Scientist turns people’s mental images into text using ‘mind-captioning’ technology

A Japanese scientist has developed a groundbreaking technique that uses brain scans and artificial intelligence to convert a person’s mental images into clear, descriptive sentences — a major leap toward true “mind-reading” technology.

The method, created by Tomoyasu Horikawa and published November 5 in Science Advances, moves beyond earlier research that translated internal speech into text. Horikawa’s new approach, called mind-captioning, can interpret complex mental visuals — objects, scenes, actions, and relationships — and turn them into coherent language.

Horikawa, a researcher at NTT’s Communication Science Laboratories near Tokyo, scanned the brain activity of six adults while they watched 2,180 soundless video clips featuring varied scenes. Large language models converted the video captions into numerical sequences, and Horikawa trained simpler AI “decoders” to match these sequences with the participants’ brain activity.

When tested with new, previously unseen videos, the decoders successfully interpreted the brain scans, enabling an AI algorithm to generate descriptive text that closely matched what participants saw or recalled. As the system learned, the captions became increasingly accurate — even producing text in English, despite participants not being native English speakers.

Experts say the study marks a major step forward.
“This is a legitimate form of brain-reading,” said Marcello Ienca, professor of AI ethics at the Technical University of Munich and president-elect of the International Neuroethics Society.

Potential for Transformative Healthcare Uses

Because the method does not rely on language-related brain regions, it could one day assist people with aphasia, ALS, or non-verbal autism. Psychologist Scott Barry Kaufman noted that the technology could offer “profound interventions” for individuals who cannot communicate verbally — provided it is used ethically and with consent.

Privacy Concerns: ‘The Ultimate Challenge’

The ability to decode thoughts raises serious ethical issues. The study warns that such technology could expose private mental content, including dreams or unspoken intentions.
“If this ever reaches consumer use, it becomes the ultimate privacy challenge,” Ienca said.

Experts emphasize the need for strict regulations, user-controlled “unlock” mechanisms, and treating neural data as highly sensitive. Social scientist Łukasz Szoszkiewicz added that mental privacy protections must evolve as quickly as the technology itself.

Not Ready for Real-World Use — Yet

Horikawa stressed that the method still requires extensive data from active participants and works best with typical, predictable scenes. It cannot decode unusual or unexpected mental images and remains far from practical mind-reading.

“While people may fear risks to mental privacy,” Horikawa said, “the current approach cannot easily read a person’s private thoughts.”

Source: CNN

Voice of Osiz

The rapid progress in AI-driven neural decoding marks a remarkable shift in how human cognition can be interpreted through technology. This breakthrough demonstrates the growing potential of AI to become a powerful communication bridge for individuals with speech and neurological challenges. At the same time, it reinforces the need for strong ethical frameworks to safeguard mental privacy as these capabilities evolve. Innovations like this inspire our continued focus on building responsible, human-centric AI solutions that respect both progress and protection.