Scientists have developed an artificial intelligence system capable of determining what a person is viewing based on their brain activity and reconstructing an image of it.
This technology, named “Brain-IT,” was developed by researchers at the Weizmann Institute of Science. During the research, specific patterns of brain activity generated while individuals viewed various images were identified using functional MRI (fMRI) scans.
The results demonstrated that the technology successfully reconstructed recognizable images of various objects, foods, faces, landscapes, and other complex visual scenes.
Although this technology is sometimes described as “mind-reading,” Brain-IT does not spontaneously read thoughts existing within a person’s mind; rather, it requires the individual to be inside an MRI scanner and view an image, whereupon the system generates an approximate visual representation based on the resulting brain activity.
A key feature of this technology is its ability to rapidly adapt to the brain patterns of new individuals; while older methods required hours of brain scan data to yield useful results from a new subject, Brain-IT needs only about an hour to learn from a new person.
“We must accept some downsides to reap the benefits of AI.”
— Sam AltmanResearchers achieved this breakthrough by focusing on shared patterns found in the brains of different individuals. They divided brain activity into thousands of tiny segments and identified signals within them that correspond to visual features—such as color and location—as well as high-level categories like faces, food, and sports.
Why it matters: Brain-IT represents a significant leap in brain-computer interfaces, potentially enabling new ways to understand visual perception and assist individuals with communication difficulties. Its rapid adaptation makes it more practical for real-world applications, though ethical considerations around privacy and consent remain important.
