Brain-Reading Machines Know What You've Seen

We may earn a commission from links on this page.

fMRI technology has been used for lie detection and identify memories of an event, but some researchers are looking to push its mind-reading capabilities even further, using brain scans of subjects to reconstruct images they've recently viewed.

In the most recent issue of Neuron, Jack Gallant and Thomas Naselaris of the University of California, Berkeley describe their process for reconstructing images from brain scans. While previous studies have used fMRI technology to identify an image a subject has recently viewed, those studies involve a subject viewing a specific picture from a limited set and then showing them all the pictures in that set to see which one they previously viewed. Gallant and Naselaris' research is focused on reconstructing visual information based solely on readings from the brain.

The researchers have worked on identifying which parts of the brain are associated with certain types of visual information. For example, different regions are active when viewing a face or a crowd of people or a building. They believe as their understanding of these regions grow more sophisticated, they will better be able to pin point from fMRI scans what a subject has seen. Currently, they are able to get a rough idea of what a subject is looking at and then pull a corresponding image from a vast database of images. It's not quite a reconstruction yet, but it is close and will improve with further research.


An article from Wired notes that full visual reconstruction is still decades away, but could have implications for devices that read dreams or machines controlled through thought. However, as fMRI technology is so often used now in court cases (arguably prematurely given the current technology), it seems likely that one of the first applications we could see from this technology will be in justice system.