Can a computer read your mind?
This story has been making the rounds in the news lately, 
Mind reading: This man can't actually read her mind, but scientists just might be able to.Courtesy Wikimedia Commonsdescribing a study in which a computer reconstructed video clips that subjects viewed. Participants in the experiment watched video clips while their brain activity was monitored by fMRI. Then the computer selected from millions of YouTube clips it had "seen" to make a composite video clip, some of which look eerily similar to the original videos that the participants watched.
No need to break out the tin foil hat just yet. Volunteers for this study had to lie motionless for hours for their brain activity to be scanned. But this technology could easily lead to systems of communication that aid patients whose normal channels of communication have broken down because they are paralyzed or even comatose. If you're interested, you can see the original paper here (subscription required). Take a look at some sample video from the study:
EmailYour Comments, Thoughts, Questions, Ideas
Explore
Recent comments
Subscribe
Connect with Science Buzz on Facebook and Twitter.
You can also subscribe to our RSS feed using any newsreader software.
This idea troubles me.
There's a Fringe episode that does something similar to this... Creepy. And totally cool.
I wonder if the parts of the brain activated while watching video are the same parts people would use to imagine or to reply memories, which we would need to know to make this technology work as a communication device. Or I guess you would just change the part of the brain you're observing? But something tells me it's a bit more complex than that. Where do the details go? Or is the limitation coming from YouTube?
Thanks for the comments, everyone!
If you're asking why the reconstructed images aren't as detailed as the images that the subjects viewed, it's because the computer had to pick the best matches from its training set--or rather it read the fMRI data of the subject and found the training image whose associated brain pattern most closely matched the pattern of the subject. For example, with the elephants walking in from the left, the closest images were just those with a dark shape on the left moving toward the right, because there were no clips of elephants in the training set that it could draw from.
I was talking to a friend about this paper and they suggested it might be used to capture someone's dreams. I though that was a neat application idea.
I believe computers can't read our minds.
Post new comment