The landscape of professional forecasting is being reshaped by artificial intelligence, as demonstrated by a British AI system that recently bested a field of human experts. ManticAI, a startup with roots in Google’s DeepMind, secured an impressive eighth-place finish in the Metaculus Cup, a competition dedicated to predicting future global events. This result has left many in the field to ponder the imminent future of their profession.
Throughout the summer, contestants wrestled with 60 challenging questions. They were asked to predict everything from election results in Samoa to the scale of US wildfires, requiring a blend of research, analysis, and foresight. The AI’s ability to consistently provide accurate probabilities across such a broad spectrum of topics showcased a new level of machine capability.
What makes ManticAI’s performance remarkable is its underlying architecture. The system functions like a committee of AI experts. It deconstructs a problem and assigns different aspects to various AI models, leveraging their unique strengths for tasks like data analysis, scenario simulation, and final prediction synthesis. This method allows the AI to work tirelessly, constantly updating its forecasts based on new information—a feat difficult for human forecasters to replicate across dozens of problems simultaneously.
Ben Shindel, a professional forecaster who competed, admitted to the “weird feeling” of being outranked by bots, a sentiment that highlights the industry’s rapid transformation. He pointed out the massive leap AI has made in just one year, climbing from the 300s into the top 10. This trajectory suggests that the era of AI dominance in forecasting may be closer than previously thought.
While leading experts like Philip Tetlock maintain that top-tier humans still outperform bots on average, especially on questions with sparse data, the momentum is clearly with AI. The ultimate outcome, as suggested by Good Judgment CEO Warren Hatch, will likely be a hybrid model. The future of forecasting isn’t about choosing between human or AI, but about combining the strengths of both to achieve unprecedented accuracy.