Brain Scanner Can Tell What You're Looking At

palegray.net writes "Wired News brings us an article about brain scanning systems that can accurately tell what you're looking at by analyzing your brain's electrical activity. Using a database constructed of readings taken on test subjects who were shown thousands of photographs, the system works in real time to decipher what you're seeing. Naturally, there are some ethical concerns over some potential applications for this technology. Definitely a new twist on "input devices.""

It isn't time for fear mongering yet

[LockeOnLogic]

Although it true that our ability to image the brain is now allowing us to detect "thought" in the brain, it really needs to be pointed out that this is very in the lab sort of stuff. It doesn't just involve sticking you in a tube and viola a little readout comes out telling you what you were thinking. It requires finicky, multi-million dollar, difficult to interpret equipment. First have to baseline a persons normal brain function then after detailed analysis by crazy smart cognitive neuroscientists we can sort of glean very simple conclusions. Are you adding or subtracting from a number (not found out in real time btw)? Looking up or down? Which, incidentally, I can also determine by looking at your eyes. Basically the stuff here and in other imaging studies is cognitive childsplay in comparison to the "reading of someones thoughts" people seem think is around the corner. We are so far off from that state of technology that ethics really aren't an issue, yet. It is kind of interesting to me that ethical concerns are beginning to become a concern in research of cognitive neuroscience, but needless worry is premature. This is like people starting to fear the atomic bomb right after discovering uranium.

Good research, but not mind reading...

[MacBorg]

Gah. Can we file this under really bad summary - this is basically an expansion of work that has been underway for a few years now (just read a paper on a similar concept from 05). What we're really seeing is a pattern-matching algorithm - train it using fMRI data from visual cortices and, with a limited subset, it's pretty accurate. Honestly, as a vision researcher, the more interesting bit isn't the so-called "mind reading" bit, although it is a good trick - it's the fact that it works across subjects with a respectable amount of accuracy (which indicates that activation in V1/V2/V3 is not overly dissimiliar between subjects). Cool work though...