Artificial intelligence should not be a mind-reading tool

On May 10, in Tokyo, Japan, participants were talking at the 7th Japan Artificial Intelligence Conference. Image source Visual China

China Youth Daily·China Youth Daily trainee reporter Zhao Tingting

Lying flat inside a functional magnetic resonance machine (fMRI), CNN reporter Donny O'Sullivan could barely hear the mechanical noise. He wore special headphones and listened to excerpts from the "Wizard of Oz" audiobook. Meanwhile, an artificial intelligence model is making predictions based on his brain's electrical activity and what he's listening to, to understand his thoughts and perceptions.

Recently, scientists at the University of Texas at Austin used fMRI technology to monitor the brain activity of three subjects while they listened to stories. They used GPT-1, OpenAI's first language model, which was built on a large database of books and websites. By analyzing this data, AI can learn the way humans speak and think.

Participating in the study was Alexander Huth, assistant professor of neuroscience and computer science at the University of Texas at Austin. During the experience, O'Sullivan discovered that artificial intelligence still has some limitations in the process of interpreting human thoughts. For example, the scene in his mind of Dorothy, the protagonist of "The Wizard of Oz" walking along the yellow brick road, was not interpreted. come out.

CNN explained that the artificial intelligence model must first analyze the brain activity of participants when they hear specific words. After learning enough knowledge, it is possible to make predictions by monitoring human brain activity. This makes many people breathe a sigh of relief: "Artificial intelligence cannot easily read our thoughts."

"We are reluctant to use the term 'mind reading' because it evokes many negative associations and lacks reality.". "Hutter said that the original intention of this research was to "help those who cannot communicate," such as autism, stroke patients or those who are unable to speak normally for other reasons. Perhaps in the future, they will not need to undergo neurosurgery to find The way to speak.

The involvement of artificial intelligence can provide more young people with opportunities to communicate and learn, which has triggered discussions on ethical and legal issues: Will it be used for interrogation? Will people's deepest secrets be revealed?

"Everyone's brain data should be protected." Jerry Tang, a PhD in computer science at the University of Texas at Austin and a participant in this study, believes that "the brain is the last line of defense for our privacy."

Huth believes that brain data breaches will not happen in the short term. "What we get is only the direction and outline of people's thinking, and we don't see the small thoughts and details clearly," he said.

Shirley Turkle, a psychology professor at MIT, told the New York Times, “Parents should tell their children that artificial intelligence has no feelings or experience. It is just a tool to help reading and thinking. Don’t use it as a replacement easily. Think, or reveal your thoughts to it.”

According to CNN, OpenAI CEO Sam Altman publicly responded to the "risk risks" that this technology may cause, saying that the development of artificial intelligence without "guardrails" may cause significant harm to the world, and he proposed legislation. These concerns should be addressed as soon as possible.

Tang told CNN that lawmakers should be careful when dealing with "mental privacy" issues and ensure that "brain data" is protected to avoid the technology being misused. The technology currently has limited scope, but don't think that will continue forever. It’s important that (humans) don’t have a false sense of security. Technology is constantly advancing, which may change our ability to decode. ”

The above is the detailed content of Artificial intelligence should not be a mind-reading tool. For more information, please follow other related articles on the PHP Chinese website!