In the results of the study, the scientists showed several images. The left image is shown to the study participants, and the right is the image created by the artificial intelligence after analyzing the brain scans taken at the time of monitoring.

the results of the experiment

Technology makes a lot of mistakes, but it already knows how to imitate the general properties, colors and shapes of objects in pictures. The authors of the study believe that in just ten years, this improvement will be available to anyone and everywhere.

It can help disabled patients improve what they see and think. Ideally, people wouldn’t even need to use their cell phones to communicate. we can only think
– says Zhao Chen, head of the study.

Examples of reading brain waves and converting them into images / Photo: Zijiao Chen

According to Associate Professor Helen Zhou, the images the system generates are thematically consistent with the original image, but not photographic replicas—perhaps because each person’s perception of reality is different. “When you look at the grass, I may think of the mountains, you the flowers, and others the river.”– says Zhou, who explains he’s human imagination may cause differences in image output. But the differences may also be the result of artificial intelligence working to produce different images from the same set of inputs.

Currently, the experiment requires the model to be trained on each participant’s brain activity for about 20 hours before it can generate images from the fMRI data.