Yeah, I think that bringing things we currently use a computer into the field of vision will be a big and important change, but when it starts doing things based on the context, that's when it really starts changing things. Some people are trying to do this kind of thing with smartphones (Color, with their attempts to piece together what's going on via photos, for example), but many of the more advanced attempts rely on an unnatural action like pulling out your phone and taking a picture or checking in or what have you. When you have an always-on forward facing camera and other sensors, and it's always accessible with a glance rather than awkwardly ignoring people and pulling something out of your pocket, things become a whole lot more interesting (and potentially creepy).

One quick example, but imagine something that annotates every person you see that gets near you and is clearly talking to you, so you never forget another name? Something like rapportive, but for in-person encounters, and provides a quick dossier to refresh you on who they are. I think it'd change things pretty radically. It'd be tough to do, but I'm sure Facebook and maybe a couple other companies have a large enough tagged face training corpus to do it.