I think Tesla is taking the correct approach, treating it as an AGI problem, and training their system to be antifragile.

I think Waymo will make money off of their system sooner.

Why is that? and what are the two approaches?

Waymo is trying to solve the environment. They're HD mapping every centimeter of areas they operate in, so that they can let their vision systems focus on deviations from that single source of truth. It's a reasonable approach, but it has some significant limitations, especially in terms of infrastructure.

Tesla, on the other hand, is trying to make their cars work anywhere. It's a larger problem space, but the end goal is a far more robust product, with much lower long-term infrastructure spending.

I personally like the robust approach, but I think both approaches are long-term viable, albeit for different use-cases.

> but the end goal is a far more robust product What about Tesla's approach makes you think it is inherently more robust?

Tesla does not want HD maps. They want to be able to place the car somewhere it's never seen and have it drive itself competently. That's by definition more robust than requiring HD maps to navigate anywhere.

What you're referring to is overall coverage not robustness. In fact, placing your car somewhere you've never been before and operate it seems less robust than one that already has HD maps.

ro·bust | rōˈbəst, ˈrōˌbəst | adjective (robuster, robustest) • (of a process, system, organization, etc.) able to withstand or overcome adverse conditions: California's robust property market.

You think this is clever, but you actually come across as though you have no clue what you're talking about. We are operating under the same definition of robustness, can't believe that your first reaction is to act like I don't understand what robustness means.

You on the other hand, have no explanation whatsoever as to why you think a non-HD map system is robust everywhere compared to a HD map system. If you claim that HD maps are less robust by definition, you clearly do not understand deep learning or computer vision.

Thanks!