First of all, before I introduce the interesting topic regarding the brain and facial recognition, I want to sincerely apologize that I haven’t been posting much. I plan to make this a weekly thing now, so stay posted for that!
So, back to the topic at hand. MIT has finally been able to reproduce how the brain recognizes faces, which is absolutely incredible. What they did essentially, was create a machine-learning system that implemented their model, training it to recognize specific faces. In essence, it’s a AI system (artificial intelligence), using a neural network. It’s so unique because it’s different. Previously, these systems didn’t have the important aspects of the human brain functioning and wiring, however, this time, they do. The model was so accurate, that it had a processing step representing the degree of rotation (with no differentiation between left and right) of the face.
Here’s the catch: this wasn’t inbuilt. In fact, it’s emergent, as it came from the meticulous training sessions. This is a good thing, because that means their model truly resembles the human brain.
From the main release site– “This is not a proof that we understand what’s going on,” says Tomaso Poggio, a professor of brain and cognitive sciences at MIT and director of the Center for Brains, Minds, and Machines (CBMM), a multi-institution research consortium funded by the National Science Foundation and headquartered at MIT. “Models are kind of cartoons of reality, especially in biology. So I would be surprised if things turn out to be this simple. But I think it’s strong evidence that we are on the right track.”
Poggio had previously believed the brain produces ‘invariant’ representation of faces and objects, which basically means that these representations are indifferent to a variety of “could-have-influenced” factors: orientation in space, distance from the person viewing, and their location in the visual field to name some. In 2010, he performed an experiment with macaque monkeys, on their neuroanatomy when their brains tries to recognize faces. In a serious of locations on the brain, some fire only to specific facial orientations, and other locations fire regardless of orientation, which is the “invariant” representation. These results combined with the results so far from this recent study using the accurate model seem to conclude that neurons are “mirror symmetric”, or sensitive to angle of face rotation, not direction. It’s important to make a distinction between intermediate, first, and final regions from the data observed. In intermediate regions and their clusters of neurons, each one cluster seems to specialize in one specific degree with no regards to direction. The first region clusters does depend on both direction and rotation, while the final region makes no difference– the same cluster can fire regardless of degree; no specifics are involved.
Here’s a short video on the topic if you’re interested.
Next week’s post will be on the cognitive disorder relevant to this article, prosopagnosia. I got interested in the topic after watching MIT neuroscientist, Nancy Kanwisher’s TED talk on the neural portrait of the human mind. If you haven’t seen her TED talk either, I highly recommend you check it out here!