Tether’s AI and neuroscience unit said a single brain-computer interface model can generalize across people for speech, vision and music tasks, including image reconstruction from macaque neural activity.
Tether Evo, Tether’s artificial intelligence and neuroscience research initiative, has published three peer-reviewed papers on brain-computer interfaces (BCIs), focusing on whether a single AI model can interpret neural signals across different individuals rather than being trained separately for each person. The work targets speech, vision and music, an approach aimed at addressing a longstanding BCI challenge because every brain produces slightly different signals. Two of the papers were developed with the University of Rome Tor Vergata (UniTOV). One study on speech was published in the Journal of Neural Engineering and examines cross-subject decoding of human neural data for speech BCIs. The research is framed as support for people who have lost functions such as vision, hearing or speech after injury or accident. In the vision study, Tether and UniTOV researchers recorded neural signals from macaques viewing thousands of images and reconstructed what the animals were seeing from that activity, with 70% accuracy using 200 milliseconds of neural data. Tether said the model produced plausible reconstructions that captured shape, color and content. A third study analyzed music by recording fMRI (brain imaging that tracks blood-oxygen changes) scans from five people as they listened to 540 songs across 10 genres. Tether linked the broader effort to privacy, saying neural data should remain protected through an on-device AI stack.