‘Mind captioning’ uses MRI scans and alogorithms to tell what people are thinking about
Scientists have created a form of digital mind-reading called ‘mind captioning,’ which uses MRI scans to determine what people are thinking.
Researchers have trained computers to interpret the scans and relay what you’re viewing – or even recalling – from your brain activity in remarkably precise sentences.
The technique could assist people with speech impairments, such as stroke survivors, to communicate more easily.
However, concerns are mounting that AI is edging closer to exposing private thoughts, feelings, and medical information that could potentially be exploited for monitoring, coercion, or other sinister aims.
The discovery, highlighted by Scientific American and detailed in a recent Science Advances paper, employed non-invasive MRI scans to analyse how the brain processes meaning whilst participants viewed brief video clips, reports the Daily Star.
A sophisticated two-stage AI system then converted those neural patterns into language describing the scene with remarkable precision.
You have a seventh sense, wolves help to save the planet, Google has plans for Crab Island, and the biggest spider web ever found – all this and more in our latest weird science newsletter
Alex Huth, a computational neuroscientist at the University of California, Berkeley, says the computer predicts what a person is looking at “with a lot of detail”.
He said: “This is hard to do. It’s surprising you can get that much detail.”
In one trial, a volunteer observed someone leaping from a waterfall. The AI’s predictions progressed from basic word groupings to a remarkably accurate sentence: “a person jumps over a deep water fall on a mountain ridge.”
The technology wasn’t limited to real-time observation – when volunteers remembered footage they’d previously watched, the system could produce corresponding descriptions from their recalled memories.
Six participants took part in the trial, with researchers describing the degree of accuracy as “surprising” for a technique that doesn’t require invasive procedures. It represents a significant advancement on previous attempts that could only produce individual keywords rather than complete, contextually accurate descriptions.
Currently, the AI requires your active participation, substantial personal training information, and a large MRI scanner to function, and there’s no indication yet it can interpret your spontaneous internal thoughts.
Nevertheless, the ethical implications are significant – as the technology advances, concerns surrounding mental privacy, informed consent and potential abuse will inevitably intensify.
