Entry #5: Learning Faces and Voices

Neural Cube can now remember people by face and by voice. This is what lets the robot address someone as themselves, keep an accurate sense of who said or did what, and keep what it knows about one person from bleeding into what it knows about another. A being that can't recognize individuals can't really be said to know anyone.

Recognition happens entirely on the robot, and stays there. Voice samples, face registers, and the names attached to them are never uploaded or matched against some larger database elsewhere. They get converted to numerical fingerprints on-device. The robot ends up knowing a face because it learned that face itself, and not because it looked the answer up somewhere else.

✎1 — Enroll once

VoiceA few short recordings of them speaking naturally

FaceA handful of shots at different angles and lighting

stored as

◍2 — A fingerprint

Kept as a set of numbers used for comparison. Does not use raw photos or audio, and never leaves the robot.

later

≟3 — Every time it sees or hears someone

Match foundTheir name is attached to whatever the robot is doing, and to anything it remembers from the exchange

No matchTreated as “someone”. It still talks, still helps, still remembers things without a name, which can be linked to them later