[Reading Science] AI Applied to Neuroengineering... Reading Your Thoughts

Brain Blood Flow → Successful Translation into Language
Recent Innovations in Neuroengineering Using AI Continue
Assistive Activities for Disabled People at 'Commercialization' Stage
Media Innovation Enables Dream and Memory Storage and Playback

Artificial intelligence (AI) is revolutionizing brain engineering. By applying AI to technologies that scan the brain and read 'thoughts,' faster and more precise semantic interpretation has become possible. Currently, assisting the movements of individuals with complete paralysis is considered 'basic.' Brain-computer interface (BCI) technology, which connects humans and computers, has diverse medical and industrial applications and is attracting global attention.

Brain image. Not related to the article. Photo source=YouTube screen capture

Translating brain blood flow into language

On the 1st (local time), a research team at the University of Texas at Austin published a representative study in the international journal Nature Neuroscience on non-invasive brainwave analysis using functional magnetic resonance imaging (fMRI) and AI. Most existing brain scanning and verbalization technologies were invasive, involving implanting chips in the cerebral cortex to measure electrical changes and convert them into speech. In contrast, the team succeeded in translating brain changes into language non-invasively through fMRI. The fMRI used by the team measures changes in brain blood flow. When a person performs an action such as speaking, blood flow increases in specific brain regions responsible for that action, which can be observed in detail.

The team achieved innovative results by employing AI to analyze the fMRI imaging data. The AI algorithm used is called large language models (LLMs). They utilized an early version of the generative AI 'ChatGPT,' which has recently become a focus of attention. It was trained to predict the next word based on preceding words when a person speaks.

The researchers had three volunteers listen to podcasts for 15 hours each while measuring changes in brain blood flow using fMRI. This procedure was to train the LLMs on which brain areas increase blood flow when the volunteers heard specific information. Through this, the team was able to create a map showing how people's brains respond when hearing particular words and sentences.

As a result, the team succeeded in verbalizing the information input to the brain while volunteers watched short animations. They were also able to grasp the gist when a volunteer thought of a sentence. For example, when a volunteer mentally thought, "I don't have a driver's license yet," the AI generated a sentence like "I haven't learned to drive yet."

Thought stealing possible

However, limitations were also identified. When volunteers thought of something else, verbalizing brain activity was impossible. Accuracy improvements are needed, as many of the words or sentences derived from the analysis were inaccurate. Moreover, even when listening to or watching the same story, brain changes varied individually, making compatibility impossible. Another limitation was that only those who consented to have their thoughts read could be interpreted. Above all, the need to use the massive fMRI equipment found only in hospitals for extended periods means commercialization is still far off.

Nonetheless, scientists are somewhat unsettled by the emergence of this innovative non-invasive brain scanning technology. Ethical concerns are foremost. Reading others' thoughts without permission infringes on privacy and could be exploited for criminal purposes. If investigative authorities or courts prematurely attempt such 'mind reading' during investigations or trials, it could lead to unintended consequences. For example, in the team's experiment, the AI described a volunteer's thought of "I just got out of the car" as "I had to push her out of the car." Gabriel Lazaro-Munoz, a bioethics professor at Harvard Medical School, said, "It's not panic-inducing, but the technology is more sophisticated and non-invasive than expected, bringing it closer to a boundary," adding, "This research is a wake-up call for policymakers and the public."

Adina Roskies, a philosophy of science professor at Dartmouth College, also said, "This technology is difficult to use and too inaccurate to worry about yet," but expressed concern that "(Inaccurate analysis of brain scan results) can make a significant difference in legal lawsuits. The technology should not be used when it shouldn't be." Another issue exists. Francisco Pereira, a researcher at the U.S. National Institute of Mental Health, pointed out, "It will be difficult for AI to translate jokes, metaphors, or sarcasm," emphasizing, "There is a big difference between AI simply predicting the next word and the brain combining words for language use."

A patient who became able to speak by implanting a neural chip in the brain to translate brain waves into language. Photo by YouTube capture

Brain engineering revolution... potential innovation across various fields

However, experts are closely monitoring the significant impact that could arise if this 'mind reading' technology advances. Technologies that analyze brainwaves to convert commands for assisting the activities of people with paralysis, enabling them to freely operate robotic arms, electric devices, and walking aids, have already entered the commercialization stage. Furthermore, various medical and industrial applications are possible. For example, the technology that reads brainwaves can be used in reverse to input information into the brain through electrical currents or magnetic fields. This could enable visually impaired individuals to see and bring revolutionary changes to virtual reality (VR) devices, TVs, smartphones, and wearable devices.

If it becomes possible to read others' thoughts and emotions accurately from a distance, new communication methods and tools will emerge. Precise remote control of cars, airplanes, and robots from afar will also be feasible. Recently, Jack Gallant, a professor at the University of California, Berkeley, attracted attention by demonstrating a technology that reconstructs videos watched by test subjects by analyzing their brains with fMRI. Such technologies may bring significant changes to human life in the future, such as storing and replaying dreams or revisiting memories.

Kim Hyung-min, director of the Bionics Research Center at the Korea Institute of Science and Technology (KIST), explained, "There are many challenges, such as equipment development, to accurately capture and interpret brain activity in real time," adding, "The brain sensing technique used in this study, fMRI, has excellent spatial resolution but poor temporal resolution." Director Kim also emphasized, "South Korea is investing heavily in research and development (R&D) of brain scanning technology and is at a world-class level in some fields. Since it has diverse applications, continuous support is needed to improve research equipment and software, as well as medical science research."