University in Texas creates AI tool that reads minds, turns thoughts into text
AUSTIN, Texas (KEYE) — The University of Texas Austin has created a system that can turn brain activity into text.
The new artificial intelligence tool could be a game-changer for some folks.
Our real hope is that this could help people who have lost the ability to communicate for some reason,” said UT computer science and neurology assistant professor Alexander Huth. “So, people who have locked-in syndrome from like brain-stem strokes.”
UT researchers have essentially created a device called a semantic decoder that can read a person’s mind by converting brain activity into a string of text.
“It’s basically a model that has been trained where it’s over 15 hours of data per person to take in their brain activity when they’re sitting in an MRI scanner to spit out the story that they’re listening to,” said UT computer science department Ph.D. student Shailee Jain.
Participants first listen to podcasts for about 16 hours in an fMRI scanner. This trains the AI to learn the person’s brain activity.
Later, they go back in to put their brain to the test.
And tell the same story without actually saying anything out loud,” Huth explained. “Since we already knew which words they were gonna say, approximately, and when those words would occur, we could then compare that to the actual words and see how well we were doing in decoding.”
Unlike other AI technology, this is non-surgical and non-invasive.
Researchers say it works, but it’s hard to call it accurate because it doesn't turn thoughts into text word-for-word.
“One example that we often use is the actual story said, ‘I didn’t even have my driver’s license yet,’ and the decoded version was, ‘She hasn’t even learned to drive yet,’” said Huth.
Researchers say they were surprised the decoder still worked even when the participants weren’t hearing spoken language.
“Even when someone is seeing a movie, the model can kind of predict word descriptions of what they’re seeing,” Jain said.
Some might be leery when it comes to mental privacy, but that’s also been addressed.
“We also tested whether it could be used on someone without collecting all this training data so, you know, could you use it on just a random person, and so far the answer to that is definitely no,” said Huth.
