> Bay area also has historically mild climate

Bay Area yes. San Francisco less so. Sure it doesn’t get snow but it is often rainy and foggy.

Beyond that, the drivers are very aggressive along with pedestrians and cyclists. It is not a very easy place to drive.

Are you suggesting that they should start development in a more extreme climate right off the bat?

Waymo has even developed custom localized climate models for San Francisco: https://blog.waymo.com/2022/11/using-cutting-edge-weather-re...

Yes. What's the point of developing something that is suitable for sunny climate only? This is how bias creeps into Al/ML systems.

You start with the easy case then gradually make the model more robust by increasing the difficulty.

What you’re suggesting is akin to trying to develop a jetliner before building propeller aircraft. Too hard too soon and you don’t really know what the target even is. Much learning happens before you can successfully execute a complex plan like that.

Plus it’s better from a business standpoint because you can achieve probability profitability sooner with your MVP.

This doesn't suit the narrative that self-driving cars are close to taking over. Up in the northeast US, drivers are actually a lot better at controlling their vehicles than bay area and southwestern drivers thanks to the fact that everyone has to drive in ice and snow.

There will always be situations where self-driving vehicles are not appropriate, but those situations will continue to decrease.