York University study finds AI processes differ from human brain
Researchers at York University discovered that artificial intelligence models do not process visual information like the human brain on November 21, 2026.
Reverse testing reveals hidden differences
Scientists tested whether recorded brain activity could predict internal processes in AI models. This reverse predictivity testing method showed that the relationship does not work in both directions. While AI successfully predicts brain responses to images, the brain cannot predict what happens inside the computer systems. The study examined 1,320 natural photographs and synthetic images featuring ten familiar subjects like bears and zebras. Researchers also included 300 additional images showing these subjects in artistic styles or outlines.
AI models lack true neural mechanisms
York University Assistant Professor Kohitij Kar led the investigation into how vision works computationally. He noted that systems described as brain-like only match activity patterns in one direction. The team included Postdoctoral Fellow Sabine Muzellec who worked with the Connected Minds program. Their findings challenge the reliability of artificial neural networks used to explain the visual system. These models rely on internal processes that the actual human brain does not use.
Questions remain for neuroscience research
The new results raise important questions about how scientists understand artificial intelligence today. Researchers must determine if computational models truly reflect underlying neural mechanisms for object recognition. The study was published by York University and challenges long-held assumptions in the field. Future work needs to evaluate similarities that conventional tests previously overlooked completely.
Reported by one outlet
Only one outlet has published this. Nothing here has been checked against a second report, so read it as that outlet's account and follow the link below for the original.
Reported by
1 independent outlet. Headline as published. Links open the original report.