You can almost set your watch to this stuff pic.twitter.com/UrIwG4AP2m
— The okayest poster there is (@ok_post_guy) October 26, 2022
The image on the right:
… The U.S. Department of Justice launched the previously undisclosed probe last year following more than a dozen crashes, some of them fatal, involving Tesla’s driver assistance system Autopilot, which was activated during the accidents, the people said.
As early as 2016, Tesla’s marketing materials have touted Autopilot’s capabilities. On a conference call that year, Elon Musk, the Silicon Valley automaker’s chief executive, described it as “probably better” than a human driver.
Last week, Musk said on another call Tesla would soon release an upgraded version of “Full Self-Driving” software allowing customers to travel “to your work, your friend’s house, to the grocery store without you touching the wheel.”…
Tesla, which disbanded its media relations department in 2020, did not respond to written questions from Reuters on Wednesday. Musk also did not respond to written questions seeking comment. A Justice Department spokesperson declined to comment.
Musk said in an interview with Automotive News in 2020 that Autopilot problems stem from customers using the system in ways contrary to Tesla’s instructions.
Federal and California safety regulators are already scrutinizing whether claims about Autopilot’s capabilities and the system’s design imbue customers with a false sense of security, inducing them to treat Teslas as truly driverless cars and become complacent behind the wheel with potentially deadly consequences.
The Justice Department investigation potentially represents a more serious level of scrutiny because of the possibility of criminal charges against the company or individual executives, the people familiar with the inquiry said…
The criminal Autopilot investigation adds to the other probes and legal issues involving Musk, who became locked in a court battle earlier this year after abandoning a $44 billion takeover of social media giant Twitter Inc, only to reverse course and proclaim excitement for the looming acquisition…
I think often about that guy in the NY Times/Hulu documentary on musk who was a huge fanboy and trusted the company's claims right up until his "self-driving" tesla couldn't see a tractor trailer in the middle of the road, at which point he, well, died
— Karl Bode (@KarlBode) October 26, 2022
Now that Tesla faces a criminal probe over it's autonomous tech, this great recent article argued that all self-driving tech firms are basically scams at this pointhttps://t.co/nck45bTsHb
— vocational politics stan account 🫳♨️ (@Convolutedname) October 26, 2022
the conclusion is that self-driving tech has one major problem that it simply cannot overcome, and that the companies just fudging their demos to get around this central limitation pic.twitter.com/j5tWak8tBw
— vocational politics stan account 🫳♨️ (@Convolutedname) October 26, 2022
'Elon Musk's Crash Course' was the documentary. It was far too kind in terms of perpetuating the supergenius engineer mythology, but it's an interesting watch regarding the early crashes (there's of course been many, many since then)https://t.co/VaQtRXQy3u
— Karl Bode (@KarlBode) October 26, 2022
But look, over there! I am about to revolutionize social media!…
guy has no genuine idea how the platform he's about to overpay for works, and neither do any of his loyal bitcoin fluffing earlobe nibblers
going to be a very keystone cops esque few years around here pic.twitter.com/JSt2dmC5gS
— Karl Bode (@KarlBode) October 24, 2022
Of course, Musk is not actually an Afrikaaner — he’s a soutpiel:
people in the 80s and 90s complained nonstop about faceless corporations so now corporations have a face and its a 55 year old afrikaaner who communicates exclusively in reddit memes
— Promo Code: Rudy (@canderaid) October 24, 2022
War for Ukraine Day 245: The Russians Have Reached the Sloppy Agitprop Stage
Goku (aka Amerikan Baka)
Woah, this sounds like a big deal. Tesla stock must be having a bad time. And yeah, I’ve assumed that the self-driving tech was never going to become fully realized how people like Musk have claimed. The best it will ever be capable of is likely stuff such as self-parking and some driver safety assistance such as braking.
Wasn’t it in like 2015 or so that news articles/enthusiasts said that fully autonomous vehicles were only 5 years away?
Most of the electric vehicles I see in my area tend to Teslas unfortunately, but hopefully that will begin to change. It was around 2020 that I started to notice their prevalence
NotMax
@Goku (aka Amerikan Baka)
Keep your eyes peeled for the Spectre.
