It's always dangerous to predict that soon we will all videoconference, but the technology really is getting better (See the Marconi ViPr, for example.) And bandwidth is getting cheaper. The Marconi system wants 8 megabits bidirectional -- the secret being not to do very fancy compression to instead push for really low latency and hi resolution -- but this is getting to be more affordable.
Here's a possible application. Nice restaurants making tables for a remote virtual dinner date. A table for one facing a screen, connected with another such table in another city. Get together for dinner with distant friends. Of course, one would also like to do it for larger groups but as the Marconi project shows, that requires losing too much resolution (at least for now) to get an acceptable experience.
Another barrier -- you want a lifesize screen here, but a bigger screen means a larger difference from the other's eyes and the camera, and that just looks wrong. The low latency requirement stops us from playing cute tricks, like having two cameras left and right and trying to calculate an eye that's looking at you. This is hard to do though some day we might solve it.
For an added touch the restaurants, if they share a menu, might even offer you a taste of the dish the other person ordered. Here, try this, it's delicious. Hokey, but who knows? Harder of course to provide physical contact, though there have been attempts to build gloves for virtual handsakes and in the extreme there is teledildonics.
Of course the system (without the teledildonics) would also be popular for business lunches. Of course there are lots of videoconferencing centers trying to sell their services for business meetings, but we have a certain attraction to meeting for lunch for a one one one or 2 on 2. It might make work what we haven't done before.
On the other hand, I think we should work to eliminate the speakerphone from the audio conference call. It's never great. Just give everybody in the room a bluetooth earbud or headset, and headsets for anybody joining remotely. Good audio, full duplex, ability to interrupt. Forget the echo cancelling and tinny acoustics.