AI Reads Mouse Brain to Replay Movies It Watched

In a new study, researchers have taken a step toward reading minds—or at least, reading movies from the brain. They recorded the activity of nearly 8,000 neurons in a mouse's visual cortex while it watched natural movie clips. Then they used a powerful AI model, similar to those that generate videos from text, to turn those brain signals back into video. The reconstructed clips were so accurate that a computer could pick the correct original clip out of a lineup 81% of the time. This means the brain activity contained enough information to identify what the mouse was seeing.

The AI didn't just copy the movie; it learned to separate two things: the layout of the scene (where objects are) and the meaning (what the objects are). By combining these, it could generate new frames that matched the original movie's content and timing. The researchers also tested the method on a simulated dataset of over 116,000 neurons from 11 different experiments, showing the approach works across larger scales. This suggests that the brain represents visual experiences in a distributed way, with different aspects handled by different patterns of activity.

Why does this matter? For one, it's a major step toward brain-computer interfaces that could let paralyzed people communicate by thought or allow doctors to see what a patient is imagining. It could also help us understand how the brain processes the world, leading to better treatments for visual disorders. But there are big caveats: this was done in mice, not humans, and required invasive surgery to measure neurons directly. The technology is far from being ready for everyday use, and ethical concerns about privacy of thought will need to be addressed. Still, it's a fascinating glimpse into the future of neurotechnology.