By Lumos Labs Science Associate, Paul Li, MS Neuroscience.
Movies like Being John Malkovich are based on the idea that one might be able to experience what another human’s mind is visualizing. Most would think that such movies are pure fantasy and science fiction, but researchers at U.C. Berkeley are one step closer to making this a reality.
Using a computational model calibrated to each individual subject, Professor Jack Gallant and his research team were able to use brain activity (measured with fMRI) to identify which of a large set of images was seen by a subject. Importantly, none of the images in the set had been previously seen by the subject, demonstrating the ability to generalize to novel situations. Though performance isn’t yet perfect, it’s impressive. Accuracy ranges from 80% when viewing 1,000 images, to 90% accuracy when viewing 120 images.
Dr. Gallant said, “there may theoretically be sufficient information available to decode memory, imagery and dreams some day, but it will likely be many decades before this is really possible.”