A new technology powered by artificial intelligence can read minds. To be more precise: it can read your brain waves, identify what you are thinking and translate it into a very high resolution image. It is so accurate that it can even capture shapes and colors. Researchers at the Graduate School of Frontier Biosciences at Osaka University in Japan, used Stable Diffusionone of the most famous image-generative AI systems.
The study published last December consisted, broadly speaking, of showing various images to a group of volunteers. These images generated stimuli in brain activity, which scientists could capture thanks to scans by MRI. The information obtained was translated through semantic decoders and, finally, it was used as input to generate a final image with stable diffusion.
The research team claims that It is the first time that one is used. tool of generative artificial intelligence as Stable Diffusionin an experiment to “read” the mind and generate visual recreations of brain activity.
How does AI that can “read” minds work?
The Osaka University team included in the publication of their study a comparison between the images they showed the volunteers and those finally generated with stable diffusion. The resulting images are not exactly the same as the originals, but they reproduce most of the essential elements. For example, shapes, colors and perspective. final pictures too varied among the participants, but this has to do with the fact that each person has their own way of interpreting reality.
The Japanese team managed to “read” the minds of the volunteers not only thanks to the artificial intelligence of stable diffusion. The study involved other key processes to transform brain activity into data, so that it could be read and used by this tool. stable diffusion It does have the merit of having simplified the creation of graphic representations of extremely high resolution and semantic fidelity.
The researchers acknowledged that some previous studies had succeeded in generating high-resolution images. They explained, however, that this was possible after a whole process of «training and adjustment» of generative models, a highly complex task.
Radboud University in the Netherlands, for example, published in 2022 that they had managed to train an AI model to “read” the mind. The method was similar. They showed various participants photos of people, generated information from MRI scans, and generated new images with their AI model.
How does the artificial intelligence of Stable Diffusion work?
Artificial intelligence to “read” the mind of Radboud University had a high match on the gender, age and pose of the people in the photos. But then again, it required the “training” process that the Japanese team was avoided thanks to stable diffusionplatform created by the company Stability AI.
stable diffusion Design images from text. It is open source, unlike similar tools like DALL-E. It also has an automatic learning logic: it offers increasingly accurate results as it is used.
A trio of artists sued the company last January for violating the copyrights of “millions of artists.” Getty Imagess, the US photo agency, followed suit in February. He assured that Stable Diffusion used millions of photos without permission from the agency to generate your images.