AI-synthesized faces cross the "uncanny valley"

February 24, 2022

Artificial intelligence (AI)–powered audio, image, and video synthesis has democratized access to previously exclusive Hollywood-grade, special effects technology. From synthesizing speech using anyone’s voice, to synthesizing an image of a fictional person and swapping one person’s identity with another, or altering what they are saying in a video, AI-synthesized content holds the power to entertain but also deceive. So-called "deep fakes" are also being weaponized for the purposes of nonconsensual intimate imagery, financial fraud, and disinformation campaigns.

In a new paper in PNAS, BIDS Faculty Affiliate Hany Farid and colleague Sophie J. Nightingale report that their recent evaluation of the photorealism of AI-synthesized faces indicates that synthesis engines have passed through the "uncanny valley" (a term used to describe a human's emotional response of unease or revulsion toward the human-like appearance of computer-generated content or a robotic object that is highly realistic) and are now capable of creating faces that are indistinguishable — and deemed more trustworthy — than real faces.

"We encourage those developing these technologies to consider whether the associated risks are greater than their benefits," say the authors. "At this pivotal moment, and as other scientific and engineering fields have done, we encourage the graphics and vision community to develop guidelines for the creation and distribution of synthetic media technologies that incorporate ethical guidelines for researchers, publishers, and media distributors."

Farid PNAS Feb 2022 - 16 faces - photo square - "A representative set of matched real and synthetic faces."
"A representative set of matched real and synthetic faces."



AI-synthesized faces are indistinguishable from real faces and more trustworthy
February 22, 2022  |  Sophie J. Nightingale and Hany Farid  |   PNAS 

Humans Find AI-Generated Faces More Trustworthy Than the Real Thing:br /> Viewers struggle to distinguish images of sophisticated machine-generated faces from actual humans
February 14, 2022  |  Emily Willingham  |   Scientific American

Featured Fellows

Hany Farid

EECS and School of Information, UC Berkeley
Faculty Affiliate