So, add a 3d space with spatial audio, and it's natural to form groups. Add really good spatial audio, and the system can handle the different unconscious movements we do to pick up snippets of activity around us and figure out which conversations we want to be closer to. (You still can't replicate the free food, though, so it's probably doomed anyway, but it's worth exploring).
His bit about how you can't really control volume in that kind of experience was really interesting to me.
"We should forget about small efficiencies, say about 97% of the time: premature optimization is the root of all evil. Yet we should not pass up our opportunities in that critical 3%"
People often ignore that second half.
> "I mean how many live on a 10 Mbps home connection with high latency yet they make perfectly fine calls."
I'm not sure I agree with 'perfectly fine', people just suffer through things being terrible because they don't know they could be better. Latency on a 1:1 call is bad, but tolerable - things get a lot worse as you move towards group interaction (dealing with group video chats often because of covid it's clear how much worse this is than in a real life room).
It's not premature optimization to get the latency down when that's probably the most critical feature tied to usable performance and better experience, arguably everything else should be lower priority.
I suspect a lot of the problems with videochat (and VR) has to do with this.
Bad latency is frustrating and wears you down, at some point it's not worth attempting to speak up because it's tiring to deal with.
In real life communication transitions between people are fast and you can overlap a bit.
VR could solve the issue of what people are looking at least which I think is the other issue, though I'm not sure how well avatars will fit in for that.