Australian scientists create AI that converts brainwaves into words
text_fields
Researchers in Australia are making strides in the field of neuroscience and artificial intelligence by developing a new system that can turn brainwaves into words.
This breakthrough AI model, created at the University of Technology Sydney (UTS), translates brain signals captured through an electroencephalogram (EEG) into readable text, effectively transforming thoughts into language.
The innovation is the result of work by PhD student Charles (Jinzhao) Zhou, alongside his supervisors Chin-Teng Lin and Dr. Leong. By using deep learning techniques, the model can detect specific brain signals and decode them into particular words.
In a demonstration, Dr. Leong wore a 128-electrode EEG cap and remained silent, yet the AI model output the sentence: "I am jumping happily, it's just me."
As reported by ABC News, the model currently operates using a limited vocabulary. This allows researchers to more accurately identify and match brainwave patterns to individual words during the early stages of development. The AI plays a critical role in distinguishing relevant signals from background noise, a complex task given that electrical activity from multiple regions of the brain tends to overlap when measured from the scalp.
Unlike Elon Musk's Neuralink, which requires surgical implants, this approach is entirely non-invasive. “We can't get very precise because with non-invasive, you can't actually put it into that part of the brain that decodes words,” said Lin, highlighting the trade-off between safety and precision.
Despite current limitations, the research holds significant promise for applications such as stroke recovery, speech therapy for autistic individuals, and restoring communication for people living with paralysis.
This project is part of a broader global movement to harness AI and EEG data to explore the brain’s functions. In a related advancement, scientists at Mass General Brigham in the U.S. recently developed an AI tool capable of predicting cognitive decline by analyzing brain activity during sleep. That tool identified 85% of future cases of cognitive deterioration with a 77% overall accuracy rate.
