A British artificial-intelligence startup co-founded by a former DeepMind researcher placed eighth in Metaculus’s international forecasting tournament, outperforming many human participants—including some professionals—while still trailing the very best forecasters. The system, which orchestrates multiple AI models from OpenAI, Google and DeepSeek to research, simulate scenarios and update probabilities, impressed contest organizers and raised expectations that AI could match top human forecasters by around 2029. Executives and forecasters said AI excels at persistence and breadth across many questions, but still stumbles on complex, interdependent judgments and logical consistency checks. Researchers including Philip Tetlock say humans retain an edge in sparse-data domains requiring judgment. The emerging consensus: hybrid human-AI teams will likely deliver faster, more accurate forecasts for investors and corporations.
Related articles:
