Researchers at the University of California Berkeley are improving the science behind mind-reading. Much like reverse engineering an encryption algorithm using both the cipher and plaintext, their machine begins with a known library of images and attempts to map brain activity to them. Subjects were shown video from YouTube movie trailers (roughly 18 million seconds known to the machine) while an fMRI mapped their brain activity. Their algorithm then attempts to correlate those activities to particular shapes, colors and movements to recreate video of what a subject is thinking about. "You're reconstructing a movie that they saw using other movies that they didn't actually see," says neuroscientist Jack Gallant, to which he caveats, "We're trying to understand how the brain works ... We're not trying to build a brain-decoding device." [Be sure to watch the embedded video]