The problem is - most of the cool and useful personal applications of glass fly in the face of various social expectations. The very idea you could be recording someone caused a backlash the last time, and that's nowhere near doing facial recognition...
If it's done in real-time, i.e. the scanned images aren't saved then surely a "This is person <X> you've met and tagged as X" is less creepy than actually videoing someone? You wouldn't get any information you haven't yourself added to the device (although it would probably need to lean on external data-sets for the training).
I wouldn't suggest facial recognition should recognise anyone you haven't met yet, I think that would be a bit weird (although I think it is the future anyway), but a way to effectively add a tag on someone you know would be great. Most people can do this without the technology to varying levels of accuracy and and breadth of their acquaintances.
I think it is creepier by definition, through the very fact that the other person can't verify you aren't recording. A camera is a camera. Even if the product officially doesn't record anything, who's to say I didn't mod my glasses' firmware to dump the video buffer? Not to mention that once video stream goes into cloud, you lose control over what happens to it.
