Shulou(Shulou.com)11/24 Report--

This article comes from the official account of Wechat: SF Chinese (ID:kexuejiaodian), author: SF

Scientists have used functional magnetic resonance imaging and large language models to achieve a non-invasive brain-computer interface, according to a paper published in the journal Nature Neuroscience. We know that the large language model created the epoch-making ChatGPT, so now let's see how it makes the leap from intrusive to non-intrusive in the field of brain-computer interface.

Wen Jing | Wen Jing

Functional magnetic resonance imaging (fMRI) is a non-invasive method for measuring brain activity. The principle is to stimulate specific senses, activate nerves in the corresponding parts of the cerebral cortex, and display them by magnetic resonance images. Recently, scientists have combined fMRI with a large language model (LLM: an artificial intelligence algorithm that is also the technical basis of ChatGPT) to interpret the thinking activities of the human brain.

Breakthroughs in brain-computer interfaces most of the existing brain-computer interfaces are invasive and require brain implants to monitor the activity of a human cortex and predict the language that the person being monitored may output.

Recently, computer scientists at the University of Texas at Austin used functional magnetic resonance imaging and large language models to describe how the brains of 16 volunteers responded to different words and short sentences. a method for constructing a non-invasive language brain-computer interface. The results were published in Nature Neuroscience (Nature Neuroscience), a sub-journal of the well-known international academic journal Nature.

The scientists asked volunteers to lie on functional magnetic resonance imaging scanners to record everyone's brain activity while they listened to or watched the same story. After the experiment, based on a large language model, the scientists coded a process that reflects everyone's brain activity by mapping the blood flow of the volunteers' brains to the details of the story.

This process can be thought of as a decoder. For people who cannot communicate by voice, this kind of thought decoder gives them a convenient opportunity to communicate with the outside world. Under certain circumstances, the decoder will be more efficient than the already popular voice control.

Research shows that the decoder records each person's response to different words and phrases more accurately, but the response itself contains the subjective thoughts of the person being monitored. Therefore, the results produced by the decoder will vary from person to person.

At the same time, if another story is imagined in the subject's brain during the experiment, the accuracy of the results of the decoder will be greatly reduced. In other words, at present, this artificial intelligence technology is easy to be "deceived".

The interpretation of thinking is that there are still many immature aspects of this technology, coupled with the limitations of equipment and cost, it can not be widely used in individuals in a short time. However, the development of this complex non-invasive technology has exceeded many expectations, which is a wake-up call for both policy makers and the public.

Some experts worry that the inaccuracy of the results of the decoder will affect judicial judgment. Adina Roskis, a scientific philosopher at Dartmouth College in the United States, said: "the difference between them [the semantics that can be mistaken by current decoders] is very obvious and can lead to a completely different direction in legal cases."

In addition, more people have expressed concerns about privacy and discrimination about the technology. The authors Jerry Tang and Alexander Huth also called on policy makers to actively address the issues under which the technology can and cannot be used.

We know that for people with mental disorders such as obsessive-compulsive disorder, sometimes there are random, unintentional invasive thoughts in the brain, which is an obsessive-compulsive behavior that the patient does not want but is difficult to control. Therefore, experts worry that in the process of mental interpretation of these people, if the person being monitored has an intrusive idea, the decoder will implement the idea, resulting in irreparable consequences.

References:

Https://www.nature.com/articles/d41586-023-01486-z

Https://www.nature.com/articles/s41593-023-01304-9

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.