The brain is becoming an image-generation device. An AI “mind-reading” tool can analyze brain scans, guess what a person is looking at, and re-create that image with remarkable precision. The same tool can also predict a person’s brain activity from the image they are viewing — a two-way link between visual input and neural signals.
Michal Irani, a researcher at the Weizmann Institute of Science in Rehovot, Israel, is among the researchers behind the work. The team hopes the “mindreading” tool will reveal more about how the brain works, although the technology’s possible uses reach beyond basic research.
From brain activity to reconstructed images
The system focuses on activity recorded in the brain through scans. Each highlighted “voxel” of activity covers around three cubic millimeters and contains around 16,000 neurons. That is a lot of neural activity compressed into a small space — and a reminder that a brain scan does not offer a neat pixel-by-pixel view of thought.
Newer datasets used voxels covering around one cubic millimeter of neurons. The researchers also trained the tool with a data mix in which around 70% of the training data came from images that were not originally paired with fMRI scans, adding another unusual detail to a method built around connecting pictures with brain activity.
The result is not just a prediction of what someone saw. The tool can re-create the image itself from the scan, while also predicting the brain activity associated with an image a person is viewing. That ability gives researchers a way to compare the brain’s response with the visual content that prompted it.
Tommy Sprague, a neuroscientist at the University of California, Santa Barbara, said, “The results seem very impressive.” The assessment is brief, but the underlying claim is not: an AI model can move between an image and the activity linked to that image.
Useful tool or privacy problem?
The researchers hope the technology will help explain how the brain processes what people see. It could also help locked-in people communicate, offering a possible route to express information when physical movement is unavailable.
Another proposed use is dream research. Scientists could use the tool to re-create the content of dreams, turning private mental imagery into something that can be examined outside the dreamer’s mind. That would open a striking research path — and a deeply uncomfortable one.
Other scientists warn that a similar approach could reveal people’s inner thoughts and mental imagery, potentially without their consent. The concern is not limited to identifying an image on a screen; the same basic approach raises questions about whether brain activity could expose content a person never chose to share.
Judy Illes, a neuroethicist and professor of neurology at the University of British Columbia in Canada, is among the voices connected to that ethical debate. The tool’s possible value for communication and neuroscience sits beside a clear privacy problem: brain scans may carry information that people cannot control once it is analyzed.
The technology remains tied to the facts its researchers can extract from brain scans and training data. Still, the direction is clear. AI can now connect neural activity with visual images in both directions, giving researchers a powerful instrument for studying the brain while forcing a basic question into the open — who controls the images inside someone’s head?
Based on
