MIT’s AI Recognizes Faces Like Humans Do
Researchers at MIT have developed a new AI that recognizes faces like humans do. The still unnamed AI, according to the researchers, replicates an aspect of the human brain function, which was missed by earlier face recognition models.
The MIT researchers described in details the newly developed facial recognition system in a paper published at Arxiv. The journal Computational Biology recently published the same paper. The MIT researchers – including Joel Leibo, also a researcher at Google DeepMind and Tomaso Poggio, director of the Center for Brains, Minds, and Machines (CBMM) – trained their machine-learning system by feeding it with a battery of sample images.
Before the development of this new facial recognition system, previous systems could not identify faces when turned in the right or left directions. Despite not being trained to do so, the newly developed AI taught itself to recognize faces when turned in the right or left directions. What the researchers only did was trained the system to recognize faces rotated at approximately 45 degrees.
An earlier findings by Winrich Freiwald, also one of the authors of the paper, demonstrated that a particular cluster of neurons will fire in a macaque monkey’s brain if a face is rotated 30 degrees, 45 degrees, 90 degrees, or anywhere in between. The new paper by researchers, including Freiwald, Leibo and Poggio showed that the new AI was able to replicate the primate’s brain behavior.
The new paper, Poggio said, is “a nice illustration of what we want to do in (CBMM), which is this integration of machine learning and computer science on one hand, neurophysiology on the other, and aspects of human behavior.”
Christof Koch, president and chief scientific officer at the Allen Institute for Brain Science, told MIT News that the new finding of the MIT researchers is a significant step forward. Koch added, “In this day and age, when everything is dominated by either big data or huge computer simulations, this shows you how a principled understanding of learning can explain some puzzling findings.”