Yes, but a narrower interpretation than used in more robust systems like those that we built. They don't need an entire SLAM system with loop closure and relocalization because you can mimic those with the simple facial recognition portions. However they added orientation functionality but without visual odometry. Functionally it works the same, and is the easiest way to describe it.

Not in the robotics sense where you map out a hallway of a building or something. But it does map facial landmarks very well (and I suppose that will also give you localization relative to the face).

No, it does face recognition and orientation, it's far simpler than slam and have nothing to do with it.