The technology that allows you to transcribe your working meetings can help people with paralysis to talk again.
Researchers at UC Berkeley and Uz San Francisco used General AI to reduce the delay between when a person with severe paralysis is trying to talk when the computer device plays the sound. Their work helped a woman named Ann, who suffered a stroke in 2005 at the age of 30, to communicate close to real time. Ann spoke with a voice that sounded like her own because the model was trained in footage of her before her stroke.
General AI's deployment in several different ways has allowed researchers to make improvements in neuroprosthesis that can last much longer, said Jol Jun Joe, Dr. Berkeley. Student in Electrical Engineering and Computer Science and Co-Water Author of the Study, who appeared in March in Neuroscience of Nature.
It is an example of how generative tools for you – using the same basic technology that authorizes Chatbots Like Chatgpt and Chatgpt and Anthropic's Chatgpt and Anthropic in Google Meet – they help medical and scientific researchers solve problems that can take much longer to solve, Joe told me. Experts and supporters of AI pointed to the use of technology in medicine as an area with a huge upside downWhether it's in devising new drugs or security Better testing and diagnosing.
“AI accelerates progress,” Joe said. “Sometimes we imagined that the timeline would be a decade or two. Now, when the pace is like three years.”
The technology that helped Ann is proof of concept, Joe said, but shows a path to tools that could be more plugins and play in the future.
Accelerating the speech
The problem with existing neuroprosthesis is latency. There is a time lag between when the person starts trying to speak and when a sentence is actually generated and heard. Joe said previous technology meant that Ann had to wait until one sentence was finished before the next one began.
Ann, seen during the first research study in 2023, was able to communicate through computers reading the signals that her brain tried to send to the speech muscle.
“The main progress is here that she should not wait until the punishment is over,” he said. “Now we can convey the decoding procedure whenever she intends to speak.”
The prosthesis includes a series of electrodes that are embedded on the surface of its brain and are connected through a computing bank cable. It decodes the control signals of the brain to the muscles that control the speech. After Ann chose the words she intends to say, AI reads those signals from the motor cortex and gives them life.
To train the model, the team had an attempt to speak sentences shown on the ambulance on the screen. They then used data on that activity to map the signals in the engine cortex, using Gen AI to fill the gaps.
Joe said the team hoped the breakthrough leads to devices that are scalable and more accessible.
“We are still in the ongoing efforts to make it more precisely and lower latency,” he said. “We are trying to build something that can be more plugin and play.”
Using AI to go from thought to speech
Joe said the team used General AI in several different ways. One was to replicate the voice before Ann's injury. They used footage of her injury to train a model that could produce the sound of her voice.
“She was very excited when she first heard her voice again,” Joe said.
https://www.youtube.com/watch?v=mgsokgGBBXK
The great change was in real -time transcription. Joe compared it to the tools that transcribe presentations or meetings as they occur.
Work built on a 2023 study It used AI tools to help Ann communicate. That job still had a significant delay between when Ann tried to speak and when the words were made. This research significantly reduced that delay, and Ann told the team that he felt more natural.
“She announced that she felt embodied, that it was her speech,” Joe said.