The AI Forecasting Agenda
...
The AI forecasting landscape has evolved, now rivaling top human performance on traditional forecasting questions, such as those found on ForecastBench and the Metaculus Cup. Specialized AI models, developed by startups like FutureSearch, Mantic, and Preseen, have reached this level, with foundation models like ChatGPT and Claude expected to match human performance by the end of the year.
This progress has solved the problem of accessing top-human-level forecasts, which previously required expensive teams of superforecasters or unreliable prediction markets. The "science of forecasting" has transformed into a more straightforward task, with AI forecasters capable of handling various types of forecasting questions. However, there remain challenges to overcome to fully utilize AI forecasters' capabilities.
Written by urgent.news from Astral Codex Ten's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.