{"stable_id":"c42c26c667a37013","slug":"ucsf-speech-neuroprosthesis","company":"University of California, San Francisco and UC Berkeley","workflow_name":"AI speech neuroprosthesis for text, audio, and avatar control","function_code":"clinical_ops","pattern_codes":["creator_to_judge"],"changed_assumption":"Attempted-speech brain signals can produce real-time text, audio, and facial animation as a faster channel.","evidence_strength":"verified","publication_tier":"showcase","freshness":"current","reviewed_at":"2026-08-23","updated_at":"2026-08-22","source_quality_summary":"1 peer reviewed; publication outcomes are verified.","caveat_summary":"Single-participant investigational system, not an approved general product. A 25% error rate requires correction and limits high-stakes communication.","freshness_caveat":null,"workflow_summary":{"before":[{"actor":"Participant","action":"Uses head tracking to select letters on a commercial assistive device.","actor_type":"control"},{"actor":"Communication partner","action":"Reads the slowly composed text.","actor_type":"control"}],"hinge":"Attempted-speech brain signals can produce real-time text, audio, and facial animation as a faster channel.","after":[{"actor":"Participant","action":"Attempts to speak while 253 cortical electrodes capture speech-motor signals.","actor_type":"system"},{"actor":"AI decoder","action":"Decodes text and synthesizes speech and avatar movements in real time.","actor_type":"ai"}],"decision_mode":"shared","decision_marker":"Human moves from creator to judge"},"before":[{"order":1,"actor":"Participant","action":"Uses head tracking to select letters on a commercial assistive device.","handoff_to":"Communication partner","control":"User intent and device selection."},{"order":2,"actor":"Communication partner","action":"Reads the slowly composed text.","handoff_to":"Conversation","control":"Human clarification."}],"after":[{"order":1,"actor":"Participant","action":"Attempts to speak while 253 cortical electrodes capture speech-motor signals.","handoff_to":"AI decoder","control":"Participant-initiated use and clinical trial protocol."},{"order":2,"actor":"AI decoder","action":"Decodes text and synthesizes speech and avatar movements in real time.","handoff_to":"Communication partner","control":"Language model, confidence, and user correction."}],"decision_rights":"The participant owns intended communication and decides use; clinical teams control the investigational device and safety.","exception_path":"Incorrect text is corrected or repeated; the participant can revert to the existing assistive device.","removed_work":["Letter-by-letter composition for many utterances"],"outcomes":[{"metric":"Communication rate","baseline":"14.2 words per minute with the participant's commercial head-tracking device.","result":"Median real-time decoding rate of 78.3 words per minute.","period":"Nature study reported August 2023.","scale":"One clinical-trial participant with severe paralysis.","attribution_caveat":"Single-participant investigational system, not an approved general product.","evidence_label":"verified"},{"metric":"Text decoding error","baseline":"Naturalistic speech speed and accuracy were previously elusive for neuroprostheses.","result":"Median word error rate of 25%.","period":"Study evaluation.","scale":"Large-vocabulary real-time decoding for one participant.","attribution_caveat":"A 25% error rate requires correction and limits high-stakes communication.","evidence_label":"verified"}],"executive_lesson":"A breakthrough accessibility outcome can coexist with material error, invasiveness, and single-user limits; all belong in the case.","anti_pattern":"Calling an investigational single-participant result a deployable cure.","questions_for_leaders":["How is consent verified when decoding is wrong?","What noninvasive alternatives remain?","How durable is calibration?"],"collections":["human-still-decides","regulated-autonomy"],"bundle_version":"1.0.0","bundle_fingerprint":"sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93","canonical_url":"https://brianletort.ai/transformations/ucsf-speech-neuroprosthesis"}