Brain-computer interface can now decode your inner thoughts
Stanford researchers have figured out how to read people's silent thoughts using brain-computer interfaces (BCIs), reaching 74% accuracy.
The study, published August 14, 2025, worked with four paralyzed participants who had tiny electrodes in their brains.
An AI model turned their imagined words—picked from a huge vocabulary—back into text.
Participants used a 'password' phrase to trigger the decoder
The system could even pick up on things like silently counting numbers.
To keep things accurate, participants used a "password" phrase that triggered the decoder only when they wanted. This method nailed over 98% accuracy for those phrases.
Unlike older BCIs that rely on muscle movement (which can be tiring), this approach taps directly into inner speech patterns, making communication feel more natural.
This could help people who can't speak
Researchers are optimistic: with better tech, this could help people who can't speak regain fluent conversation just by thinking what they want to say.