A.I. NOW SMART ENOUGH TO AVOID SONOMA LEMONS
Radical leap in machine learning lets artificial intelligence enter crap race, think better of it, then totally flake–completely without human guidance.
After starting with no knowledge of the games but their rules, the original AlphaZero became the world’s best chess, Go, and shogi player in a matter of hours. Researchers were attempting to add Grand Theft Auto V to that list when a Sonoma team signup mysteriously appeared in a folder marked [snickerdoodle recipes–nothing to see here] deep in the system’s core memory.
“By the time we caught the anomaly,” said research head Dr. Daewoo Maepsy-Na, “AlphaZeRomeo had entered a Lemons race, bid on three non-running 200SX Sportbaks, and gotten Mike Carr and Dave Montoya to sign up as drivers.” After cashing Mr. Carr’s and Mr. Montoya’s checks, the program immediately withdrew its team, closed its own email account, and stopped taking its ex-teammates’ phone calls. “Which, I’d point out, is a much smarter strategy than most Lemons captains employ,” observed Dr. Maepsy-Na.
Fears that AlphaZeRomeo will dominate Sears Pointless @ Hooptiecon, Confusingly Presented By Yokohama just as its predecessor dominated professional chess and Go are probably overblown; statistical models have found zero correlation between strategy, intelligence, and success in a Lemons event. There’s also the obvious question of how a virtual entity would meet Lemons’ various physical requirements such as sleeping through the race start, losing the car’s tech sticker, and mounting the transponder nine inches up one’s own colon.
“Nevertheless,” series organizers warned, “while the difference in driving skill between a disembodied machine-learning algorithm and a typical Lemons entrant is negligible, AlphaZeRomeo‘s likelihood of correctly registering for Sonoma by this Saturday’s deadline is orders of magnitude greater than the average team captain’s. So, meat-puppets beware.”