It appears machines may already be catching up to humans, at least in the world of computational biology. A team of researchers at the MIT-based Center for Brains, Minds and Machines (CBMM) found that the system they designed to recognize faces had spontaneously come up with a step that can identify portraits regardless of the rotation of the face. This adds credence to a previous theory about how humans recognize faces that was based studies of MRIs of primate brains.
The as-yet-unnamed system is a computational model of how the human brain recognizes faces, and was trained to identify particular visages from a battery of sample images it was fed. In the process of learning to spot faces, the program created an intermediate processing step that looked at “a face’s degree of rotation – say 45 degrees from center – but not the direction.”
In layman’s terms, this means the system, which was looking for invariance (or non-difference) between faces, was able to do so regardless of whether a face was flipped, as long as it was rotated in the same angle. That property is known as “mirror symmetry.”
This discovery excites scientists because it duplicates a previously observed feature of how primates process faces, indicating that the system might be doing something similar to the brain. However, it’s not for sure. “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 CBMM.
The researchers’ machine-learning system in this case is a neural network, which has been employed by tech giants such as Microsoft, Google and Facebook. These companies all have their own facial recognition systems in place, and have been investing in machine-learning to enhance their tools.
Understanding how we recognize people could help facial recognition systems get significantly better and more accurate, which has vast applications in tech. Face unlock is an increasingly popular feature of phones and laptops, and identifying people in photos lets companies like Facebook, Apple and Google better sort your pictures. The downside, if you choose to see it that way, is that surveillance systems could also get accurate at finding the exact individuals they wish to seek from the endless amount of security camera footage and DMV photos they have. While this is clearly in early stages, and a tiny step towards implementing human-level facial recognition in machines, it certainly is a sign that artificial intelligence is capable of replicating specific functions of the human brain.