The world of forecasting is hosting the new chess match of the century: a quiet, intellectual battle between the best human “superforecasters” and the most advanced predictive AIs. A recent tournament, the Metaculus Cup, was the latest round in this ongoing match, and while the AI played a brilliant game, the human grandmasters are still ahead—for now.
The AI, a system named ManticAI from a British startup, made a series of powerful moves, securing an impressive eighth-place finish. It demonstrated classic machine strengths: massive computational power, flawless memory of historical data, and a relentless, logical approach. Its performance was a quantum leap from previous years, showing it has been studying the game intensely.
However, the world’s top human players, disciples of the “superforecasting” methods pioneered by Philip Tetlock, ultimately won the match. Their advantage lies in a different style of play. They excel at strategy, intuition, and understanding the “psychology” of the game board—the messy, often irrational world of human events. They can spot subtle logical traps that the AI still falls for.
This isn’t a simple game, however. The AI is a learning machine, and its capabilities are improving at an exponential rate. The CEO of the tournament’s organizing body predicts the machine could be the world champion by 2029.
The most interesting phase of the match is yet to come. It will likely evolve from a human-vs-machine setup to one resembling “centaur chess,” where human players are allowed to use AI partners. These human-AI teams, combining human strategy with machine calculation, are expected to outperform either player alone, signaling a new, collaborative future for the sport.