AI forecasters hit parity by 2026

Updated: 2026.01.14 14D ago 5 sources
The Forecasting Research Institute’s updated ForecastBench suggests AI forecasters are on track to match top human forecasters within about a year. Phil Tetlock’s 'best guess' is 2026, contradicting longer 10–15 year timelines. — If AI equals superforecasters soon, institutions in policy, finance, and media will retool decision processes around AI‑assisted prediction and accountability.

Sources

Silver Bulletin pollster ratings 2025 archive
Nate Silver 2026.01.14 62% relevant
Silver’s ratings address the same core problem that ForecastBench and superforecaster research target—how to evaluate and weight probabilistic political forecasts. The pollster ratings provide a reproducible, evidence‑based prior (Predictive Plus‑Minus) that forecasters and institutions should combine with model outputs (including AI) when producing election probabilities.
Silver Bulletin pollster ratings, 2025 update
Nate Silver 2026.01.14 45% relevant
The updated human pollster accuracy benchmarks matter for claims that AI forecasters will reach or exceed human forecasting skill — Silver’s ratings supply the empirical baseline against which AI forecasting systems should be compared and audited.
So, who’s going to win the Super Bowl?
Nate Silver 2026.01.07 72% relevant
This article is a concrete example of the same broader phenomenon: algorithmic forecasters (here ELWAY plugged into QBERT) producing probabilistic predictions that can outperform or correct human consensus; it illustrates how domain‑specific models are shaping public expectation and betting markets, the very arena ForecastBench and parity discussions claim AI systems will dominate.
What I got wrong in 2025
Matthew Yglesias 2026.01.05 60% relevant
Yglesias’ admission that human punditry produced notable misses (especially in foreign elections) strengthens the case for improving forecasting methods and tools; it concretely illustrates the human error this existing idea argues AI/forecast‑bench systems may soon be able to match or outperform.
From the Forecasting Research Institute
Tyler Cowen 2025.10.09 100% relevant
Tyler Cowen’s post citing FRI’s ForecastBench update and Phil Tetlock’s 2026 estimate (via tweet/Substack).
← Back to All Ideas