> at some point don't you have to get this stuff out in the wild to see how it behaves or else we are resigned to not making progress on this
It's not my problem whether Waymo makes progress on their technology, but they make it my problem when they hit me with their car.
I do have sympathy for wanting to make the progress: we might make better pacemakers faster by testing out 20 designs on 20 people. But nobody wants to be the one that gets the one that doesn't work. And in this case the victims aren't even involved. They didn't opt-in to beta testing the road full of half-baked robots, they're just trying to get to work.
In the interest of having a discussion, let's assume that AVs are a meaningful goal to work towards for whatever reason. But given that, how do we get vehicles from the drawing board to actually being viable products without on-road testing? I've never seen a complex product that went from non-existent to perfect in the first deployment, so it doesn't seem realistic to expect that here.
Instead, they should be developed iteratively, with design prototypes that proceed from closed-course testing to supervised public testing to closed course autonomous testing, to on-road autonomy over the course of many years. This is what Waymo did. There's a reasonable argument to be made that they did this too quickly, but I can't reconcile that with your argument that they shouldn't have done it at all.
In an ideal world, there'd also be effective government oversight and public safety monitoring at every stage of the above process. Regulators haven't stepped up to do this, though AV companies have done quite a bit to stymie the oversight process as well.
Sadly, I come only with problems and not solutions. I take it as axiomatic that beta-testing with peoples' lives that didn't agree to do so is unacceptable. That closes off a lot of the solution space that you're proposing. That sucks and you're free to disagree but again I take it ethically unassailable.
Teleportation would also be a societal game changer but if the only way there is to beta test it on unwilling participants I'd also believe that, well, we just don't get teleportation then.
It's up to Waymo to figure out how to get there, not to me. I do not take it as axiomatic that just because it'd be useful that the ends justify the means. And it certainly isn't up to Waymo whether you or I can be sacrificed.
And Waymo and Cruise are only rolling out gradually into settings they have reason to think they can handle. And unlike for humans, every time they learn a new lesson, the improvement spreads to their whole fleet. Humans lack that ability -- in a sense, we're the ones who are untrainable.
(If you want to point at Uber before they stopped, I agree they were irresponsible.)
It's not my problem whether Waymo makes progress on their technology, but they make it my problem when they hit me with their car.
I do have sympathy for wanting to make the progress: we might make better pacemakers faster by testing out 20 designs on 20 people. But nobody wants to be the one that gets the one that doesn't work. And in this case the victims aren't even involved. They didn't opt-in to beta testing the road full of half-baked robots, they're just trying to get to work.