An Algorithm Now Breeds Video to Steer Human Brain Activity
The team's evolutionary search breeds AI video prompts using an fMRI model that predicts each voxel's response in the visual cortex.
- NEvo's bred videos beat decades of handcrafted localizer clips still standard in neuroscience labs.
- The lateral visual stream, tuned to social movement, showed the biggest AI-over-old gap.
- Six authors, including EPFL's Amir Zamir and MIT's Martin Schrimpf, posted the paper July 2, 2026.
Why it matters: Mapping which images seize a brain circuit is one step from choosing which images to show it.
NEvo project page (EPFL) / arXiv:2607.02317 ↗ · 10 ביולי 202610.7.26 · ✓ Checked✓