If I installed the CAN bus hack, I’d have that system today. No, I’m not going to do that. But for a wide variety of routes (including my work commute), FSD drives perfectly today with zero interventions.
The regulatory question is a big one. I’m not expecting wide deployment of fully driverless cars in the near future. But ones that can at least fail safe in almost all situations are getting close, IMO.
I don’t think people quite realize how good FSD is right now. It beats Waymo already in many (but not all) respects.
Oh, you mean a specific person, one with an easy commute, an absence of inhospitable weather, and no complicated parking lots to negotiate. Well, sure, L4 is here already, I’ll grant that. Somewhat.
The city of 400,000 pr so where i live is like this. For many, FSD would be a god send, allowing something better to do than a boring commute in rush hour traffic
In the sentence immediately preceding the one you quoted, I mentioned Level 5 as a contrast to what I was talking about. So yes, obviously I meant something short of that, although really I think the level system is somewhat dubious in the first place. And in the (second) immediately following sentence, I specifically said “[not] every route” or the endpoints.
It will have to be something more than the Mercedes L3 system. That can’t drive curb-to-curb or any variation on that. It will have to be what I have today with FSD, except with interventions driven down to a point where I don’t have to actively monitor the road, and where it will tell me with ample warning if I need to take over.
Ed Niedermeyer, interviewed in the podcast linked below, defines level 5 as when the manufacturer takes liability, rather than the driver or car owner having liability. He said it as an aside, but I think it’s as good a definition as any.
Ironically, I was listening to the podcast while FSD was driving my Tesla.
It was interesting, even if he did fall into a few of the same holes that people (including in this thread fall into). First, is criticism of FSD and autopilot from someone who has little experience using it. Second, is letting (justified) hatred for Musk get in the way of rational criticisms of Tesla and their software.
His primary argument is that FSD and autopilot are just versions of speed detecting cruise control and lane keeping, with a bunch of marketing layered on top claiming that it is self-driving, rather than just driver aids. Tesla/Musk make these claims as a way to distinguish their cars in the marketplace. Even if there is some truth to that, I think there are lots of things wrong with the argument, so don’t @ me about it, but feel free to yell at Ed & Ed.
Then there is lots of talk about how many people FSD and autopilot have killed. I’m totally willing to listen to evidence for this. The number is supposedly over 400. However, all of the well publicized cases I’m aware of where the investigations have reached a conclusion have pretty much exonerated FSD. The driver was drunk, watching a movie, or FSD wasn’t even engaged prior to the crash.
A case I would blame on FSD is something where the driver’s attention wavered for a split second, and FSD made a fatal mistake in that time. Not something where the driver was distracted for extended periods of time, and not something where an attentive human would also have had an accident (for example, a pedestrian emerging from behind an obstruction into the path of the car).
Finally (for me), they predict that the NHTSA will at some point ban FSD, not just require a name change and a few more steering jiggles.
It’s one thing to be bad at predictions… a certain someone has that quality as well. But to be wrong that consistently takes real talent.
I don’t think the “liability” definition is particularly good. Tesla will take liability when your car is part of the robotaxi fleet, but that doesn’t require L5. On the other hand, auto manufacturers never take responsibility for anything not related to a product defect, so I see no reason why anything would change here.
I wasn’t aware of his background, but that is interesting.
My biggest complaint I had is what he didn’t say. Maybe a word or two about Tesla dragging the car industry into electrification. A big tech lead that Musk seems to be squandering, but very far from his insistence that it was all something worse than a scam (because it causes deaths).
Also all of the doom saying about drops in Tesla sales, ignoring the fact that they made the best selling car in the world last year, even if it is selling somewhat less than now that, and repeating the headline that is more truthfully written as “parking lots can be seen from space.”
My reactions are straying from the topic of this thread, but the podcast is very much in line. Unfortunately the podcast is [Tesla = Musk = Self driving], while I would say [(Tesla \approx Musk) \ne Self driving].
I do generally like Ed Zitron’s podcast though. He is rightfully angry at many tech things in generally well researched ways. This particular one it was almost completely Niedermeyer talking.