Memo to Rolls: Isn’t that the organizational nemesis of James Bond?
;)
James E Powell
@NotMax:
The Spectre looks like a futuristic car design from the 1950s.
James E Powell
@Goku (aka Amerikan Baka):
Don’t remember how long Elon’s been promising it or exactly when he said Tesla would make it happen. Duncan Black at Eschaton keeps track & occasionally posts reminders.
Anne Laurie
‘Mid-Century Modern’ design is very chic right now.
There’s an actual theory that ‘retro’ or ‘revived’ fashions become cool when the people who were fascinated by the “cool” stuff their parents wore / used are old enough to have their own money to spend on it. Late Boomers who watched early James Bond movies and The Dick Van Dyke Show are being courted by advertisers looking for their retirement-crisis money, I’m guessing…
We’re spending our childrens’ inheritance — on an electric Rolls Royce!
Origuy
I have a Honda Insight which has a limited driver assist capability. It has a few glitches. If you get to close to an object such as a car or person, it will apply the brakes on its own. I had it enabled on the freeway when a mylar balloon drifted across my lane and the brakes slammed on. That’s only happened once. The system that detects when you drift across a lane marker can be fooled if you’re on a street which has been recently widened and the old lane markers are still visible. The steering wheel gives you a little more resistance when you cross a lane marker without signalling.
prostratedragon
A Spectre Haunts Europe, silent era horror movie made by the Ukrainian Soviet movie unit. Says in the article, based on “The Masque of the Red Death.” As a youngster I remember some spooky sounding fellow on tv who adapted that title to his anticommunist programming.
Amir Khalid
@NotMax:
Not to worry. James Bond is an Aston Martin driver.
danielx
@Origuy:
Got a Mazda CX-5 that does all that, including a really annoying Driver Alert when the vehicle thinks one has been driving too long without a break.
Ladyraxterinok
@prostratedragon:
Was that Fred Schwartz and his Christian Anti Communist Crusade?
I remember hearing him on the radio in high school in the 50s with his voice of doom
There’s a 1974 Atlantic article on the internet about him and his organization
NotMax
@Amir Khalid
Cars we don’t need — the Rezvani Vengeance.
NotMax
@Anne Laurie
Kidney-shaped occasional tables made of blond wood coming back in vogue?
Say it ain’t so.
sab
@Anne Laurie: Yay. Maybe now I can sell my parents teak Danish modern bedroom suite (vintage 1950.)
rikyrah
Self driving has always seemed like a dangerous scam😠😠
sab
@rikyrah: Yes! I am still waiting for the hover cars we were promised (I am 68) but if you don’t want to drive take a cab or a bus.
The Thin Black Duke
In spite of Elon Musk’s hallucinations, I’d prefer riding on Amtrak to a self-driving car.
NotMax
@The Thin Black Duke
Wow, looky what bobbed to the surface of the memory pond. Lurve me some Tracy Nelson.
Take a plane or a train
It’s all the same
Just get out of here
Take a bus if you must
Make sure you just
Get out of here
;)
Brachiator
@Goku (aka Amerikan Baka):
Fully autonomous vehicles are always 5 years away.
VR and driverless cars are two technologies that seem very hard to get right. And these two have some overlap in LIDAR and similar tech.
Some of this tech is already providing improved driver assistance, and some group might find a way to make it truly workable one day.
Gvg
@sab: yes you can. My parents are thrilled and quote estate sales prices. They tell me so I won’t sell too cheap when they die by which time probably something else will be in vogue. Naturally condition matters and not all mid century is actually equally in vogue but danish modern definitely is. I like it myself and have a nice dining set.
Keith P.
I could never believe that the DOT was letting Tesla et al essentially beta test driving software with real drivers on real roads. NASA can’t risk lives like that when it’s in bunches of 5.
Gvg
@Brachiator: For a brief period this blog was sure self driving trucks were going to put truckers all over the country out of work in just a few years. I was doubtful then but there were long discussions about how to retrain them and the social impact on society. Then other events came up and then the crashes and difficulties became more obvious.
To do self driving cars, I think we would have to build roads designed for them just for starters. I don’t think it’s going to happen in my lifetime.
