Beating Google — One Face at a Time

Facial recognition technology was advancing faster than the legal and regulatory frameworks designed to govern it — and around 2011, a wave of startups and established tech companies were competing to build the most accurate commercial systems, with Google among the most closely watched participants.

The technology works by mapping the geometry of facial features — the distance between eyes, the shape of the jaw, the contour of the nose — and creating a numerical representation that can be matched against a database of stored images. Systems trained on large photograph collections were already reaching accuracy rates that, under good conditions, could exceed human recognition.

Google's entry into the space — through its 2011 acquisition of the face-recognition startup PittPatt, born out of Carnegie Mellon research — raised questions the company itself seemed ambivalent about. Chairman Eric Schmidt said publicly that facial recognition was "the one technology Google has decided to hold back" over privacy concerns — a rare acknowledgment from a company that typically moves aggressively into new technological territory.

The concern was specific and concrete: a mobile application capable of identifying strangers on the street by photographing them would fundamentally alter the possibility of anonymous public life. Stalkers, abusers, and authoritarian governments all have obvious use cases for technology that eliminates the anonymity that physical public space has historically provided.

Competing companies were less hesitant. Facebook began deploying its own facial recognition across hundreds of millions of user photographs with its tag-suggestion feature. Law enforcement agencies started building large biometric databases.

The question of who would set the norms for this technology — companies, governments, or users — remained entirely unresolved. The faces, meanwhile, kept accumulating in databases.