A new AI system has been developed that can translate brain activity into a continuous stream of text, representing a significant leap in brain-computer interface technology. The system, developed by researchers at the University of Texas at Austin, uses a non-invasive functional MRI (fMRI) scanner to measure blood flow in the brain. Participants listened to hours of podcasts while in the scanner, and the AI model, a custom-trained GPT variant, learned to associate specific brain activity patterns with the meanings of words and phrases. In tests, the system could produce text that closely captured the gist of a participant’s thoughts, though not a verbatim transcript. The research raises important ethical questions about mental privacy and the potential future applications of such technology. For the full details, read the complete article at https://technologyreview.com/2024/05/01/1092338/ai-reads-brain-activity-to-reconstruct-thoughts/.