Shape representations in humans and Artificial Intelligence systems

What is it about?

We analysed how an artificial intelligence (AI) system, trained to recognise large datasets of images, responds when these images are deformed. While humans are hugely sensitive to deformations that change relations between parts, we observed that the AI systems do not show such a sensitivity. This difference suggests that human visual representations do not simply emerge as a consequence of learning to classify large datasets of objects.

Featured Image

Photo by Paul Skorupskas on Unsplash

Photo by Paul Skorupskas on Unsplash

Why is it important?

This study shows significant differences in how the human visual system and recent artificial intelligence (AI) models represent objects. This difference between the two systems likely stems from their different goals—while AI models are built to classify objects, humans must additionally reason and interact with them. Our results suggest that the human visual system represents objects in a manner that enables us to perform these additional tasks.