Using embodied AI to study how eyes and vision evolved

Topic: Using embodied AI to study how eyes and vision evolved

When: 27 May, 2026 12.00 to 13.00 CET

Where: Online

Speaker: Dan-E. Nilsson, Lund Vision Group, Lund University

Abstract

Our understanding of how of eyes and vision has evolved is largely based on the astonishing diversity of visual systems we see today, their structure, function, neural connections, their molecular signatures, and underlying genes. Very little is based on fossil evidence, simply because not that much has been preserved in useful states. It is every evolutionary biologist’s dream to be able to travel back in time and directly study how thing evolved. Using embodied artificial intelligence, we have come close to such time travel. We designed agents that randomly modify their sensory equipment and then gave them different visual tasks to solve. By continuously selecting the most successful agents we were able to mimic real evolution of vision, and by changing tasks and conditions we could learn about the reasons vision evolved the way it did. This is a novel and powerful way of using AI in evolutionary biology.