A new AI tool developed by researchers at the Weizmann Institute of Science in Rehovot, Israel can reconstruct images that a person is looking at based on their brain scans, and vice versa. The system, led by Michal Irani, uses high-resolution fMRI data to predict both the structure and content of viewed images with remarkable precision. The team trained their "brain decoder" on publicly available datasets from eight individuals who viewed around 9,000 images during scanning sessions.
The decoder splits predictions into two branches—one for image structure and another for content—feeding into a diffusion model that generates accurate reconstructions. Earlier attempts produced blurry, indistinct images, but advances in fMRI resolution and AI modeling have dramatically improved results. The researchers hope the technology could eventually help locked-in patients communicate or enable scientists to study dreams.
However, concerns about misuse have been raised. Tommy Sprague, a neuroscientist at the University of California Santa Barbara, warns that such technology could potentially extract private thoughts without consent. Judy Illes, a neuroethicist at the University of British Columbia, praised the work as "magnificent" while acknowledging therapeutic potential for people with neurologic conditions.
Source: MIT Technology Review · Summarized by HeadlinesBriefing