Slash Boxes

SoylentNews is people

posted by Fnord666 on Sunday June 18 2017, @06:04PM   Printer-friendly
from the never-forget-a-face dept.

Scientists have reconstructed faces nearly perfectly by analyzing the activity of neurons in macaque brains:

[Using] a combination of brain imaging and single-neuron recording in macaques, biologist Doris Tsao and her colleagues at Caltech have finally cracked the neural code for face recognition. The researchers found the firing rate of each face cell corresponds to separate facial features along an axis. Like a set of dials, the cells are fine-tuned to bits of information, which they can then channel together in different combinations to create an image of every possible face. "This was mind-blowing," Tsao says. "The values of each dial are so predictable that we can re-create the face that a monkey sees, by simply tracking the electrical activity of its face cells."

Previous studies had hinted at the specificity of these brain areas for targeting faces. In the early 2000s, as a postdoc at Harvard Medical School, Tsao and her collaborator electrophysiologist Winrich Freiwald, obtained intracranial recordings from monkeys as they viewed a slide show of various objects and human faces. Every time a picture of a face flashed on the screen, neurons in the middle face patch would crackle with electrical activity. The response to other objects, such as images of vegetables, radios or even other bodily parts, was largely absent.

Further experiments indicated neurons in these regions could also distinguish between individual faces, and even between cartoon drawings of faces. In human subjects in the hippocampus, neuroscientist Rodrigo Quian Quiroga found that pictures of actress Jennifer Aniston elicited a response in a single neuron. And pictures of Halle Berry, members of The Beatles or characters from The Simpsons activated separate neurons. The prevailing theory among researchers was that each neuron in the face patches was sensitive to a few particular people, says Quiroga, who is now at the University of Leicester in the U.K. and not involved with the work. But Tsao's recent study suggests scientists may have been mistaken. "She has shown that neurons in face patches don't encode particular people at all, they just encode certain features," he says. "That completely changes our understanding of how we recognize faces."

Also at Singularity Hub and The Guardian:

Professor Rodrigo Quian Quiroga, a neuroscientist at the University of Leicester who was not involved in the work, described it as "quite a revolution in neuroscience". "It's solving a decades-long mystery," he added.

The puzzle of how the brain identifies a familiar face dates back to the 1960s, when the US neuroscientist, Jerry Lettvin, suggested that people have hyper-specific neurons that respond to specific objects, a notion that became known as "grandmother cells", based on the idea that you have a specific neuron that would fire on seeing your grandmother.

More recently scientists found "face patches", clusters of neurons that respond almost exclusively to faces, but how recognition was achieved had remained a "black box" process. In the absence of proof otherwise, the grandmother model continued to appeal because it tallied with the subjective "ping" of recognition we experience on seeing a familiar face.

"This paper completely kills that," said Quian Quiroga.

The Code for Facial Identity in the Primate Brain (DOI: 10.1016/j.cell.2017.05.011) (DX)

Original Submission

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 2) by fyngyrz on Monday June 19 2017, @05:08PM

    by fyngyrz (6567) Subscriber Badge on Monday June 19 2017, @05:08PM (#528021) Homepage Journal

    Until next year when we read: ""(S)he has shown that neurons in face patches don't encode certain features at all, they just encode certain people," he says. "That completely changes our understanding of how we recognize faces."

    Aside from the fact that this is a monkey-based model and it's not been established that humans use the same one (though it seems probable), this mechanism would not be invalidated by additional feature recognition for rare or subtle features and faces.

    For instance, moles can appear anywhere, and animals have faces these particular systems might not be specialized to recognize, but it might be that moles aren't used for recognition at all, or there might be another mechanism for positionally-based rare features like these (moles, scars, dimples, facial hair, etc.), or entirely another system might exist for recognizing, for instance, specific predators of another species.

    Clearly, for these monkeys, the feature-based identification and triggering is operating and can serve to characterize human faces, the experiment demonstrates this explicitly; the assumption being made is that's what is used for the actual recognition downstream from there. Might not be, but if not, it would be interesting to know what this is being used for as well as what system(s) is/are doing the recognition.

    Also, there's a base assumption that this drives recognition of humans in monkeys, as that's strongly suggested by the predictive algorithm; it may not. It may only act to characterize and be thrown away. Perhaps they don't recognize humans this way, or the system is unused in favor of something else. For instance, perhaps what this really does is track expressions, and its ability to characterize a face is incidental. Lots of possibilities suggest themselves, and of course, validating it in humans is down the road a bit anyway.

    But! It does mean that a mechanism that can perform facial recognition in nature has been positively identified. I find it interesting as some of the methods we have already worked out reduce faces to a relatively small set of vectors describing feature (contrast) distances from one another. Fractional-based weightings vs. interval-based weightings are almost irrelevant, they're just data encoding differences. It took us quite a while to dig down to that solution, though.

    Biology contains, IMHO, every lesson we need to achieve actual AI (conscious, reasoning machines as opposed to simplistic LDNLS. []) We're just not as good at reading the lessons out as we need to be if we're not to have to figure out every step from first principles.

    Having said that, considering the chaotic state of our society and politics, I'm willing to speculate that it's probably better if it takes us longer to figure all this out. I can at least hope things will be better in the future.

    The eyes are the windows to the soul.
    Sunglasses are the window shades.
    Starting Score:    1  point
    Karma-Bonus Modifier   +1  

    Total Score:   2