Over the past years, scientists have been developing various technologies that aim to read human thoughts, decipher brain puzzles, and discover what a person is thinking without the need for speaking, in an attempt to break through in the field of treating numerous diseases and successfully communicate with individuals who are speech-impaired.

These technologies vary from brain imaging devices to other advanced methods, even to the point of implanting electronic chips into human brains to decode thoughts.

Last September, Neuralink, a biotech startup founded by entrepreneur Elon Musk, began recruiting participants for its first human clinical trial after obtaining approval from the U.S. Food and Drug Administration (FDA) to conduct it.

Neuralink's trial focuses on surgically implanting a chip into a part of the brain of quadriplegic patients in order to record brain signals and send them to a computer application, with the primary goal of "allowing individuals to control a computer cursor or keyboard using only their thoughts."

Musk is working on achieving Neuralink's ambition to use implants to connect the human brain to computers within five years, although the company so far has tested it only on animals.

Non-invasive System

In a new effort to control the brain, researchers from the Center for Artificial Intelligence at the University of Technology in Sydney, Australia, developed a non-invasive, portable system, the first of its kind in the world, that can decode silent thoughts in the brain and convert them into text. The results were presented at the "Neural Information Processing Systems" conference held in mid-December in Montreal, Canada.

The researchers explained that this technology could help communicate with people who are unable to speak due to illness or injury, including stroke or paralysis patients and might also enable seamless communication between humans and machines, such as operating an electronic arm or robot.

This attempt is different from previous technologies for translating brain signals into language, which either require surgical procedures to implant electrical electrodes in the brain, like Neuralink's device, or scanning using Magnetic Resonance Imaging (MRI), which is bulky, expensive, and difficult to use in everyday life.

Speaking to Al-Sharq Al-Awsat, Professor C.T. Lin, the lead researcher of the study and the director of the Center for Artificial Intelligence at the University of Technology in Sydney, stated that the new technology primarily relies on individuals wearing a cap-like device on the head, recording the brain’s electrical activity through the scalp using an Electroencephalogram (EEG).

An EEG is a test that measures electric activity in the brain using small metal discs (electrodes) connected to the scalp. The activity is displayed as wavy lines on an EEG recording and is one of the primary tests for diagnosing epilepsy. An EEG can also play a role in diagnosing other brain disorders.

While EEGs are not novel, the team's innovation lies in the artificial intelligence model called DeWave. DeWave analyzes the EEG results into distinct units that capture specific characteristics and patterns of the human brain and translates those signals into words and sentences by learning from large amounts of EEG data.

Lin explains, "By learning from brain signal data associated with speech in EEG data, deep learning models can decode non-invasive (non-interventional) human brain signals and convert them into language, as well as transforming an EEG wave signal into a linguistic sentence without prior processing or eye-tracking assistance through deep learning models."

According to the researchers, these methods also struggle to convert brain signals into speech texts without additional assistive tools such as eye-tracking, which can limit the practical application of these systems. In contrast, the new technology can be used either with or without eye tracking.

The team tested the new technique on 29 participants, using EEG signals received through a cap instead of electrodes implanted in the brain.

Advanced Performance

Regarding the translation of EEGs in terms of converting signals into texts, the study reported advanced performance that exceeds previous standards; researchers achieved results of 41.35 on the BLEU scale a measure of linguistic accuracy used to evaluate the quality of machine translation. The score is calculated by comparing the translated text with a set of high-quality reference translations, taking into account the number of matching words and their order.

The researchers stated that their new model achieves meaningful results by aligning keywords and constructing similar sentence structures. They hope to see this improvement evolve to a level comparable to traditional language translation or speech recognition programs, which are approaching 90% accuracy.

As for the practical applications useful for the research results, Lin pointed out that they could contribute to transforming the thoughts of patients unable to speak into language, as well as enhance the control of robots and smooth communication with them.

He saw that "this research represents a pioneering effort in directly translating raw EEG brainwaves into language, marking a significant advance in this field; it presents an innovative approach to neural decoding, and this integration with large language models also opens new horizons in neuroscience and artificial intelligence." He noted that his team will continue their research to improve the accuracy of brainwave decoding, collect more data, and phrases related to everyday conversation or robot control.