'Mind reading' cap turns thoughts into text during trials

Year of AI closes with breakthrough in non-invasive decoding of thoughts

Researchers in Australia manage to turn brain activity into text

Researchers in Australia have made a breakthrough in the field of translating electroencephalogram (EEG) waves directly into text.

The technology has been developed by a team at the University of Technology in Sydney, Australia.

In tests of the device, participants were told to read silently passages of text while an electroencephalogram (EEG) recorded their electrical brain activity.

An artificial intelligence model called DeWave was then used to decipher the thoughts into written text. The model had an accuracy rate of between 40 and 60 per cent.

"This research represents a pioneering effort in translating raw EEG waves directly into language, marking a significant breakthrough in the field," said CT Lin, a professor at the University of Technology Sydney.

"It is the first to incorporate discrete encoding techniques in the brain-to-text translation process, introducing an innovative approach to neural decoding."

Professor Lin also raised the prospect of integration with large language models, and stated that this would "open new frontiers in neuroscience and AI".

It's a prospect that sounds faintly dystopian, although it should provide conspiracy theorists with plenty of podcast material. More positively, the technology could could help people who are unable to speak due to illness or injury.

The accuracy of this new, non-invasive brain-computer interface recorded lower accuracy rates that existing interfaces which use MRI scans or more invasive approaches such as nose or skull implants. Perhaps unsurprisingly, Elon Musk is working in this field with his startup Neuralink which aims to create a "generalised brain interface." Neuralink is currently recruiting for clinical trials involving the implantation of a chip in the brain by a surgical robot.

Invasive methods may be more accurate in the most recent trials, but the team behind the non-invasive method believes it has potential to reach a level closer to 90 per cent.

"Despite the challenges, our model yields meaningful results, aligning keywords and forming similar sentence structures," said Yiqun Duan, one of the scientists behind the study.

The research was presented at the NeurIPS conference in New Orleans on 12 December.