imagine a world in which machines can read people's minds just by tracking their brain activity--in which, say, cyborg detectives can tell what you're thinking by monitoring the flow of blood and other physiological responses inside your head. sound like science fiction? don't be too sure.
a team of neuroscientists have announced that they've made a "decoder" that can tell what people are looking at just by tracking their brain activity using functional magnetic resonance imaging (fmri). for now, the decoder only works in a controlled setting, but scientists say it's a big leap forward.
for the study, two test subjects were shown 1,750 images while researchers used fmri to track their brain activity. basically, the researchers taught their decoder how the subjects' brains processed visual information.
then, each subject was shown a new set of images. the researchers knew which images were in the new set, but they didn't know which one a test subject would see--sort of like knowing which cards are in a deck but not knowing which one your opponent will draw.
the decoder's task was to figure out which image a subject was looking at, based on brain activity. it did it remarkably well--though it became less accurate the larger the "deck" of possible images was. when a subject "drew" from a set of 120 images, the decoder found the right one 9 times in 10. with a set of 1,000 images, its accuracy dropped to 8 in 10.
the implications
of course, impressive as that card trick is, it's not the same as trying to figure out what someone is looking at without knowing what's in the deck to begin with. to do that, says lead researcher jack gallant, "you'd need a very good model of the brain, a better measure of brain activity than fmri, and a better understanding of how the brain processes things like shapes and colors seen in complex everyday images."
no one has that knowledge yet, but it may well come. in fact, this study is already helping brain scientists better understand how our brains process visual information. and it accomplished more with existing fmri technology than many researchers thought possible.
all of which brings us back to the prospect of real-life mind-reading machines. "it is possible," says gallant, "that decoding brain activity could have serious ethical and privacy implications downstream in, say, the 30- to 50-year time frame." he and his team are quick to insist that "no one should be subjected to any form of brain-reading process involuntarily, covertly, or without complete informed consent." sure. but try telling that to a cyborg detective.
--steve sampson
özet olarak denebilir ki, fonksiyonel manyetik renozans görüntüleme, beyindeki kan akışı ve diğer psikolojik tepkileri gözlemlemek yardımıyla insanların ne düşündüklerini yüzde seksen doğrulukla tespit edebiliyorlarmış.
