Artificial intelligence tools helped restore speech to a paralysis woman: “I felt embodied”

Technology that allows you to copy your business meetings may help people with paralysis to speak again.

The researchers at UC Berkelegi and UC San Francisso used the Tructe intelligence to reduce the delay between when a person who suffers from very paralysis tries to speak and when the computer plays the sound. Their work called Ann, who suffered from the stroke of the brain in 2005 at the age of thirty, helped in actual time. Ann spoke with a voice that looked like because the model was trained in its records by a stroke.

California, California, said that Gen AI’s deployment in several different ways allowed researchers to make improvements in neurons that may have taken much longer. A student in electrical engineering, computer science and co -author of the study, which appeared in March in Natural neuroscience.

It is an example of how to generate artificial intelligence tools – using the same basic technology that works Chatbots Zhou, like Chatgpt from Openai and Claude’s Claude or Google Meet – helps medical and scientific researchers to solve problems that may take longer to solve them. The artificial intelligence experts and office have pointed out the use of technology in medicine such as A huge high -side areaBoth in Place new drugs Or savings Better test and diagnosis.

“Artificial intelligence speeds up progress,” said Zhou. “Sometimes we imagined that the schedule will be a contract or two decades. Now that the pace was like three years.”

Zhou said that the technology that helped Ann is evidence of the concept, but it shows a way towards tools that could be more delivery and operation in the future.

Speech acceleration

The problem with the current neurons is cumin. There is a time late between when a person begins to try to speak and when the sentence is already created and heard. Zhou said that previous technology means that it had to wait until one sentence ended before the next day.

“The main penetration here is that it does not need to wait until it ends from the sentence,” he said. “Now we can actually stream the decoder whenever you intend to speak.”

The artificial limbs include a group of electrodes planted on the surface of their brain and connecting them across the cable to the bank of computers. It thinks of the control signals that the brain sends to the muscles that control speech. After Ann chose the words she intended to say, artificial intelligence reads those signals of the kinetic cortex and give them life.

To train the model, the team had to try to speak with the camels shown on the screen. Then they used data on this activity to draw a map of signs in the motor crust, using Gen AI to fill the gaps.

Zhou said that the team hopes that the hack will lead to developed and easier devices.

“We are still in the ongoing efforts to make it more accurate and less transmitted,” he said. “We are trying to build something that can be more delivery and operation.”

Using artificial intelligence to move from thinking to speech

Zhu said the team used Gen AI in several different ways. One of them was to repeat the voice of Ann before the injury. They used recordings by her injury to train a model that could produce her voice.

“She was very excited when she first heard her voice again,” said Chu.

https://www.youtube.com/watch?

The significant change was in the actual time. Compare them with tools that copy presentations or meetings when they happen.

Work based on a 2023 studies That uses artificial intelligence tools to help Ann communicate. This work still had a great delay between when I tried to speak and when the words were produced. This research significantly reduced this delay, and told the team that it felt more natural.

“She stated that she felt incarnate, and that she was her own speech,” said Zhu.