Brain implant cuts first-sound latency to ~1 second for paralyzed speaker
coursiv_ · reddit · 2026-11-04
In a 2025 speech-implant study, UC Berkeley and UCSF researchers worked with Ann, severely paralyzed: an implant recorded speech-production brain signals and an AI model mapped them to speech. The earlier approach had 8-second delays per sentence; the new streaming system produced the first sound within about a second of detecting her attempt to speak, and used pre-injury recordings to make the voice sound like her own.
Notably, the system began generating speech before she finished the sentence. It's an experimental single-participant system decoding attempted speech, not arbitrary private thoughts. The authors ask what would make a synthetic voice feel truly yours — accent, laugh, emphasis? Paper in Nature Neuroscience.
More from AGI Musings
- Sam Altman says AI is entering the RSI era: how smart should a system be allowed to be? — ChrisGPT · 2026-11-04
- AI vs bureaucracy will be bloody: 30-minute work sitting in a 7-day queue — airkatakana · 2026-11-04
- Researcher: 'not quite there yet' on LLMs often means superhuman within a year — proteinrosh · 2026-11-04
- Why LLM-generated lessons may kill educational YouTube videos for good — signulll · 2026-11-04
- Your end-of-day to-do list reveals exactly where AI still falls short — alliekmiller · 2026-11-04
- Why attack velocity is the underrated cyber metric AI agents are collapsing — _shreya_s · 2026-11-04