If my anecdote is useful for anything, my car drives about 90% of my 25mi commute today on its own, including leaving one highway and going over a ramp to then merge into another highway. Needless to say that I’m extremely happy with what has been delivered so far.
I think a careful interpretation shows how far that is. A self-driving system ought not be considered reliable until it can drive O(100 million miles) without disconnecting once, in order to match human reliability (that's about the distance between fatal accidents currently). A guaranteed disconnect within 25 miles is many zeros of missing reliability.
A Disney park engineer once relayed to me the philosophy for designing safe attractions in the parks: "If there's a one in a million chance of it happening, it'll happen multiple times per year," given attendance numbers which are in the millions.
A self-driving car needs to handle ordinary commute circumstances with 100.0% reliability, and one-in-a-million circumstances (which statistically you will have never personally encountered) with reliability literally above 99%.
I don't know many humans that drive 100 million miles without disconnecting. The disconnect thing is more analogous to pulling over to check the map for which 5000 miles might be ok.
If the computer needs to pull over and check itself into a motel and nap for a few hours, I'm not going to object. But right now the computer "takes a break" by simply disengaging in the middle of the road. Presumably when this happens, the computer isn't even confident in it's ability to safely park the car on the shoulder, as a human driver would attempt in the event of an emergency (blown tire) or undrivable conditions (which for a human might be an intense blizzard or torrential downpour.)
This is the comment thread beginning with "Musk claims 2020 for sleep while the car drives". I think the point of this discussion is that Musk, has been implying that.
I am expecting significant improvements as more and more vehicles start providing data for the entire system to learn and I don’t care at all if it takes longer than that. I don’t want the technology to fail just because Musk is overly optimistic.
And if your trip on autopilot is 99.9% reliable, how would you feel about that stat?
Do you find yourself paying attention to the road context the entire time? I'm curious if your mental acuity has dropped over time as the car has driven you.
You start trusting it more as you use it and learn its flaws, so you can anticipate when it may do something stupid.
I pay attention the entire time, but it is a much more relaxed experience. I’d still pay attention the entire time even if it were supposedly 100%.
It is very sad that a lot of people in this forum are hoping this never works. It is one of the most exciting advancements in technology that can benefit us all, but people here seem to be more interested in seeing Musk and Tesla fail rather than hoping they achieve this and bring the whole industry forward one more time, affecting millions of lives.
That is not what is happening here. People are rightly skeptical of the claims. As Theranos showed, you can't "fake it till you make it" when people's lives are involved.
> It is very sad that a lot of people in this forum are hoping this never works.
Top level OP here. I am personally rooting for self-driving to succeed and catch on. I just don't think Tesla's current strategy is likely to work, and their cavalier, unjustified overconfidence is either going to sink them or kill people, or both, neither of which is good for the future of self-driving.
