Rise of the forecasters … Startups betting on AI to predict the future

Several startups are leveraging artificial intelligence (AI) to create more accurate forecasts than even the most specialized experts in fields ranging from stock market movements to the performance of Shakira's songs on music charts. Dan Schwarz, co-founder of the forecasting company FutureSearch, predicts that within a decade, traditional decision-making methods will seem like "complete Stone Age thinking."

AI forecasting models are already matching the abilities of the best superforecasters – individuals whose estimates are significantly superior to those of ordinary people – thanks to vast amounts of data gathered from both public and private sources.

In early September, two AI models achieved historic victories by topping the Metaculus Cup, a competition featuring both human and non-human prediction champions. FutureSearch's in-house superforecaster, Tom Liptay, acknowledges that his role may become obsolete in the medium term, as AI systems surpass him.

Theo Summer, co-founder of Preseen, used AI tools to transform $35 into $1.9 million in just seven months on the American prediction-making site Kalshi. Their aim is not to guarantee outcomes but to provide a probability of occurrence, enabling the examination of assumptions across interconnected events, such as fluctuations in oil prices and geopolitical conflicts.

While hedge funds have traditionally been the quickest to act on AI-driven predictions, AI forecasting is poised to further automate financial markets. Toby Shevlane, co-founder of Mantic, a forecasting startup, believes that AI is already having a small impact on trading decisions made by customers. As AI tools become more prevalent, retail investors, who lack access to expensive AI resources, may struggle to profit on prediction markets like Kalshi and Polymarket.

Beyond finance, AI forecasting models can also formulate their own questions about the future and assign probabilities based on possible outcomes. These advanced models could impact government policy, corporate decision-making, scientific research, and insurance risk assessments. Mantic provides American companies with likely scenarios of regulatory developments in their sectors and has collaborated with the British government.

Although widespread access to these tools can bring benefits, there are potential drawbacks. Theo Summer expresses concern about the possibility that countries like Russia or China might develop superior AI systems, posing a challenge to their use.