"A device capable of" reading "the brain and producing a visual image of the experience of a person could soon become a reality", the journal Nature announced a lacónico release, the purpose of results discreetly published in its online edition today.
Does understand well? Are they really talking to detect, in the midst of our neurons, the images that shape our mind when we look at something? Because they are. It is now for tomorrow, but the truth is that Jack Gallant and his team of the University of California, Berkeley, have just taken a fundamental step in this direction.
Using only the analysis of the patterns of brain activity generated by natural viewing of images, showed that it is possible to identify, with great precision, the image that the human brain is to see.
The experience they have done is similar to that trick which is to choose a letter without the magic show, look at it and make embedding it in the deck. Instantes later and some abracadabras later, the magician pulls some of the letter and the public applauds. Except that here it is not magic.
Dir you - I will see what
In the lab, two participants (two of the authors of the study, Thomas and Kendrick Naselaris Kay) began by viewing, randomly, 1750 photographs of various objects. "People, places, animals, those things usually shoot during the holidays," said Gallant to the PUBLIC. And thanks to the technique of functional magnetic resonance imaging (fMRI), which allows you to view real-time brain activity, there was what was happening in the visual areas of the brain, where the signals are processed nervous from the retina.
From these records, and based on a mathematical model that simulates the processes brain, the researchers built a "decoder" which produces, from any image, a particular pattern of brain activity. The model takes into account, inter alia, "hints" that the visual neurobiólogos know that the brain uses to recognize what you see, such as the location in space and orientation of the elements present in the image presented.
After this phase of "training" of the device, asked to the participants to view each for a single image, this time chosen at random from a collection of 120 photographs, all of them different from that had seen until then. Once again, there was the brain activity of both.
But now, in parallel, the scientists also appealed to the mathematical simulation. More specifically, for each of these 120 photos, the decoder system has generated a pattern of brain activity. The final step was to choose among all these standards possible, the one that most seemed to the pattern of brain activity actually produced by the participants themselves for the photo while it was submitted. The results of multiple tests were spectacular: the model chosen the right photo in 92 percent of cases for Naselaris and in 72 percent to Kay!
When the number of images possible increases, performance decreases, but is still remarkable: with 1000 images, the decoder yet to identify 82 percent of the images presented to Naselaris. "Our calculations suggest - the authors say - that a number of billion images (of the order of indexed by the Google on the Internet), the image would be some identified nearly 20 percent of the time. That is, if a person choose and olhasse for a photo at random on the Web, yet be able to identify that image from the brain activity once every five. "
Dreams and memories
If it were possible to systematically identify any photo without having any prior knowledge of the set where he was chosen, then yes, it could be said that the device is able to "see" the image within the brain. But this has not happened: "No one can reconstruct the photograph that the person seen without knowing the set of images where she was taken," the authors explain.
"However, our results suggest that there may be insufficient information on the data of fMRI to achieve there in the future." Indeed, one thing that this study showed is that the information contained in fMRI data is much richer than previously thought.
Scientists go even further: if it is that when dreaming or mentally evocamos a scene, the images produced by these mental experiences are processed by brain images as if they were real, coming from the outside world through the eyes, nothing prevents think it is possible one day, also decode these images "virtual" with the same kind of
Technique. "Since the measurements of brain activity and mathematical models of the brain have the sufficient quality, it should be possible, in theory, decode the visual content of mental processes such as dreams, memory and imagination."