FSD (by whatever name) is such a seductive invitation to misuse (e.g. DWI or movie-watching) that it must be presumed to be used that way most of the time by most of the users. Unless it was proven to be fully switched off at the time and the driver was lying about FSD driving, any misuse is in fact use and therefore FSD’s fault.
It almost becomes what the law calls an “attractive nuisance”, something with legit uses but that also contains an inevitable downside that requires active precautions against. The classic example being an unfenced backyard swimming pool that the neighbor’s kid drowns in. Kids will jump into pools, and pools inevitably contain drowning depth water, so fences are needed and the absence of a fence is per se negligence on the part of the pool owner.
In reality, the argument is much more specious than that. In the law they call that “but for causation”. The theory that “But for [this], [that] would / could not have happened. So therefore [this] is at fault.” In the law that’s also recognized as total BS; not a valid argument.
In FSD’s case the argument amounts to “If FSD hadn’t been engaged (or installed), the person would not have been watching a movie when the car rear-ended a semi. Therefore FSD killed him”. Pure BS. But persuasive BS when laid out more slowly in a news article or hatchet-job video.
What do you mean it’s “clear BS?” The argument isn’t that people can’t help but watch movies. It’s that a system like FSD that almost provides autonomous driving except in rare situations causes people to lose focus. They are physically not able to react to something as quickly as if they were actively driving.
This is not some unanticipated consequence of people using the feature incorrectly. It is completely expected, and is much more analogous to the unfenced pool than you state.
And it’s actually worse than the pool, because we have the fence solution for that situation. I can’t even conceive of a solution to an almost capable self driving system other than somehow building more responsible humans.
Any system will be circumvented. No system is completely idiot proof. (Generally). Around 100 people per day die in car accidents in the US. Nearly all of those are in non-autonomous cars.
Per mile, which system causes fewer injuries/deaths/accidents? That’s all that matters. cars are always going to kill people. Which is the safer system?
People are unable to pay attention when they are actively driving. All of the arguments about vigilance tasks also apply (to some degree) to actively driving.
This is the real question, and currently we don’t have data to answer it. Tesla probably has data to answer it, but they really don’t release enough to trust the answer. The government is maybe just starting to wake up to the need to answer the question.
People are terrible drivers. Is automation an even more terrible driver, or a somewhat less terrible driver? If it is a more terrible driver, is there a path for it to being a less terrible driver?
… and discard the McD baggy, the empty Red Bull cans and the used condoms in the rear seat…?
In Austria a car rental that focussed on Teslas just went bankrupt, for this very reason … certain strata of people rented the Teslas on the weekend to race responsibly drive them to the clubbing scene and have “a good time” with others…
Regular users complained about the weed smell, trash and assorted bedroom junk (forgotten underwear!) and ADN-samples left behind from somebody before … and eventually stopped using the company all together.
so … I am not seeing this happen on a wider scale … that’s just Elon talking trying to talk things up … (Teslas will be appreciating assets… yeah right)
If Tesla tells me that they have achieved L5 (for whatever definition), then I’d say they are warrantying that the system will handle whatever is thrown at it (including things potentially actually being thrown at the car). So they should have liability for avoidable crashes (once again, for whatever definition of ‘avoidable’), since the crash is a defect in the system?
If the rental company wasn’t cleaning cars between rentals, then they deserved to go bankrupt.
The one Tesla-specific factor would be people renting them for sex while driving on FSD, so maybe a couple more condoms or panties? I can’t imagine it’s enough to make a significant difference.
But until I see evidence, I’m guessing we’re not getting the accurate story.
During FSD the driver needs their hand on the wheel and to be facing forward. Maybe a blowjob but that’s about it. You can get road head in any car. There’s lots of room in the back for a party with the rear seats folded down but that’s not Tesla specific.
A Corolla isn’t a fair comparison because there’s no room. But I can see that Teslas became more trendy than a SUV or something from a different maker. Regardless, the story is bullshit.