The Paris-based Meta Fundamental Artificial Intelligence Research (FAIR) laboratory and the Basque Center for Cognition, Brain and Language (BCBL), Spain, have worked together to develop new artificial intelligence models that can read and interpret brain signals to reconstruct typed sentences and map the precise neural processes that transform thoughts into spoken or written words.
The Paris-based Meta Fundamental Artificial Intelligence Research (FAIR) laboratory and the Basque Center for Cognition, Brain and Language (BCBL), Spain, have worked together to develop new artificial intelligence models that can read and interpret brain signals to reconstruct typed sentences and map the precise neural processes that transform thoughts into spoken or written words.
These are new opportunities for restoring communication in people with traumatic brain injuries without the need for surgical interventions.
The first study used magnetoencephalography (MEG) and electroencephalography (EEG) to record the brain activity of 35 healthy volunteers as they typed sentences, Techspot writes .
The system has a three-component architecture consisting of an image encoder, a brain encoder, and an image decoder. The image encoder builds a rich set of image representations independently of the brain. The brain encoder then learns to match MEG signals to these images. Finally, the image decoder generates a plausible image based on these brain representations. Character decoding accuracy has reached up to 80%.
The second study focused on understanding how the brain converts thoughts into language. By using artificial intelligence to interpret MEG signals as participants typed sentences, researchers were able to pinpoint the precise moments when thoughts were converted into words, syllables, and individual letters.
This study shows that the brain generates a sequence of representations, starting at the most abstract level (the meaning of a sentence) and gradually translating them into concrete actions, such as finger movements on a keyboard. The study also demonstrates that the brain uses a “dynamic neural code” to create a chain of successive representations, storing each one for a long time.
Although this technology is promising, several challenges remain before it can be applied in clinical settings. Models still need to improve decoding accuracy and optimize the use of MEG, which requires special conditions such as a magnetically shielded room and patient immobility.
Meta plans to address these limitations in future research by improving the accuracy and reliability of the decoding process, exploring alternative non-invasive brain imaging methods that are more practical for everyday use, and developing more sophisticated AI models that can better interpret complex brain signals.
The company also aims to expand its research to cover a wider range of cognitive processes and explore potential applications in areas such as healthcare, education, and human-computer interaction.
