Researchers at Japan’s Kyoto University have developed a new artificial intelligence program that can be used to decode human thought patterns, translating brain scans into pictures of not only what the subject is looking at, but also simply remembering, into digital images that can be reviewed.
Although this technique is not new, earlier experiments into machine mind-reading were only capable of producing simplistic binary images. The Kyoto team has refined this technique, through a deep neural network AI program, successfully decoding more complex "hierarchical" images, such as that of an owl along with a less distinct human depicted in the background. The information processed by the AI is produced from magnetic resonance imaging (MRI) scans, and the team’s research illustrated that the human brain processes images in a much more complicated way than one would assume.
"Our previous method was to assume that an image consists of pixels or simple shapes," explains Computational Neuroscience professor Yuki Kamitani, head of Kyoto University’s Department of Neuroinformatics. "But it’s known that our brain processes visual information hierarchically extracting different levels of features or components of different complexities."
Kamitani’s team found that the technique also worked when the subjects were simply remembering the image, and not just when they were actively looking at it, although the clarity of the image wasn’t as reliable — scientists are still uncovering the complexities of how we store and access our memories.
The applications for a technology capable of producing images from human thought range from being able to produce art directly from one’s imagination, to psychiatric treatments for patients suffering from hallucinations. Since the technology relies on a combination of bulky MRI scanners and large computers to run the AI that decodes the thought patterns — remember that this also only works under optimal conditions — we don’t have to worry about having our cellphones or laptops read our thoughts without our knowledge.