A team of four researchers from Japan-based Kyoto University are developing an artificial intelligence system to "decode" a person's thoughts based on brain scans, CNBC reports.
In December, the researchers released results of their research on BioRxiv, an archive of unpublished science papers. "We have been studying methods to reconstruct or recreate an image a person is seeing just by looking at the person's brain activity," Yukiyasu Kamitani, PhD, one of the researchers, told CNBC.
Here are six things to know about the research project.
1. Previous research has successfully used machine learning methods to generate visualizations of what an individual is thinking based on MRIs of the participant's brain. This research has typically focused on simple images, such as black and white letters or shapes.
2. However, the four researchers from Kyoto University are using an advanced machine learning technique called deep learning to visualize more complex thoughts from brain scans, such as images of animals with multiple layers of color and structure.
3. During the course of a 10-month study, the researchers displayed images such as photographs of animals, geometric shapes and letters to three participants. In some cases, researchers analyzed brain activity while the participant viewed an image. In others, they measured brain activity after the participant had viewed the image, while asking the participant to recall it.
4. After the researchers scanned a participant's brain activity, the AI system attempted to reconstruct the image using reverse-engineering. The researchers' goal was to generate visualizations of the subjects' thoughts.
5. The researchers considered their success in reconstructing images for both types of participants as a breakthrough, although the AI system had a more difficult time recreating an image from the recalled brain scans.
6. Dr. Kamitani told CNBC future applications of the AI system, which the team is continuing to improve upon, may extend to helping individuals visualize their dreams or even providing clinicians with images of psychiatric patients' hallucinations.
To view reconstructed and original images from the research project, click here.