Japanese scientists have achieved a significant milestone in AI accuracy by successfully reconstructing images solely from people's brain activity with over 75% precision, a first of its kind accomplishment. Typically, image reconstruction from brain activity is possible only when subjects view the images with their own eyes or when specific types of images, like faces or letters, are predefined. However, researchers from Japan's National Institutes for Quantum Science and Technology (QST) have demonstrated the potential to accurately reconstruct complex images with AI, primarily relying on a person's thoughts.

‍

In their study published in Neural Networks, the scientists recorded brain activity from subjects who viewed 1,200 different images while inside a functional magnetic resonance imaging (fMRI) machine. They created extensive "score charts" with approximately 6.13 million factors, such as color, shape, and texture, enabling the AI to recognize the images. Additionally, they developed a neural signal translator program that linked brain activity to the scoring chart and generated new scoring charts for new brain activity inputs. Using this method, they achieved a remarkable 75.6% accuracy rate in reconstructing original images, marking a significant improvement over previous efforts that reached only 50.4% accuracy. This breakthrough could have profound implications, particularly in the medical field, potentially aiding individuals who have lost their ability to verbally communicate and opening doors to innovative forms of communication.