> Creating a self driving car which can navigate a dark, rainy parking lot is something we do not know how. Not even in theory.

I do not recognise the reality you are in.

Navigating dark and rainy parking lots is not hard. At all. Not hard in theory, and not hard in practice either.

We have lidars which work very well in rain. We have cameras with excellent dynamic range. Parking lots are slow environments where everyone moves slowly and you are generally allowed to stop if you are suddenly spooked or need a bit more time to check things.

There are hard problems about self driving cars, but dark and rainy parking lots are not the stumbling block.

Where do you even get this idea?

Let me tell you two harder things about self driving cars: "How many nines do you want in your certainty that the car won't hit anyone?" and "How do you want that proven? With stats or with fault tree analysis, or a mix of both?"

You have the definition slightly wrong.

Level 5 driving just means it can drive a child or blind person around who can’t take over in an emergency. That’s the difference between Little Boy and a Moab.

However current level 5 cars aren’t something a consumer would buy. A car that refuses to drive in 99% of situations isn’t marketable. The minimum threshold for that might be a car that can only drive in Hawaii, not a large enough market to pay for R&D, but still plenty use for a blind person living in the area.

I don't know if I agree with that particular example. You only need to go 3-5mph in a dark, rainy parking lot.