Scientists from Meta have created a model AI that can transform brain activity into written sentences, without the need for implants. This discovery could revolutionize brain-computers interfaces and could help people with communication difficulties.
A new model AI, capable of “reading thoughts”. Photo: Archive
According to Live Science, the discovery shows that this method works by scanning neuronal activity with non -invasive techniques, thus opening the way to new possibilities for people with communication impairments.
In a first study, the researchers used magnetoencephalography (Meg), a method that measures the magnetic fields generated by the electrical activity of the brain, to analyze how participants type sentences.
By training a model you have on this data, scientists have managed to decode brain signals and rebuild the sentences written by volunteers. The accuracy obtained was 68%, and the letters used frequently were recognized with more accuracy than rare, such as “Z” or “K”.
The second study focused on how the brain generates language. The researchers registered 1,000 instantaneous meg per second to capture each stage of the sentence formation process. Thus, they found that the process begins with formulating the sense of sentence, followed by decomposition in words, syllables and, finally, letters.
“These results confirm the long -term predictions that language production requires a hierarchical decomposition of the significance of the sentence in progressively smaller units that eventually control the motor actions”, The authors of the study explained.
Another important aspect discovered by the team was the existence of a “dynamic neuronal code ”which prevents interference between successive letters and words. This code constantly adjusts the place where each information is represented in the brain, which allows an effective organization of language and keeping information for a longer period of time.
What impact could the technology have?
This technology could be an important step towards developing portable devices to facilitate the communication of people with brain lesions. However, for now, the equipment used are too bulky to be used outside the laboratories.
The computational neurologist Alexander Huth, who was not involved in research, believes that it is a significant progress: “They certainly do everything we can do with current technology in terms of what they can extract from these signals”.
In the future, the improvements of Meg technology could make it possible to use this system in everyday life, opening new possibilities for human communication assisted.
