I don't think anyone in their right mind thinks cameras are best. It's just that lidars are expesive in general, and the type of lidars required for anything remotely close to real self driving is insanely expensive all things considered (I believe the ginormous things on Waymos are a few grand at automotive builder rates if bought by the boatload), and that the hope was to make cameras as good eventually.
This is close to insignificant in the price of say a self driving quarry dump truck that costs millions already, but accounts for quite a chunk in the price of a passenger vehicle.
And this is where reality takes a turn while muskality flies straight ahead on impulse alone, as we are absolutely not there yet, despite real progress.
And then the real achievements of engineers and teams that are genuinely savvy and talented and are making genuine progress gets drowned in the "If a pregnant woman gives birth after nine months, nine pregnant women will give birth after one month" sales pitch. And there's not even nine of them. And it all rolls to Musklandia from there.