Scientists taught a neural network to draw pictures according to thoughts

Scientists from the Japanese University of Osaka taught the Dall-E 2 neural network, created by the American company OpenAI, to draw according to a person’s thoughts. Artificial intelligence can read thoughts with an accuracy of up to 80%, the study was published in the electronic archive of scientific articles on biology BioRxiv.

Four volunteers participated in the experiment. They were alternately shown a set of pictures and at the same time their brain activity was recorded using MRI. After that, the results of magnetic resonance imaging were loaded into the neural network. And based on this data, artificial intelligence drew what he “saw”.

Despite the fact that some details turned out to be blurry, the images of the neural network turned out to be quite similar to the original.

After training, scientists managed to ensure that artificial intelligence reproduced pictures with an accuracy of 80%.

“We demonstrate that our simple framework can reconstruct high-resolution images from brain activity with high semantic fidelity. Unlike previous image reconstruction studies, our method does not require fine-tuning of complex deep learning models,” the study authors write.

Earlier, Russian and Scottish psychologists created a neural network that can slow down aging by changing hobbies. Artificial intelligence has revealed a formula for the dependence of human aging on hobbies, taking into account his personal characteristics. It is enough to enter the data of a particular person, and the neural network will recommend what to do to prolong life.