On September 12, 2017, Apple unveiled the iPhone X and with it Face ID, a system that unlocked the phone by recognizing the owner’s face. Apple’s press release described a TrueDepth camera made up of “a dot projector, infrared camera and flood illuminator,” which projects “more than 30,000 invisible IR dots” to build a depth map of the face.
The recognition itself ran on the device using machine learning. According to Apple, “the IR image and dot pattern are pushed through neural networks to create a mathematical model of your face,” and that model is compared against the stored one to confirm a match “while adapting to physical changes in appearance over time.” This neural processing happened on a new, purpose-built component, Apple’s first neural engine, built into the A11 Bionic chip. Crucially, Apple stressed that “all of the processing is done on-device and not in the cloud,” with face data kept in the chip’s secure enclave.
Face ID put deep-learning-based facial recognition in the pocket of a mass-market audience and helped popularize the idea of dedicated AI accelerators in consumer hardware. It also made on-device, privacy-preserving machine learning a selling point rather than a footnote.
For business readers, Face ID marks the moment neural networks moved from data centers into a chip in a phone, and the moment a hardware company turned on-device AI and privacy into a competitive advantage. It also kept face recognition - a technology with serious bias and surveillance concerns elsewhere - framed by Apple as a local, consensual unlock rather than a tracking tool.