Yeah, that's not what fair use means in US law. But I really doubt that a case could be made that this was fair use.

It may be a good decision in some ways, but it's clearly illegal.

No, it's going to need to let YOU navigate the grass lot. That it CAN manage the driving doesn't mean that it has to do it all the time.

Well, the real advantage is that if they guess wrong on which flu strain is going to dominate next year they can change the formula quickly. There are other advantages, but most of them don't affect as many people.

Actually, above about a 20 mph difference in speeds I don't believe that humans CAN gage distance and speed well enough. And I may well have overestimated the minimum speed delta. (Admittedly some of that is reaction time rather than an estimation problem.) We do better with things that are off to one side, because there we can judge based or angles.

I suspect that they are currently better, on the average, than the average human driver, but that they have DIFFERENT edge cases. That would explain both the statistics reported and the news stories.

Sorry, but that's too extreme. It has *less* missing information to infer. And how much less is going to depend on both the sensors and the environment. (And it's also possible to be overwhelmed by information to the extent that you miss the obvious problem.)
I expect that automated cars are better at driving than people are already, but they can easily be oversold.

You are overselling current automated cars. They may be better on the average, but there are lots of specific scenarios where they aren't. (And "better" is a lot short of perfect.)
However, I'd prefer one over my capabilities, even if I just had to avoid the edge cases.

Highways are generally an easier problem to solve, but the errors are more dangerous. And there are some places that are quite poorly designed...so they AREN'T easier in those places. (I'm thinking of one particular intersection where two freeways merge and then quickly split, so one frequently needs to cross 6 poorly marked lanes in less than 1/4 mile or take a 1/2 hour detour through local streets.)

It's not guaranteed that a system with lidar+camera will do better than a different system with camera alone. Certainly not always. That it could, in principle, do better is clear, but it depends on a lot more than just the sensory modalities.

No. It just needs to recognize situations it can't handle sufficiently in advance. "Going off road" is something that should be easily recognizable in advance if you mean doing it intentionally.

That depends on what statistical measure you use. It's statistically significant enough that there's a special law against it, so it's not trivial.

And even that's wrong, because people also use their ears when driving. And even a bit of "sensing vibration". So people don't JUST use their eyes when driving, even though that's the primary mode.

True, but it's almost always faster than the flow of electrons...well, if it goes at all.

They're both quite different from the US, but saying they're about the same is both weird and wrong. The Chinese are rebuilding the imperium, which was sometimes good and sometimes quite bad. The Russians...they seem to be in the process of destroying their country. The reports I have heard are now a couple of decades old, but they indicated that they country was largely run by non-government-based gangs.