You are right. On reflection, I don't think Robotaxi (in the sense of replacing Uber and Lyft) is as important as Musk thinks it is, although not of zero importance. But I do think that personally-owned L4-capable cars [which is what I was conflating Robotaxi-as-Uber with] will be a huge deal, for both safety and convenience.
If I'm visiting another city or country, I'm just as happy to have a human drive me around, although I wouldn't mind a driverless taxi for boring routes. But closer to home, I'm very much looking forward to being able to go on e.g. a family roadtrip in my own autonomous car, and not have to keep my eyes on the road at all times instead of engaging with my family. (Or surfing the web, watching a movie, etc.) In other words, my autonomous car will replace my own unpaid job of staring at the road for hours on end. I think that's a pretty nice and important feature.
Yes, a level 4 (or especially 5) car would be very nice, and probably far safer than human-piloted cars. Even without the benefit of swarm communication, cars being able to look all around all the time should improve safety. As an old codger, I find that my processing power is not what it used to be, but I remember some white-knuckle driving from decades ago where it would have been nice to see all around all the time, instead of swiveling my head around.
Swarm communication should, in many locations, eliminate the need for traffic lights and stop signs. With accidents reduced by 90%, people could feel safe in much smaller vehicles that could get 10 miles per kilowatt hour instead of 3-4. (In fact, I have such a vehicle in my shop.) Smaller vehicles also consume less energy in their manufacture, and take up less space on the roads, so traffic could flow more freely and some asphalt parking acres could be replaced with trees. Owners could spend travel time reading, catching up on work, enjoying the sights, chatting with other drivers...
Imagine if Tesla had a leader interested in 1) making really good electric cars 2) outfitting those cars with good driver assist features to include wipers that work properly and TACC that functions correctly, 3) providing good ergonomic solutions instead of distracting bells and whistles, 4) building the infrastructure to feed these great cars, 5) making restitution to all the owners who paid for fake full self-driving software, 6) providing great customer service, and 7) starting work on developing a viable plan to develop a level 5 swarm-enabled personal car, probably working with a consortium of car companies, universities, and government entities.
I could get enthused about that.
That sounds better to me than a come-from-behind competition with Waymo, Zoox, Argo, Motional, and the rest.