← Back to archive

Why IQ Is a Misleading Benchmark for AI

Sam Altman recently remarked that AI's IQ has been rapidly improving, estimating that AI gains roughly one standard deviation of IQ per year.

However, IQ is a misleading benchmark for AI. IQ tests were designed for humans, measuring logic and abstract reasoning but not practical intelligence.

AI models have an unfair advantage due to vast memory and are trained on common patterns. Intelligence is far more complex and IQ tests are biased and unsuitable for AI.

We need to figure out better AI evaluation methods, rather than directly comparing AI to human cognition.

Why IQ Is a Misleading Benchmark for AI

Originally posted on X

PS: I'm pretty active on 𝕏 if you'd like to follow along

Archive

2026

2025

2024

2018