AI has come closer to 'reading minds' – scientists have demonstrated a new technology

Scientists have developed an AI system capable of reconstructing images a person is viewing based on fMRI data. The system analyzes brain activity to identify objects, their location, and colors. The model was trained on data from eight volunteers, each viewing about 9,000 images. Adapting to a new person requires about an hour of data. The technology is still imperfect: it mistook a cake for three sandwiches and a dog for a goat. Future plans include reconstructing video, sounds, and dreams. The development could help paralyzed individuals but raises questions about mental privacy.

Scientists have presented an AI-based system that reconstructs images perceived by a person using brain scan data (fMRI). The technology analyzes activity in various brain regions and matches signals with visual information, identifying the object, its location, colors, and composition. The model was trained on data from eight volunteers, each viewing about 9,000 images. Adapting the system to a new person requires about an hour of data—significantly less than previous developments. The technology is still far from perfect: in one case, the system recognized a cake as three sandwiches and a dog as a goat. Researchers plan to expand capabilities to reconstruct video, sounds, imagined scenes, and even dreams. Scientists emphasize that the technology cannot literally 'read minds,' but only reconstructs visual information based on patterns of brain activity. One potential application could be helping people with severe paralysis. However, the development of the technology raises questions about protecting mental privacy and the possibility of using brain activity data without a person's consent.