Bots on the Rise: Multiple AIs Now Outperforming Professional Forecasters

The recent Metaculus Cup was not just a victory for one AI, but a coming-out party for AI forecasters as a group. While British startup ManticAI led the charge with an eighth-place finish, its success was part of a broader trend that saw several bots outperform human professionals, signaling a new and highly competitive era in the field of prediction.
This marks a dramatic shift from just a year ago. As professional forecaster Ben Shindel noted, the top bot in a similar contest last year was ranked around 300th. This year, the presence of multiple AIs in the top echelons has left many, including Shindel, with the “weird feeling” of being outdone by their silicon counterparts.
The competition’s 60 questions, which covered a wide range of global events, provided a level playing field for man and machine. The success of multiple bots indicates that the underlying technology—primarily large language models—has reached a new level of maturity. These systems are now capable of the kind of sophisticated reasoning required for real-world forecasting.
ManticAI’s system, for example, uses a collaborative approach, assigning different AI models to specific analytical tasks. This multi-agent method is likely a common thread among the top-performing bots, as it allows for a more robust and comprehensive analysis than a single model could achieve alone.
While the very pinnacle of forecasting is still occupied by elite human “superforecasters,” the rise of a competitive class of AI bots is a game-changer. It suggests that high-quality forecasting may soon become more accessible and automated. The future of the industry will likely see human experts increasingly working alongside, and in competition with, these rapidly improving AI systems.