A group of researchers from Japan, representing Kyoto University, presented an artificial intelligence that can “understand” what a person thinks, and try to portray it. Naturally, the technology is far from perfect, however, and the results sometimes look quite fascinating.
For ten weeks, the researchers showed the volunteers various images. In parallel using different equipment recorded the reaction of the brain of study participants for each of the images, and scientists developed the algorithm processed the accumulated information to then “learn” them. Information was collected not only when the volunteers were directly looking at any picture, but when they were asked to recall the picture that they saw before.
After scientists found that artificial intelligence is sufficiently “trained”, the experts went to his trial. The participants were asked to look at the image of a geometric shape or picture of the animal, and the algorithm tried to “guess” what they’re looking at. In most cases, looking at the image created by this program is quite difficult to say with certainty what it looked person, but sometimes the similarities between the “original” picture and the one that was created based on the thoughts of the volunteer, have a rather obvious similarities.
Deep image reconstruction: Visual imagery, GIF version
Left: the Imagined images
Right: Images reconstructed from brain activity (being optimized) pic.twitter.com/m9ZwPRFwGD
— ‘Yuki’ Kamitani (@ykamit) January 4, 2018
Although the researchers do not deny that their project requires very significant improvements, they believe that in the future artificial intelligence such fascinating prospect — it is possible that one day with its help you can view “video” of human sleep.
Scientific work focused on the results, specialists were published in the online library preprints and BioRxiv.
The most interesting day in the “MK” in one evening newsletter: subscribe to our channel in the Telegram.