Except you'd need to map "everywhere" in high-fidelity 3D, save it somewhere in the car, and have it accessible near-realtime. The real reason Waymo can't service "everywhere" is that their approach doesn't scale.
And don't get me wrong - it's clearly a better service where it works (at this point in time), but it'll realistically only ever work in pre-mapped cities. Which of course would remove a ton of drivers and accidents, so still a win.
"How do we get self-driving into millions of regular consumer cars without doubling the price by adding expensive sensors and redesigning the vehicle around huge chunky sensors"
Waymo is focused on Taxis for a reason because it's likely going to be unaffordable, except to people driving Lambos. But that also may be fine for a big chunk of the public who don't want to own cars (or want to rent their own as a service).
Some consumer car companies are experimenting with adding smaller LIDAR sensors like recent Volvo Ex90 which costs ~$100k. But they aren't as sophisticated as Waymo
Back then Comma AI did the same thing choosing cheap cameras, Geohot the founder is similar to Elon being someone who prefers bucking trends and taking risks. He made a good public argument around that time for using Cameras and betting on more data = better output. And I guess Elon thought so too. It also happened to involve less hardware engineering and let them focus on software. Which is an easier decision as a business.
If it doesn’t work out I’m sure it will die a quick death and LIDAR will take over. State regulators won’t tolerate it long.