PARIS: Scientists in the United States have developed a system that uses brain scans and artificial intelligence (AI) to “decode” the meaning behind a person’s thoughts. The technology was described as a step towards mind reading but also raised concerns about mental privacy.
The language decoder, developed by neuroscientists at the University of Texas at Austin, aims to help people who have lost the ability to communicate.
It works by mapping out how words, phrases, and meanings prompt responses in the regions of the brain known to process language. This data is then fed into a neural network language model that uses GPT-1, a predecessor of the AI technology used in the popular ChatGPT.
To test the system’s accuracy, three participants spent a total of 16 hours inside a functional magnetic resonance imaging (fMRI) scanner listening to spoken narrative stories, such as podcasts.
The decoder was able to “recover the gist” of what the participants were hearing, even when they thought up their own stories or watched silent movies.
While previous research has found that a brain implant can enable those people who can type to spell out words or even sentences or no longer speak, the Texas team’s decoder “works at a very different level,” according to co-author Alexander Huth. “Our system works at the level of ideas, of meaning, of semantics.”
Technology raises questions about mental privacy
However, the study’s authors acknowledged that the technology raises questions about mental privacy. The researchers ran tests showing that their decoder device could not be used on anyone who had not allowed the decoder to be trained on their mental activity over long hours inside the fMRI scanner.
But David Rodriguez-Arias Vailhen, a bioethics professor at Granada University in Spain not involved in the research, warned that the technology could possibly be used against people’s will, such as when they are sleeping.
The researchers also called for regulations to protect mental privacy. “So far, our mind has been the guardian of our privacy,” said Rodriguez-Arias Vailhen. “This discovery may be a first step towards compromising that freedom in the future.”
The study, which is the first to be able to reconstruct language without an invasive brain implant, was published in the journal Nature Neuroscience. The team hopes to speed up the decoding process so that it can be done in real-time.
