Last Friday, Ripstone Games ran a promotion for their upcoming cross-platform title Pure Chess, inviting chess grandmaster Simon Williams to face off against Twitch chat. Viewers would suggest a move during a designated “voting period” and the most popular move would be played. Ripstone was billing it as a rematch of the 1999 game “Kasparov versus the World,” in which more than 50,000 people on a chess message board failed to vote a set of moves that could beat the reigning Chess champ.
Chess as a novelty isn’t particularly common in 2016.
Chess as a novelty isn’t particularly common in 2016, but that definitely hasn’t always been true. As mechanical craftsmanship in Europe reached its peak in the eighteenth century, there were a number of attempts to use those developments to capture the public imagination. While the average citizen might not have been blown away by inventor John Harrison’s accurate marine chronometer, the chess-playing machine constructed by Wolfgang von Kempelen probably had a better chance of snapping up headlines. But The Chess Turk was not in fact a robotic man that played perfect chess—it was a very fancy box. One of several grandmasters would sit in the box and move the Turk’s arms in order to beat Napoleon or any other human player.
The name of Amazon’s Mechanical Turk service is a nod to this machine. Their Mechanical Turk is a framework that allows developers to offload certain complex tasks to a network of humans. Machine learning continues to develop and improve, but, for now, it’s still cheaper and faster to use human labor to identify images or name colors in situations where computers might struggle. When it turned out that Chess was beyond an automaton, though, the solution was simply to make human work appear to be that of an automaton. Amazon’s service applies the same logic in reverse: simple tasks computers can’t do are performed by a different sort of distributed cloud, and the results look the same because either way the processing is done off-site. By now, we’re long past computers that can beat humans at Chess, and if the match between AlphaGo and Lee Sedol this spring is anything to go by, the same is true of Go, the supposed last bastion of human intelligence.
Twitch chat is a mob on the internet.
As a novelty though, why not try putting a chess pro up against the distributed cognitive power of Twitch chat? Even though the message board in 1999 couldn’t take Kasparov, more recent scholarship suggests that crowds are often more intelligent than even their smartest members. Looking at Twitch chat through the lens of “crowd psychology” tends to turn into an examination of harassment or memes-for-memes’-sake. In James Surowiecki’s The Wisdom of Crowds, he lists the conditions required to form a wise crowd, among them “independence” and “diversity of opinion” (whether either can be found on Twitch chat is an open question).
In other words, Twitch chat is a mob on the internet. This is true enough, but playing Chess against Twitch seems to make good on Amazon’s joke—a distributed human cloud suggests moves, and all 600 of them somehow overcome a real chess champion. Over the course of the three games—the third of which Twitch managed to win—Williams described the competition as “tense” and “unbelievable.” “Where did you get these people on Twitch chat!” he asked. In chat, the popularity of certain moves would ebb and flow before voting closed, often leading to a situation where one person suggested a particularly strong move that was recognized and repeated extensively: one such move was the final E5 to D6, which host Mike Rose noted received as many as 200 votes. As the third game ended, the RipstoneGames folks running chat asked, “Are you all cheating?” The responses came in: “we just rock ;D” “our hive mind is that of a gm” “WE DID IT POGCHAMP.”