top professional poker players in the world in a 20-day marathon poker tournament that ended on Tuesday.”
Headlines have trumpeted Libratus’ accomplishment around the world. Here are just a few examples:
• Machine beats humans for the first time in poker (Reuters)
• Computer manages to beat 4 of world's best poker players (FOX News)
• A Computer Just Clobbered Four Pros At Poker (FiveThirtyEight)
• A Mystery AI Just Crushed the Best Human Players at Poker (Wired magazine)
• Artificial Intelligence Goes All-in …on Texas Hold’em (Wall Street Journal)
Developers compared the victory to that of Deep Blue 20 years ago. The team certainly faced a challenge in engineering their AI to adjust to betting differences, imperfect information, unorthodox play, and that unique aspect of poker that differs it from most other games, bluffing.
Players were given a certain amount of “play money” and Libratus would go on to notch a computer's first victory in the no limit variety of Texas Hold'em (a previous computer had already mastered Limit Hold'em).
“Yes, poker is just a game," University of Michigan professor Michael Wellman, who specialises in game theory and closely follows AI poker, said to Wired magazine. "But the game theory exhibited by Libratus could help with everything from financial trading to political negotiations to auctions.”
Some have hailed the entire spectacle as great for the game of poker – and no doubt – there is some nice PR benefit that comes with it. But from a simple poker-playing perspective and in regards to its relevance among poker fans, the whole thing seems a bit too much. As a massive fan of the game of poker, this whole spectacle lacks the impact of Deep Blue’s win.
To me, this matchup of man versus droid/computer/software/techno-gizmo lacks the one aspect of poker that makes it so unique: risk. It’s the reason that playing poker online for free or playing with your grandmother for matchsticks (or cheerios or whatever) is so lame; there is no risk of losing one’s own money.
Chess is a game with merely risk of losing one individual match itself. The two combatants may have some kind of extrinsic monetary motivation, such as tournament payouts, appearance fees, etc., but there is not an inherent expected loss of one’s own personal earnings.
In poker, players must square off against each other with their (usually) hard-earned money and that risk of one’s own cash is a huge part of poker’s appeal. Financial risk is inherently about losing money, and if you’re not playing with risk in the game, you’re not really playing poker.
“If you’re afraid to lose your money, you can’t play to win,” said Johnny Moss, a Texas poker legend and winner of the first two WSOP Main Events.
That attitude is something inherently flawed in making so much hoopla about Libratus' accomplishment; a machine/software/robot has no real inherent sense of loss or risk.
And when it comes to the art of the bluff, it seems engineering a machine to make these kinds of moves misses the key component of the risk involved in doing this: the pulse-racing feel of having all your chips in on a pot when you know your hand is “squadoosh” as ESPN WSOP analyst Norman Chad likes to put it. A highly-engineered AI topped four poker sharks with no real money on the line.
As a poker fan, this whole event doesn’t even seem like real poker and just left me asking: “So what?” Poker is a game that is extremely dependent on human emotion and temperament.
Artificial intelligence has no fears about losing the mortgage payment in a pot.
Artificial intelligence has no fears about losing the mortgage payment in a pot or being down to that last bit of the poker bankroll and having to look for a real job to build it back.
Another aspect of this matchup with Libratus that is really missing for me, and I think for many poker fans, is that the self-reliant, mano-a-mano, battle of minds that takes place at the poker table. Sure I can concede a machine can get the better of humans in this type of setup, but poker’s appeal for me is seeing players squaring off against each other and matching skills.
A battle against a computer lacks the panache of seeing real-life humans battling it out for their own cash. Libratus may have massive amounts of computing power, but it lacks the humanity that makes poker great and now watchable on television.
Many poker insiders and those with deep roots in the game may forget that, to casual fans, seeing thousands of dollars won and lost on a single game of cards is extremely bizarre yet extremely appealing. That appeal, along with the game’s unique characters and history, is the reason poker has grown into the international game it is today.
Poker is great because the human aspect is so important to excelling; it is not simply a series of moves on a game board or your old Commodore 64. Players who master the game can read other players and keep their own emotions in check.
They must master the subtleties and games within the game to excel. They benefit themselves by timing their actions correctly based on other players’ tendencies, outlooks and general gameplay. Players like Jason Mercier and Daniel Negreanu have mastered these nuances.
Don’t read my hand wrong here, I am not a poker pessimist who thinks the game is moving in the wrong direction. Quite the contrary: I think the game is moving in the right direction in general after massive growth in the 2000s.
Libratus is not the next Big Blue and these four players were not Gary Kasparov.
Actual growth of the game depends on continuing presentations of the game in its real context – on the felt and focusing on the players.
Some of those include: continued growth of the WSOP and live ESPN broadcasts; the World Poker Tour’s continued success and international growth; great broadcasts like Poker Central’s Super High Roller Bowl (with great commentary catering to fans and hard-core players alike); progress (thought slow) of state-by-state legalized online poker; the growth of the game by appealing younger players via Twitch; and the success of middle-tier tours catering to average Joe poker players (which are still needed to grow the game) like the Heartland Poker Tour and Mid-States Poker Tour.
The AI win seems like a minute footnote in comparison. Libratus may have won the battle against mankind, but was there ever really a war? I’m not sure this is a battle that means a whole lot in the big picture of modern poker.
Libratus may have won the battle against mankind, but was there ever really a war?
Libratus is not the next Big Blue and these four players were not Gary Kasparov. It may have been an interesting technological endeavor, but I’m sure these players in the "Brains vs. Artificial Intelligence,” as the event came to be known, would much rather bring home a WSOP gold bracelet or WPT title if they had to pick. That hardware (not software) would be tangible and real and it would certainly be a nice real-life check to cash.
Sean Chaffin is a freelance writer in Crandall, Texas, and writes frequently about gambling and poker. If you have any story ideas, please email him at email@example.com or follow him @PokerTraditions. His poker book is RAISING THE STAKES: True Tales of Gambling, Wagering & Poker Faces and available on amazon.com.
The opinions expressed here are those of the authors and do not necessarily reflect the positions PokerNews