AI Is Learning to Predict Future, Beating Human Forecasting

An AI model developed by UK startup ManticAI has achieved a top-ten finish in the Metaculus Summer Cup, an international forecasting tournament. The model, which placed eighth among more than 500 competitors, demonstrated a remarkable ability to predict geopolitical and social events, outperforming numerous human experts.

The performance of ManticAI significantly exceeded the expectations of the competition’s organizers, who anticipated a top bot would only reach around 40% of the accuracy of the best human forecasters. ManticAI, however, achieved over 80%.

A Multi-AI Approach to Prediction

The AI’s success is attributed to its multi-AI approach. Rather than relying on a single system, ManticAI divides each forecasting problem into tasks for multiple AIs from different providers, such as OpenAI, Google, and DeepSeek. It then synthesizes their analyses, which incorporate real-time news, historical data, and scenario simulations, to generate its final prediction.

Co-founder Toby Shevlane noted that their AI often breaks away from “groupthink,” offering novel perspectives that differ from the human consensus.

A Milestone for AI Forecasting

The achievement is seen as a significant milestone in the field of AI forecasting, which has long been dominated by human experts. The success of AI in accurately predicting events, from election outcomes to diplomatic conflicts, bolsters trust in probabilistic forecasting and demonstrates the growing maturity of AI models.

“If you can’t beat them, merge with them,” remarked Lubos Saloky, a human forecaster who placed third in the Metaculus Cup, acknowledging the increasing capability of AI.

Combining Human and Artificial Intelligence

Still, the debate continues over whether AI can fully replicate human judgment, particularly for events with nuanced social, moral, or contextual factors. Skeptics point to the limited number of questions in the tournament and argue that top human forecasters still hold an edge.

Warren Hatch, CEO of forecasting company Good Judgment, suggests that a combination of human and AI expertise may be the optimal path forward.

“The main point for us is that the answer isn’t human or AI, but instead human and AI to get the best forecast possible as quickly as possible,” he stated.

Implications for Future Decision-Making

For institutions and policymakers, ManticAI’s performance signals a future where AI tools could become an invaluable asset for anticipating major events and improving decision-making in unpredictable domains.

The rapid advancement of AI forecasting, as demonstrated by ManticAI, shows that the next frontier is not just in computation, but in foresight, with broad implications for governments, businesses, and health organizations. As AI continues to close the prediction gap with humans, organizations must adapt their strategies to leverage this new resource for more informed and robust decisions.