Researchers at Stanford University have developed a brain-computer interface that can translate the unspoken thoughts of people who have lost the ability to speak, according to a study published in the journal Cell. The system achieved 98.75% accuracy in a small trial involving four participants, including ALS patient Casey Harrell.
The technology addresses a key limitation of existing BCIs, which typically require users to attempt speech, a process that can be exhausting for people with conditions like amyotrophic lateral sclerosis (ALS) that weaken airway muscles. "If we could decode that, then that could bypass the physical effort," Stanford neuroscientist and coauthor Erin Kunz told the New York Times. "It would be less tiring, so they could use the system for longer."
In the study, researchers trained AI models to link thoughts to words, enabling the computer to translate complex sentences such as "I don't know how long you've been here." However, they discovered that the system sometimes picked up words participants were not imagining saying aloud, raising concerns about mental privacy.
A Mental Password to Protect Private Thoughts
To address this, the team introduced an "inner password" that turns decoding on and off. Participants imagined the phrase "Chitty Chitty Bang Bang" before and after the sentence they wanted spoken. The computer complied 98.75% of the time, according to the study.
Kunz described the trial as a "proof-of-concept," but it represents a step toward making BCIs more practical and privacy-conscious for people with speech-inhibiting disorders.
Editorial Notes