This is a weird amount of hubris for someone who either doesn't understand the goals of current work on self-driving cars or is just clueless about the field.
Tesla's attempts at self-driving capture absolutely nothing besides... Tesla's attempts at self-driving.
There are companies with realistic sensor stacks, not artificially imposing restrictions so they can pre-sell their results, actually focused on self-driving.
A true L5 vehicle is an impossibility because that'd mean handling the frozen tundra equally as well as rush hour in Mumbai, or a wildfire in California. And laypeople tend to really lean into that...
But for a realistic definition of L5, "feasibility" is low bar already proven by major players except Tesla. So to say "completely infeasible" kind of kills credibility of everything following it.
Dismissing "snow" as something that happens only in "the frozen tundra" is fairly provincial. You'd be surprised by how much of the USA (and other industrialized countries) have roads and road edges mostly obscured by snow for weeks at a time multiple times throughout winter. There's a reason almost all autonomous driving tests are done in places like Arizona. And the ones done in winter settings tend to be absolute positioning based (ground penetrating radar maps, etc) or non-human vision modalities which don't actually match how human drivers flock.
> Dismissing "snow" as something that happens only in "the frozen tundra" is fairly provincial.
Who did that? My comment certainly didn't by any reading of it.
> You'd be surprised by how much of the USA (and other industrialized countries) have roads and road edges mostly obscured by snow for weeks at a time multiple times throughout winter
Again, maybe you would, but I'm not.
> There's a reason almost all autonomous driving tests are done in places like Arizona
Yes, because they have good weather. It turns out that compartmentalizing a problem as large as autonomous vehicles is a smart idea.
It's not like you can't work on participation just because the local weather is nice. Regardless of climate we use things like rain and temp chambers, and there are AVs in places with bad weather, but that's not where service gets launched because again, compartmentation is a good thing.
I guess all that is to say, no one in the AV field is unaware that bad weather is a thing, and yet we're still in the problem space.
> non-human vision modalities which don't actually match how human drivers flock
Is this supposed to be a gotcha or did you not realize AV companies are not about blindly to replicate human vision in the AV field?
You'll find Tesla is the only company silly enough to imply nonsense like that.
>Is this supposed to be a gotcha or did you not realize AV companies are not about blindly to replicate human vision in the AV field?
Unless the machines fail the same way human drivers do and go off the absolute positioning based lanes to follow the emergent human based ones they'll cause crashes. Sometimes doing the right thing is the wrong thing because everyone is doing the wrong thing.
Again you're showing that you're not familiar with the field you're talking about.
Being able to follow vehicles in front of you outside of lines is below table stakes in this field, you wouldn't even get into the casino. Mobileye was doing that 11 years ago with nothing more than a OTS radar and a single BW camera.
I mean, again, do you really think saying "humans don't perfectly follow the rules of the road" could possibly be a novel concept to anyone in the field of AVs?
There's nothing like being in a field to realize how far HN hubris actually goes. Someone reading your first comment would think you're a thought leader with how flippantly you write off an entire industry, yet here you are 2 comments in and slowly re-discovering the ground truths that the industry is already built on.
I'm not talking about following other vehicles in front of you. I'm talking about follow the paths the other vehicles, now long gone, have left in the snow/ice/mush.
>do you really think saying "humans don't perfectly follow the rules of the road" could possibly be a novel concept to anyone in the field of AVs?
I do think that driving in cold regions seems to be a novel concept because every time I talk to someone in the know like yourself they latch on to the wrong side of the idea immediately.
Not realizing that we can follow other cars is somewhat understandable. Bringing up following tire tracks in the snow at this stage is so utterly nonsensical, it simply doesn't pass muster when considering the ways to parse the statement.
Millions upon millions of people who can be served without having to checks notes solve following tire tracks on the ground... but that's your current bastion for why self-driving is "completely infeasible"
No good deed goes unpunished apparently, because here you are now wearing the misguided nature of our charitable interpretations as a badge of honor.
Tesla's attempts at self-driving capture absolutely nothing besides... Tesla's attempts at self-driving.
There are companies with realistic sensor stacks, not artificially imposing restrictions so they can pre-sell their results, actually focused on self-driving.
A true L5 vehicle is an impossibility because that'd mean handling the frozen tundra equally as well as rush hour in Mumbai, or a wildfire in California. And laypeople tend to really lean into that...
But for a realistic definition of L5, "feasibility" is low bar already proven by major players except Tesla. So to say "completely infeasible" kind of kills credibility of everything following it.