Science fiction is becoming a reality as Japanese scientists reach new heights in Artificial Intelligence, announcing a program capable of “mind reading” through recreated images one pixel at a time.
The program, developed by a team of experts from the University of Kyoto, is based on a neural network system, decoding brain activity and reproducing images presented to test subjects.
Photos of stained glass windows, owls, aircrafts and red post boxes were presented to three volunteers. FMRI scans were then used to measure the information, analyzing electrical activity as well as the variants of blood flow to the brain.
The results varied according to its volunteers, researchers said, but the majority showed an astounding level of accuracy recreating the objects as well as others imagined by test subjects such as bowling balls, leopards, goldfish and swans.
"We found that the generated images resembled the stimulus images (both natural images and artificial shapes) and the subjective visual content during imagery,” the team of scientists said in a report published by BioRxiv.
"While our model was solely trained with natural images, our method successfully generalized the reconstruction to artificial shapes, indicating that our model indeed 'reconstructs' or 'generates' images from brain activity, not simply matches to exemplars."
Experts say the program “opens a unique window into our internal world” and could potentially provide a conduit for communication between those in a vegetative state and their family members.
Access to daydreams, memories as well as other imaginary scenarios could also be unlocked through further developments into the project.