There may be new hope for stroke patients and others who are unable to communicate through traditional means. It comes in the form of the experimental new BrainGPT system, which is capable of reading users' thoughts and converting them into readable text. Paralyzed patients suffering from "lockdown" are not only unable to speak but also unable to communicate using hand gestures or head gestures. Some of them are able to move their eyes and therefore use eye-tracking communication systems, but others don't even have this ability.

In recent years, we've seen alternative devices that convert brain waves into words. However, these systems typically either require electrodes to be surgically implanted in the user's brain, or they can only allow basic yes/no communication. This is where BrainGPT could one day come into play.

A team of scientists at the University of Technology Sydney is developing the technology, which would allow users to simply wear an EEG cap connected to a computer. No eye tracker or other additional hardware is required.

The custom DeWave software used by BrainGPT was trained by recording and analyzing the electrical signals produced by the brains of 29 volunteers as they read passages of text silently.

Simply put, DeWave's AI-based algorithms learn which specific brain signals correspond to which written words and phrases. Then, when the user is not reading the text, it detects these signals and knows that the person is thinking about the corresponding word or phrase...

Currently, the system's translation score is about 40 percent of the BLEU (Bilingual Evaluation) scale, a measure of the accuracy of machine-translated text. Still, the team hopes to increase this number to around 90% as the technology develops further.

"This research pioneers the direct translation of raw brain waves into language, marking a major breakthrough in this field," said chief scientist Professor C.T. Lin. "It incorporates discrete encoding techniques into the brain-to-text translation process for the first time, introducing an innovative approach to neural decoding."

BrainGPT is demonstrated in the following video: