Technology News | Scientific ecology

Artificial intelligence deciphered brain signals and makes words out of them

Artificial intelligence deciphered brain signals and makes words out of them

Human thoughts can be read with the help of a new development of scientists. The unique device is endowed with artificial intelligence. The device can read brain signals, decrypt them, and make words and sentences from them. The experiment conducted by scientists demonstrated the accuracy of decoding signals, it amounted to 97 percent. Now the world is on the verge of introducing powerful devices that are based on artificial intelligence.

A few years ago, the inhabitants of the planet could not imagine that such interfaces would appear in the future and they would develop rapidly, helping people and simplifying their lives. Researchers at the University of California at San Francisco are reporting new projects related to artificial intelligence. A group of scientists led by neurosurgeon Edward Chang used a new method for decoding an electrocorticogram.


The electrical impulses arising during the cortical activity were recorded. The passing activity was recorded by electrodes. These electrons were implanted into the brain of patients. Four people took part in the experiment, all participants had the same diagnosis - epilepsy.

They agreed to the use of implants, which monitored their health, recorded the number of seizures and controlled speech. Scientists suggested that the experiment participants first read and then speak out a few sentences out loud. While they were speaking out loud, electrodes recorded brain activity.

After that, the obtained data were redirected to the neural network, and she analyzed the patterns of brain activity, determining speech signatures, such as the pronunciation of vowels, consonants. Then another neural network decrypted this data, made up of 30-50 sentences that patients made earlier out loud.


Decryption was carried out only on the signatures of words. When the experiment ended, the decryption system showed the detected errors. There was only 3 percent of them, and all of them were associated with only one patient. Scientists attribute these errors to incorrect speech perception, noting the high efficiency of the new method of using artificial intelligence.