I had a fascinating conversation with another Preparing Future Faculty fellow in the neuroscience & psychology department last week about the kind of visual processing that happens outside explicit awareness. He was talking about perception studies that are able to use brain activity to predict identification of visual stimuli. A few days later, I came across this article about research conducted in the Gallant Lab at UC-Berkeley. Three subjects watched movie trailers, and the researchers matched the recorded brain signals to a database of 18-million seconds of random YouTube video clips (which did not include the trailers they had watched). The computer program they designed was able to pick out clips for each of the brain signals and put together a composite video that mirrors the scenes the subjects saw in the trailers.
Here is their side-by-side comparison of the trailer they watched (on the left) and the composite generated from the brain signals:
The software selected multiple “matching” clips from the database and created a composite:
Although this technology has been used in the past, this is the first time the images have been dynamic: movies of the mind.
The range of ethical implications are vast, of course: what sort of right-to-privacy issues can we glimpse on these horizons? As it stands, the technology requires images that have already been presented to the subject (and we’re talking specifically visual cortex here; this is not quite “mind reading”… yet), but the potential for one day “decoding” the perceptions of unconscious subjects (dreaming to comatose) is just incredibly cool.
More related links:
- The Quest to Read the Human Mind (PopSci)
- The Gallant Lab, UC Berkeley (source for YouTube videos above)
No Comments