A new AI-powered brain-computer interface has enabled a person with advanced ALS to speak fluently in real time, achieving over 99% word accuracy and a processing delay of just 30 milliseconds—fast enough for natural conversation. Over two years, the participant generated 2.7 million words, modulated intonation, and even sang.
The Research
Led by Dr. Sergey Stavisky, associate professor of neurological surgery at the University of California, Davis, the team implanted high-density intracortical electrode arrays in a man who had lost intelligible speech due to ALS. The device captures firing patterns from hundreds of individual neurons simultaneously—a data load that traditional methods cannot handle in real time.
The decoding system uses a dual-stage deep learning pipeline. First, a phonetic layer translates raw neural activity into phonemes, the basic sound units of speech. Then a large language model (LLM) processor arranges those phonemes into words and sentences. A separate deep learning voice synthesis model converts the text into audible speech modeled on the patient’s own pre-ALS voice recordings. The entire process runs with 30-millisecond latency, matching natural human conversation.
Published in conjunction with the Chen Institute and Science Prize for AI Accelerated Research, the study reports a word accuracy greater than 99% during home use. Over two years, the patient independently communicated with family, controlled a computer, and maintained full-time employment.
Why It Matters
For people with locked-in syndrome or conditions like ALS, losing the ability to speak is often more debilitating than losing motor control. This research demonstrates that high-fidelity speech restoration is clinically feasible. The dual-stage AI approach—phonetic decoding plus language modeling—could be adapted for stroke survivors with aphasia or individuals with cerebral palsy, potentially restoring conversational independence to millions.
What You Can Do
While an implanted BCI is not available for general use, you can actively maintain your speech and language networks. Practice articulating complex thoughts out loud, read aloud daily, and challenge your vocabulary with word games or learning a new language. These activities strengthen the neural pathways that support fluent speech.
Source: Neuroscience News
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.
