Photo by Phyllis Lilienthal on Pexels

A psychological audit from 2025 casts doubt on the near-term arrival of Artificial General Intelligence (AGI), revealing significant limitations in AI’s ability to generalize beyond its training data. The audit, spotlighted on Reddit’s Artificial Intelligence forum (link: https://old.reddit.com/r/artificial/comments/1owpmmq/the_generalisation_illusion_a_2025_psychological/), suggests that even advanced models like GPT-4 struggle with true out-of-distribution (OOD) reasoning, indicating that achieving human-level cognitive flexibility remains a major challenge for AI research.