I guess my argument against the aggressive timeline is that I have been working in MV/AI for the last decade or so, and am pretty deeply connected into NVIDIA, AMBA, a few SS lidar companies and so forth (though I do not deal with autonomous driving, I am in physical security/surveillance).
While there has been a lot of progress made to date, much of it (IMO) has been around the low-hanging fruit kind of stuff. Sure, we've mapped a lot of roads and covered many of the basics. But none of the systems are really able to do human-level context awareness, like you see a bunch of kids running around near, but not on, the road a few hundred feet up, might want to slow down or at least be extra aware, or people clearly driving in "tourist mode", etc.
My personal assessment is that we are still 20+ years away from the point where the human driver is mostly along for the ride (and thus to the point of this thread, can fiddle with a poorly implemented touchscreen UI without risk). I am led to believe the ultimate solution is going to involve altering/enhancing the road infrastructure, along with the general improvements for the in-vehicle stuff (which itself has at least a decade of development still to go). Combine those things together, and we're still quite a ways off from the vision that was sold 10 years ago.
So the "argument" you give is your gut feeling when you look at how things are now.
I expected something like this. That's why I said "quantifiable". We humans are animals of habit and have a hard time imagining a changing world. But every time a few decades pass, we look back and see - damn! - a lot has changed.
A way to get away from feeling and towards a rational prediction is to look at rate of change. In my experience, doubling the performance of a new technology (self-driving is only 13 years old) is usually doable. And that means 1000x improvement in 10 years. And from where we are now, that gives us Waymo in every major city and Tesla with super human capabilities.
So your response is essentially your gut feelings then?
I do not think your rate of change assessment is accurate. With many newer technologies, like the DNN/CNN frameworks, we see a more or less immediate order of magnitude improvement, and then declining rate of improvements as the easier bits are done and we are forced to do deeper development or refinement.
Saying that Tesla has super human capabilities sounds to me like you are not assessing it rationally. Super human would mean, to me, that it manages to do better than humans in extreme conditions. Instead, we have seen a number of very preventable causalities in the current generation systems.
While there has been a lot of progress made to date, much of it (IMO) has been around the low-hanging fruit kind of stuff. Sure, we've mapped a lot of roads and covered many of the basics. But none of the systems are really able to do human-level context awareness, like you see a bunch of kids running around near, but not on, the road a few hundred feet up, might want to slow down or at least be extra aware, or people clearly driving in "tourist mode", etc.
My personal assessment is that we are still 20+ years away from the point where the human driver is mostly along for the ride (and thus to the point of this thread, can fiddle with a poorly implemented touchscreen UI without risk). I am led to believe the ultimate solution is going to involve altering/enhancing the road infrastructure, along with the general improvements for the in-vehicle stuff (which itself has at least a decade of development still to go). Combine those things together, and we're still quite a ways off from the vision that was sold 10 years ago.