A new generation of analysts is being trained, not in universities, but in the data centers of tech companies. These analysts are bots, and a recent competition has shown they are learning to see the future with remarkable skill. A system from UK startup ManticAI ranked eighth in the Metaculus Cup, outperforming many of its traditionally trained human counterparts.
This new generation of AI analysts is being developed by companies like ManticAI, which was co-founded by a former Google DeepMind researcher. They are not just programmed with facts; they are trained on vast datasets of text and events, learning the patterns and relationships that govern how the world works.
Their analytical process is both rigorous and scalable. ManticAI’s bot uses a team of specialized AI agents to tackle a problem. Some act as researchers, others as scenario planners. This digital team can work around the clock, simultaneously analyzing dozens of complex global issues—from political leadership contests to the severity of wildfire seasons.
The results of this training are impressive. The AI’s predictions are not only accurate but often original, avoiding the cognitive biases like groupthink that can affect human analyst teams. As its creator noted, the bot often “strongly disagreed” with the human consensus, providing a valuable, independent perspective.
While they still have things to learn and currently struggle with certain types of complex logical reasoning, the trajectory of these AI analysts is steep. The performance of ManticAI is a clear sign that the future of analysis, in fields from finance to intelligence, will be shaped by these powerful new digital minds.