Thoughts can be read by recording brain activity with functional magnetic resonance imaging, as shown by scientists from the University of Texas at Austin (USA) who have applied the technique to a group of volunteers.

Although for now the test only works if the person analyzed agrees to cooperate, the authors of the research call for standards to protect the mental privacy of citizens to be developed before more sophisticated thought-reading techniques are developed.

“We hope that this technology will help people who have lost the ability to speak due to injuries or diseases such as ALS,” Jerry Tang, first author of the research, declared at a press conference. But “nobody’s brain should be decoded without their cooperation; (…) it is important to enact policies that protect mental privacy”.

The research, in which specialists in neurosciences and computing have collaborated, has been carried out in two phases. First, a computer system has been taught how language is processed in the brain. To do this, brain activity was recorded with functional magnetic resonance imaging while two men aged 23 and 36 and a woman aged 26 listened to stories for 16 hours.

To instruct the computer system, the MRI data has been supplemented with that of an artificial intelligence GPT language model, which generates sequences of words based on which one is most likely to come next.

In the second phase of the investigation, when the computer system was already trained, it was verified whether it was capable of interpreting the thoughts of the same three volunteers. They have been asked to imagine that they were going to say something but not say it, to listen to new narrations and to watch videos without sound.

While they were performing these activities, their brain activity was again recorded with fMRI. From the magnetic resonance records, the computer system has decoded the thoughts of the three volunteers.

“We have seen that the decoder can predict what the user is imagining or seeing” even if they don’t express it in words, points out Jerry Tang.

The results of the project are presented today in the journal Nature Neuroscience.

Previous research has turned brain signals into language by implanting electrodes in the brain with neurosurgery. Although some patients have partially recovered the ability to communicate, it is an invasive intervention that cannot be applied on a large scale.

The new research is the first to decode language in a non-invasive way. According to the director of the research, Alexander Huth, the advance has now been possible thanks, on the one hand, to the ability to process a greater amount of data from the brain than in the past; on the other, to the development of language models such as GPT.

The system does not reconstruct the exact sentence that the volunteers have in mind, but it does reconstruct the idea. “For example, when the user heard the phrase ‘I still don’t have my driver’s license’, the decoder predicted ‘she hasn’t started learning to drive yet,'” explains the researcher. What the system registers is “something deeper than language that becomes language,” Huth clarified at the press conference.

Although the experiments have been carried out only in English, “there is no reason to think that it would not work in other languages; [language] representations in the brain are shared across languages,” adds Huth.

Researchers have found that the language decoding system developed for each person does not work when applied to a different person. Therefore, it is not possible to read the mind of a person with MRI without their cooperation, since they must collaborate to fine-tune the system that will read their mind.

But “we don’t want to give a false sense of security by saying it’s never going to happen; this could change as technology improves; there is much more information in the brain data than we initially expected,” Tang says.

Furthermore, “although you need [users’] cooperation to achieve accuracy, you don’t need accuracy to be misused. [This technology] could have negative consequences if it is misused or misunderstood,” adds Tang, who recalls that “the polygraph was not accurate and still had negative consequences.”

Faced with the risk that brain activity records will make it possible to read the thoughts of citizens in the future, researchers from the University of Austin in Texas defend that “it is very important to regulate what brain data can be used for and what it cannot be used for.” they can be used. In this way, if in the future it is possible, there will be some bases to use them well”.