By Mailonline’s Fiona Jackson and The Daily Mail’s Science Editor Victoria Allen
Updated May 2, 2023 11:00, May 2, 2023 12:18
- Artificial intelligence (AI) is trained to convert people’s thoughts into text
- This involved participants lying down on the MRI for 16 hours while listening to the story.
- Then when they made up a new story, the AI could read it from their brain activity
If your friend got a terrible new haircut, it’s probably a good idea to keep your opinion to yourself, but you may soon run out of options.
That’s because scientists at the University of Texas at Austin have trained an artificial intelligence (AI) to read people’s minds and translate their innermost thoughts into text.
Three study participants listened while lying in an MRI machine while an AI “decoder” analyzed their brain activity.
Next, they were asked to read another story or compose their own, and the decoder was able to convert the MRI data into text in real time.
This breakthrough raises concerns about “mental privacy” as it could be the first step in being able to eavesdrop on other people’s thoughts.
Lead author Jerry Tan said he was not trying to give a “false sense of security” that the technology might not be able to eavesdrop on people’s thoughts in the future, and is currently being “abused”. said it was possible.
However, he said:
“And to avoid that, we want to spend a lot of time moving forward.
“Right now, technology is in a very nascent state, but I think it’s important to be proactive and get a head start. It’s about enacting policies that give you rights to your thoughts.Brain data.
“We want these to be useful and only when people want to use them.”
In fact, the technology still poses no threat to privacy, as it took 16 hours of training before the AI could successfully interpret participants’ thoughts.
Even after that, I couldn’t exactly recreate the story they were reading or making, I just caught the gist of it.
For example, someone hearing the statement “I don’t have a driver’s license yet” was translated as “She hasn’t started learning to drive yet.”
Participants were also able to “sabotage” the technology and prevent their thoughts from being read, using methods such as enumerating animal names in their heads.
The study, published in Nature Neuroscience, reveals a decoder that uses language processing technology similar to ChatGPT, an AI chatbot.
ChatGPT has been trained with large amounts of text data from the internet and can generate human-like text in response to specific prompts.
The brain has its own “alphabet” made up of 42 different elements that refer to specific concepts such as size, color and position, all of which combine to form complex thoughts.
Each “letter” is processed by a different part of the brain, so by combining all the different parts you can read people’s minds.
The US-based team did this by recording MRI data from three parts of the brain associated with natural language while participants listened to a 16-hour podcast.
The three brain regions analyzed were the prefrontal cortex network, the classical language network, and the parietal-temporal-occipital association network.
The algorithm was then given a scan and compared the pattern of speech with the pattern of recorded brain activity.
I was able to pick up what the person was thinking in half the time.
That meant it generated text that closely and sometimes exactly matched what people were hearing, and it solved this using only brain activity.
The technology could also interpret what people were seeing when watching silent films, or the thoughts they imagined when telling stories.
Unlike other mind-reading technologies, it works when you think of any word, not just words in a setlist, but struggles with pronouns like “he” and “me.”
It detects activity in the language-forming regions of the brain rather than imagining how someone moves their mouth to form a particular word.
“We were a little shocked that this actually worked,” said Dr. Alexander Hoet, senior author of the study.
“This is a problem I have been working on for 15 years.”
Researchers say this revolutionary method can help people who are mentally aware but unable to speak, such as stroke victims and patients with motor neuron disease.
Silicon Valley is very interested in mind-reading technology that could one day allow people to type just thinking about the words they want to say.
Elon Musk’s company, Neuralink, is working on brain implants that allow direct communication with computers.
But the new technology, which is relatively rare in the field, reads thoughts without the use of any brain implants, so no surgery is required.
Today, it requires bulky and expensive MRI machines, but in the future, we may wear patches on our heads that use waves of light to penetrate the brain and provide information about blood flow.
This could potentially detect the thoughts of people moving around.
Dr. Huth added:
Addressing concerns that technology could be used by someone without their knowledge, such as by authoritarian regimes interrogating political prisoners, or by employers spying on their employees, the researchers said the system could be used by individuals to It states that only after training the patterns can it read an individual’s thoughts. , so sneakily could not apply to someone.
Dr. Huth said: “If people don’t want to decipher something from their brain, they can control it using only their cognition. And then it all falls apart.”
