Why Are Large Language Models So Terrible at Video Games? (spectrum.ieee.org) AI

An IEEE Spectrum interview with NYU’s Julian Togelius argues that large language models still struggle to play video games because they lack general game AI, are not trained for the diverse mechanics and spatial reasoning that games require, and benchmarks for games haven’t shown clear progress like coding benchmarks have.

June 01, 2026 09:40 Source: Hacker News