Functional magnetic resonance imaging (fMRI) makes brain activity visible
According to their own statements, US scientists have succeeded in using brain scanners and artificial intelligence to roughly capture what people think – and thus come a whole lot closer to reading minds. The researchers at the University of Texas at Austin use a speech decoder that works ‘on a completely different level’ than current devices, said Alexander Huth, a neuroscientist at the US university and co-author of a study published in the journal Nature Neuroscience on the subject on Monday at a press conference.
According to their own statements, US scientists have succeeded in using brain scanners and artificial intelligence to roughly capture what people think – and thus come a whole lot closer to reading minds. The researchers at the University of Texas at Austin use a speech decoder that works “on a completely different level” than current devices, said Alexander Huth, a neuroscientist at the US university and co-author of a study published in the journal Nature Neuroscience on the subject on Monday at a press conference.
In the future, speech decoders could enable people who have lost their ability to speak with a brain implant to spell words and even whole sentences. The “Brain-Computer Interfaces” focus on the part of the brain that controls the mouth when trying to form words.
“Our system works (…) on the level of ideas, semantics, meaning,” said Huth. It is the first system that can reconstruct speech without a brain implant.
For the study, three subjects spent a total of 16 hours in a functional magnetic resonance imaging (fMRI) scanner – a device that maps physiological functions in the body using magnetic resonance imaging – and listened to spoken podcasts. In this way, the researchers were able to record the reactions that words, sentences and meanings triggered in the brain regions responsible for processing language.
They fed this data into a neural network language model using GPT-1, the predecessor of the AI technology that also uses ChatGPT. The AI model was trained to predict how the brain would respond to the speech it heard, and then narrow down the options and match the most likely sentences. For example, if the study participant heard the sentence “I don’t have a driver’s license yet”, the model replied: “She hasn’t even started learning to drive”.
Huth explained that even when the subjects made up their own stories, the decoder was able to capture the essentials. “This shows that we are decoding something deeper than speech and then turning it into speech,” the researcher said.
Bioethics professor at the University of Grenada in Spain, David Rodriguez-Arias Vailhen, who was not involved in the experiment, confirmed that the system goes beyond what had been achieved with previous brain-computer interfaces.
“This brings us closer to a future where machines are able to read and transcribe minds,” he said, warning that this could potentially happen against a human’s will, such as when they are asleep.
HOME PAGE