MIT AI Deepens Knowledge of Human Facial Recognition

0
facial recognition
Corepics VOF } Shutterstock.com

Using a computational model of how the human brain recognizes faces, scientists at MIT have discovered a critical step in the brain’s recognition process. This sheds light on how our brains function, and it also shows that AI can replicate our neural processes.

Neural networks have been the primary medium for testing AI technology. MIT‘s latest work indicates that working with machine brains may have taught us something about our own. At the Center for Brains, Minds and Machines (CBMM) at MIT, researchers found that a facial recognition system that they designed had spontaneously invented a procedure to identify portraits despite image rotation. This discovery reinforces a facial recognition theory based on MRIs of primate brains.

“As the AI learned to spot faces, it created its own step in the process to recognize a face that had been rotated.”

Different Degrees of Facial Recognition

The system, which is not yet named, was designed to identify faces from a batch of images. As the AI learned to spot faces, it created its own step in the process to recognize a face that had been rotated.

The system could recognize a flipped face, as long as it was rotated at the same angle. This is the property known as “mirror symmetry.”

This discovery resembles an observed feature of how primate brains process faces, which indicates that the system may be unintentionally imitating organic brain processes. However, according to Tomaso Poggio, director of the CBMM, “This is not a proof that we understand what’s going on.” More research is needed before a connection between the two types of minds is confirmed.

What Kind of Future do we Face?

Big companies like Apple, Google and Microsoft invest heavily in facial recognition technology. While a connection between the organic and machine minds is an intriguing proposition, tech giants could turn the idea into big money.

Advancing facial recognition software could assist law enforcement and security firms. The amount of data in surveillance systems is enormous, and neural networks can quickly process huge amounts of data.

AI may soon recognize faces as well as humans. As Poggio asserts, more research and development will be necessary. Their curiosity will surely be spurred by an AI mimicking the human brain.

banner ad to seo services page