SetupAI
繁
SourceStalecollected in 51m

ALS patient uses brain implant to speak with 99% accuracy

Read original on The Next Web (TNW)
#bci#neurotech#als#healthcare-ai

A major breakthrough in BCI longevity and accuracy, enabling real-world, long-term independent communication.

30-Second TL;DR

What Changed

Patient achieved 99% speech accuracy over 3,800 hours of use.

Why It Matters

This milestone demonstrates the viability of long-term, high-accuracy BCI systems for restoring autonomy to patients with severe motor impairments.

What To Do Next

Review the latest Nature Medicine publication to understand the signal processing latency and decoding architecture used in this BCI study.

Who should care:Researchers & Academics

Key Points

  • •Patient achieved 99% speech accuracy over 3,800 hours of use.
  • •System enabled full-time employment despite advanced ALS.
  • •Average communication speed reached 56 words per minute.
Key numbers92%99%

Deep Insight

Background and context from public sources — not the original article. 25 sources cited.

Enhanced Key Takeaways

  • •The UC Davis system allows for independent at-home use without researcher support, a significant advancement over previous brain-computer interface (BCI) systems that typically required lab settings or constant researcher presence.
  • •The decoded speech is rendered in a synthesized voice designed to sound like the patient's pre-ALS voice, created from existing audio samples.
  • •The implant consists of four microelectrode arrays, specifically Blackrock NeuroPort arrays, placed in the left precentral gyrus to record activity from 256 cortical electrodes.
  • •The study, published in Nature Medicine, highlights that in daily use outside the lab, the patient rated 92% of sentences as accurate or mostly correct, communicating over 183,000 sentences.

Competitor Analysis

UC Davis (BrainGate2 Consortium)
Implant Type/Method
Four microelectrode arrays (Blackrock NeuroPort)
Target Brain Region
Left precentral gyrus (speech coordination)
Key Functionality
Speech decoding, text-to-speech, real-time voice synthesis
Reported Accuracy/Speed
99% accuracy, 56 WPM (latest); 97% accuracy, 32 WPM (earlier)
Invasiveness
Invasive (craniotomy for electrode placement)
Neuralink
Implant Type/Method
N1 Implant (1024 tiny electrodes)
Target Brain Region
Motor cortex
Key Functionality
Computer cursor control, typing, AI-cloned voice
Reported Accuracy/Speed
Not directly comparable for speech WPM, but enables computer control
Invasiveness
Invasive (craniotomy for chip and thread insertion)
Synchron
Implant Type/Method
Stentrode (endovascular device)
Target Brain Region
Blood vessel near motor cortex
Key Functionality
Control of digital devices (e.g., Apple Vision Pro, Amazon Fire tablet), texting, online activities
Reported Accuracy/Speed
Not directly comparable for speech WPM; focuses on device control
Invasiveness
Minimally invasive (via jugular vein)
Stanford (BrainGate Consortium, often with Blackrock implants)
Implant Type/Method
Intracortical sensors (Blackrock NeuroPort array)
Target Brain Region
Speech-related regions
Key Functionality
Speech decoding, text-to-speech
Reported Accuracy/Speed
62 WPM, 97% accuracy (Pat Bennett study)
Invasiveness
Invasive (craniotomy for sensor placement)

Technical Deep Dive

  • Implant Type: Four microelectrode arrays (Blackrock NeuroPort array) are surgically implanted.
  • Electrode Count: The arrays record activity from 256 cortical electrodes.
  • Brain Region: The implants are placed in the left precentral gyrus, the brain region responsible for coordinating speech.
  • Decoding Algorithms: Machine learning algorithms, part of a software platform called BRAND (developed by Nicholas Card), translate neural activity into English-language phonemes.
  • Speech Synthesis: The system maps phonemes to words and sentences, which are then read aloud by a synthesized voice created from the patient's pre-ALS voice samples.
  • Real-time Speech Synthesis (newer development): The system can directly translate neural activity into synthesized speech in real-time, allowing for natural interruptions and vocal modulation.
  • Processing Speed: The entire processing chain, from neural signal acquisition to speech synthesis, occurs within 10 milliseconds, comparable to the natural delay in hearing one's own voice.
  • Decoding Model: A multilayer Transformer-based model is used to predict acoustic speech features.

Future ImplicationsAI analysis grounded in cited sources

BCI technology will become more widely accessible for at-home use.
The UC Davis study demonstrates the first sustained independent at-home use of a BCI for communication, overcoming a major barrier to real-world adoption.
The integration of AI will lead to increasingly natural and nuanced BCI communication.
Advanced decoding algorithms and real-time speech synthesis, including vocal modulation and personalized voices, indicate a trend toward more human-like interaction.
BCIs will significantly enhance the independence and quality of life for individuals with severe paralysis.
The ability to communicate independently, work full-time, and interact digitally empowers patients to regain agency and participate more fully in society.

Timeline

2023-07
UC Davis team implanted microelectrode arrays into Casey Harrell's brain.
2024-08-14
Initial UC Davis study (Casey Harrell) published in New England Journal of Medicine, reporting 97% accuracy and 32 WPM.
2025-04-28
UC Davis team won The Herbert Pardes Clinical Research Excellence Award for their BCI work.
2025-06-12
UC Davis published research in Nature on real-time speech synthesis.
2026-06-15
UC Davis study published in Nature Medicine, highlighting independent at-home use, 56 WPM, and 99% accuracy over 3,800 hours.
Read original article on The Next Web (TNW)

Weekly AI Recap

Read this week's curated digest of top AI events →

PostLinkedIn

AI-curated news aggregator. All content rights belong to original publishers.
Original source: The Next Web (TNW) ↗

This is a summary, not the original. Read the source, or get the weekly briefing.

The weekly digest

One email a week. Unsubscribe anytime.