AI and bigger data sets are improving some kinds of forecasting

In short: Better data and AI tools are making some forecasts more accurate, but they can still be misled by manipulated or low-quality information.

What's going on

A Financial Times opinion column argues that the world can feel chaotic, but in some areas, predictions are getting better. The main reason is that we now collect far more data, and AI models can spot patterns people would miss.

Weather is one example. The column says the best AI-supported seven-day forecasts now match the quality of three-day forecasts from 1980. It credits better sensors, more data, and more computing power.

The same approach is being applied to politics and business risk. Anthony Vinci, a former US intelligence officer, says he uses AI to help estimate the chances of future events. He describes four common approaches, expert individuals, group analysis inside organisations, prediction markets like Polymarket or Kalshi (betting-style markets where people buy and sell “odds”), and AI models that combine multiple sources.

Vinci’s company, Vico Technologies, is one example. The column cites a case where Vico assigned an 89% probability to a US attack on Iran by March 31, and the attack reportedly happened on February 28.

What to watch

The column warns that forecasts depend on the information fed into the system. Data can be polluted on purpose, including efforts to influence large language models (chatbots that generate text) by pushing misleading content. Another risk is synthetic data, meaning AI-made data, which can be like photocopying a photocopy until details get blurry.

Source: Financial Times