Could a machine really outguess the world’s best human forecasters?
That’s exactly what happened this summer when a British AI system cracked the top 10 of an elite global prediction contest.
It outperformed dozens of professionals on everything from US wildfire acreage to political bust-ups between Donald Trump and Elon Musk.
ManticAI, co-founded by a former Google DeepMind researcher, landed eighth in the Metaculus Cup.
The Metaculus Cup is a San Francisco-based competition beloved by hedge funds and big corporations for its eerily accurate forecasts.
“It’s certainly a weird feeling to be outdone by several bots,” admitted professional forecaster Ben Shindel, who trailed the AI before clawing his way ahead.
What’s The AI’s Secret Sauce?
Breaking big questions into smaller tasks and farming them out to a fleet of machine-learning models from OpenAI, Google and others.
“You can’t predict the future by regurgitating old data,” co-founder Toby Shevlane told reporters.
“Our system often strongly disagreed with the crowd — that’s how you beat groupthink.”
Experts caution that humans still win on nuanced, data-scarce questions.
But with AI improving fast, even forecasting veterans are hedging their bets.
As human forecaster Lubos Saloky, who placed third, quipped: “I do not plan to retire. If you can’t beat them, merge with them.”
