A team of US scientists has created an AI system that can interpret a person’s brain activity.
New York: Researchers at the University of Texas at Austin have developed an AI system that is capable of converting a person’s brain activity into text.
The researchers trained the semantic decoder using transformer models, which power Google’s Bard and OpenAI’s ChatGPT. Participants listened to podcasts while in an fMRI scanner, and the machine then generated corresponding text from their brain activity alone. This system is a significant advancement because it can decode continuous language with complicated ideas instead of only single words or short sentences.
The researchers ensured that the decoder would only work on cooperative participants who had willingly participated in its training, and it would not function on untrained participants, thus addressing concerns about potential misuse of the technology. The team hopes to adapt the technology for use with other portable brain-imaging systems, such as functional near-infrared spectroscopy, despite its current impracticality outside of the lab. The system does not produce a word-for-word transcript, but rather captures the essence of the spoken or imagined words.