Businessweek: The Golden Era of AI Chess Makes Things Tricky for Players

The chess world was reeling when 13 of America’s best players convened in early October for the US Chess Championship in St. Louis. The previous month, top-ranked Magnus Carlsen, 33, suffered a surprising defeat in the same city at the hands of 19-year-old upstart Hans Niemann. Carlsen accused his opponent of cheating—without providing evidence—while Niemann proclaimed his innocence.

There was little prospect of closure as Niemann, broad-shouldered with a mop of curly hair, settled into the match room in St. Louis. (Carlsen wasn’t competing.) Chess seemed to return to its usual form: a game of profound intensity played in churchlike quiet. Niemann started the two-week-long event strong, stumbled through a string of losses, then pulled it together to finish in the middle of the pack, about what you’d expect from a player at his level. Nothing about his tournament play raised any eyebrows. “Even people who were extremely critical of him are saying that his performance is not really noteworthy,” says chess journalist Greg Keener.

Hours after the last game was played on Oct. 20, and shortly before the first drinks were poured at the awards ceremony, the next bombshell dropped. Niemann had filed a $100 million defamation suit against Carlsen and a number of co-defendants. The game had never seen anything like it. As chess YouTuber Levy Rozman put it, “This is probably the most shocking development in the world of chess ever.”  (Carlsen’s manager didn’t respond to an interview request; Niemann declined to comment.)

Although ostensibly about allegations against a single player, the scandal also touches on a more universal concern: the disruptive impact of artificial intelligence on the very top level of human intellectual expertise. It suggests that chess could provide a warning of what’s ahead for other pursuits once thought to be uniquely human, such as writing and making art, raising philosophical questions about authenticity and fairness. “If we come to a point where we can’t really be sure if we’re dealing with a computer or a person, that’s going to cause a lot of angst, anxiety and frustration,” says Frank Pasquale, a professor at Brooklyn Law School and author of New Laws of Robotics: Defending Human Expertise in the Age of AI.

The tension within the chess world had been building for decades, since computers first began to overtake the best human players. The early chess-playing programs—now commonly called chess engines—required mainframe computers. In 1987 a single game on one leading program cost $80,000 worth of computing time. Today anyone can access them anywhere for free from a phone or laptop.

The software was transformative. It revealed strategies no human had ever thought of before, allowing players to achieve a deeper understanding of the game. Together with the internet, AI democratized the pursuit of excellence: You no longer had to live in New York or Moscow to learn from the masters, because the best players existed entirely in the cloud.

But engines skewed the competitive dynamics of the game in destructive ways, too. Cheating is rife in online chess; in tournaments, competitors must play in view of a streaming camera, but it’s not hard to hide another computer out of view. Even before September, suspicions were circulating about Niemann’s rapid rise. In 2021 the 17-year-old achieved the title of grandmaster—the highest rank awarded by the World Chess Federation—and entered the ranks of the 100 best players in the world.

The rumors remained unsubstantiated as Niemann faced Carlsen on Sept. 4 at the Sinquefield Cup, a prestigious tournament held at the Saint Louis Chess Club. Carlsen had held uncontested dominion over chess for a decade and is considered by some to be the best player ever; Niemann, a last-minute replacement, was the lowest-ranking player in the competition. His victory was surprising but not incredible—upsets do happen, after all. But Carlsen was convinced that something untoward had happened and withdrew from the tournament. The next day, Niemann admitted that he had twice cheated during online competitions, once as a 12-year-old, when a friend gave him help, and again as a 16-year-old, with the assistance of electronic devices. But he said he’d never cheated as an adult and never during in-person play.

No one could come up with a plausible mechanism Niemann might have used to cheat. In the past, cheaters at over-the-board, or OTB, tournaments have been caught consulting their phone during bathroom trips or receiving signals using devices in their shoes. One received suggestions from a co-conspirator in the audience who signaled by moving around the competition hall. But OTB tournaments have begun implementing safeguards, such as removing audiences and electronically scanning players as they arrive, and no one suggested Niemann had gotten around them. Instead, people worried that he may have found some new, as-yet-undetected method.

The next time the two met to play, Carlsen resigned from the game after a single move. The largest online chess platform, Chess.com, rallied to his cause, stating in a report that “Hans likely cheated online much more than his public statements suggest” and that as a result “we uninvited Hans from our upcoming major online event.” (Skeptics pointed out that a month before, the company had made a deal to buy Carlsen’s website, Play Magnus, for $83 million.) At the time, Niemann said, “I’m not going to back down, and I’m going to play my best chess here regardless of the pressure that I’m under.”

The controversy split the chess world. Some said it had been unfair of the most powerful man in chess to make such a serious charge against a junior player without evidence. Others said it was high time that someone spoke up against the broader issue of cheating. “It’s been a problem for a long time,” says Lenier Domínguez, the 12th-ranked player in the world. “Some people say, ‘You need proof.’ But in chess there is no system that can give you your proof.”

In the absence of hard physical evidence such as a transmitter in a shoe, there are two ways to detect cheating. The first is intuition, an expert’s sense that the moves a player is making are not right. “You know how a human mind works. People tend to make decisions a certain way,” Domínguez says. Computers use a different computational process and can reach very different conclusions. But sensing weirdness is something only top players are equipped to do, and it’s hardly proof by itself.

The second approach is to analyze the games algorithmically. Computer scientist Ken Regan, regarded as the world’s foremost chess-cheating expert, has developed a statistical approach that detects patterns of cheating by using chess engines to measure the utility of each move—that is, how useful they are in bringing the player closer to victory. “The idea is to project how well fallible human beings will perceive that utility,” he says. If they find moves much better than a player of their skill level should be able to, he says the odds are good they’re cheating. The method is similar to that used by securities regulators to detect potential insider trading, but it also has its skeptics.

The worst-case scenario, which many fear is the most likely to happen, is that no convincing evidence that Niemann was cheating will emerge, but he still won’t be able to clear his name. If Niemann’s lawsuit is successful, it might redeem his image. But chess engines will continue to improve, and unscrupulous players will continue to seek an unfair advantage. It’s the unknowability of what they might be doing more than the damage of any one deed that is the threat to competitive chess.

Such disruption will probably become more common with the spread of so-called synthetic expertise, says Ron Fulbright, a professor of informatics at the University of South Carolina Upstate. With the emergence of Dall-E, which can produce elaborate illustrations from a handful of cue words, and GPT-3, which can spew out detailed, comprehensible text, artists and writers are confronting the once-shocking possibility that their expertise, too, is automatable. “We’re going to eventually have millions of synthetic experts out there, in almost every domain that humans involve themselves in, whether chess or lawyering or landscaping or plumbing,” Fulbright says. Once that happens, what will mark a real expert? Does it matter if we don’t have them anymore?

At the party that followed the US Chess Championship awards ceremony, an unchesslike sense of festivity prevailed, lubricated by an open bar and the thumping beats of a DJ. Niemann, striding the room in a floral shirt unbuttoned to his clavicle, batted aside an interview request, then sat at one of the chess boards arrayed near the dance floor to play a round of bullet chess. Laughing, frantically moving pieces and slapping the one-minute time clock, scrambling after flying pieces, he seemed less like the focus of an international controversy than a normal teenager at a party. This is one thing, perhaps, that superhuman AI, for all its transformative powers, will never be able to do: play the game for fun.

This article originally ran in Businessweek on November 11, 2022.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.