Artificial Intelligence Unveils Secrets of the Brain's Language System
The application of AI models to recordings of human brain neuron activity during conversations has allowed scientists to reveal a pattern of brain activity reflecting key language features. The study provides insight into how neurons encode linguistic information.
Using AI to decode brain signals is revolutionizing neurobiology. New machine learning methods enable the correlation of vast amounts of text data with terabytes of neuron activity information.
Brain activity data were obtained using microelectrode arrays implanted in 8 patients for epilepsy monitoring. Specialists from Massachusetts General Hospital in Boston took this opportunity to record natural conversations in English on various topics.
Read this article in full and everything else in Telegram, ad-free
Open in TelegramThanks, I just want to finish reading here
Researchers correlated text transcriptions of conversations with the activity of hundreds of neurons in the frontotemporal cortex. The analysis showed that recordings of neural activity just before participants began speaking allowed for the prediction of many properties of subsequent speech. The work was published in the journal Nature.
Division of Responsibilities and Context
The authors discovered a clear division of functions among the studied cells. Some neurons reflected basic information, such as the meaning and syntactic roles of specific words, while others took on more complex tasks, including combining phrases into structured sentences.
The mathematical models created by the scientists successfully distinguished similar phrases and words, indicating the ability of neural activity to capture the unique context of sentences. These results reveal how individual cells encode language during conversation.
