387: It Might Not Be as Cool as You Think

Face ID Adaptation

John Siracusa and Marco Arment explore how Face ID technology adapts to users wearing masks. John suggests that successful unlocks, even when not perfect, help the system learn and adjust over time 1. Marco shares his experience of Face ID starting to recognize his masked face after repeated use, hinting at a learning algorithm at play 2. Casey Liss adds that Face ID has worked for him even when part of his face was covered, suggesting the technology focuses on the overall facial structure 3.

I'm happy it is this way, because it's one of those things where it would be better for all of society right now if it was more convenient to wear a mask.

--- Marco Arment

Marco believes that any improvement in Face ID's ability to work with masks could encourage more people to wear them, despite potential security trade-offs.

   

Learning Mechanism

The hosts discuss the potential learning mechanisms behind Face ID's adaptation to masks. Marco notes that his phone began unlocking with a mask after several attempts, suggesting an unintentional learning process 4. He speculates that Apple may have quietly adjusted the algorithm to be more accepting of masked faces, though this could reduce security 4. John explains that when Face ID fails to recognize a user, entering a passcode helps the system learn and improve its accuracy 5.

I can't tell whether this is a side effect of the way Face ID works, where it's training itself.

--- Marco Arment

This discussion highlights the balance between convenience and security in evolving technology.