A new AI is capable of reading our minds and creating the images we are thinking
Japanese researchers have developed a system for creating images from brain activity using machine learning.
The latest advances in generative Artificial Intelligence can make our imagination come true. Even if we don’t know how to draw, projects like Midjourney allow us to create drawings or photographs based on our explanations.
However, it is not as simple as that. The key to getting a good AI-created image is precisely, and the best results are achieved if we write with the limitations of the AI in mind and the way it interprets our commands. But what if we didn’t have to write anything for an AI to draw what we’re thinking?
AI that reads a mind
That is what researchers at Osaka University have achieved after a study (PDF) that has lasted more than a decade but that has come out, curiously just when generative AI is on everyone’s lips.
The researchers acknowledge that lately, generative models are having success in recreating realistic images, as long as they are based on sentences that are very faithful to what the user is thinking, which is a challenge. To solve this problem, they propose a new diffusion model capable of reconstructing images from the user’s brain activity.
In the experiments, the researchers obtained brain activity data using functional magnetic resonance imaging (fMRI), capable of generating images of activity in various brain regions. The system they created was capable of interpreting that data to create high-resolution images of what the volunteers were thinking.
Specifically, the volunteers were shown various images and the AI was able to “interpret” their brain activity and create an image very similar to the original one they had seen with their own eyes. The results speak for themselves, and although the resulting images are not exactly the same as the original ones, the general idea is maintained, such as the perspective, the colors, and the position of the elements. The results were different for each test subject since each person “thinks” differently, but they all had elements in common.
That is not the first study to try to “see” the images we think of; The achievement of these researchers has been in the use of a diffusion model that has allowed obtaining these results so faithful to the original. To do this, they had to map specific components of the model with specific regions of the brain.
Related Post