Baud
@Gvg:
Not just truckers. People were hot and bothered about automation generally.
Steve in the ATL
@Baud: I feel bad about all the A/R clerks who lost their jobs when I started using automatic bill pay. Maybe they can be retrained as coders.
Eolirin
This stuff will eventually get solved, though we may need new techniques to be developed. (To note: the quote about not being able to recognize slightly different pigeons is actually bullshit, the lack of contextual understanding is a real problem, but the AI models do in fact generalize to some degree, and them generalizing incorrectly is a bigger problem than identifying different kinds of birds would be; mistaking a small child for a bird and not breaking is a bigger danger, and resolving that will take more work than knowing, broadly, how to deal with birds) And the only way it’s going to get solved is if it’s being worked on under realistic conditions, but what Tesla is doing is incredibly reckless.
No one else is skipping working out the problems before rolling out to consumers. Consequently everyone else working in this space have way fewer accidents. Fewer accidents than humans, per mile driven, even, though they’re in more controlled settings for a reason.
I don’t think it can be stressed enough how bad humans are at driving, and how much unnecessary death is caused by widespread car use, it’s about on par with flu deaths in the US, a number that could be brought down substantially if people just wore goddamn masks, but we seem to like killing people unnecessarily here.
This tech is important if we’re going to keep having car centric transportation. It may take over a decade to actually make it work, but we really should keep at it until we do, if we’re not going to spend what would probably be trillions in rebuilding our entire transportation infrastructure. (Which would be a better solution, but the hardest of lifts politically)
It also has the potential to significantly improve transportation independence for people with disabilities who can’t drive.
Let me tell you how inadequate getting a cab is as a solution to that problem. Especially if you need things like wheelchair accommodation. It’s very easy to take for granted the privilege of being able to go places when you want instead of when you can schedule a ride. And to not have to worry about how long you can spend at your location, or how long you’ll be stuck waiting there.
Uber isn’t a solution for people with special transportation needs and is too expensive as a regular form of transportation for low income people, which is a lot of the disabled community. Even with car ownership being expensive in and of itself, it’s much easier to envision government subsidization of a one off capital expenditure if the prices can be brought low enough than perpetual taxi vouchers in amounts adequate to need.
Baud
@Steve in the ATL:
I blame WordPerfect for the loss of legal secretaries.
Steve in the ATL
@Eolirin:
Indeed. Will this new technology keep slow cars out of the left lane?
Steve in the ATL
@Baud: gone the way of buggy whip makers!
Credit to WordPerfect—“reveal codes” is the greatest tech invention of all time
different-church-lady
I’ll say it again: when the aliens land and wonder how our civilization disappeared, the answer will be, “The monkeys got to be too clever for their own good.”
gene108
@Steve in the ATL:
It’s not the A/R clerks that are out of a job. Someone still has to match the payment to the correct invoice. Computers can sort of do this, if everything is set up perfectly, but very little in this world is perfect.
The people put out are the folks who worked at banks, in a back office, and ran all the checks people deposited into the bank’s system, so the check can clear. I have an aunt who used to do this. She retired. It was basically a night shift data entry job processing the checks the bank received.
different-church-lady
@Eolirin: We can’t even make auto-correct work right, but we’re gonna nail autonomous cars?
Eolirin
@different-church-lady: Better than humans (which is a low bar), and eventually? Yes.
Things are always impossible until they’re suddenly not.
Matt McIrvin
@Goku (aka Amerikan Baka): There was incredibly rapid progress happening in this area for several years. The problem is that autonomous vehicle driving is the kind of thing where the last 1% of the problem requires most of the effort, and the practical difference between 99% and 100% is immense and deadly.
different-church-lady
I’ll be Elon Musk has never seen a Warner Brothers cartoon.
Bruce K in ATH-GR
I look at it this way: there are a lot of tasks for which pure logical deductive reasoning, the sort a modern computer is good at, just doesn’t cut it. It’s one thing to cut down the number of required operators to a small number (such as the reduced crew of a modern supertanker, or the elimination of the need for a flight engineer on modern airliners); it’s another thing entirely to eliminate those operators entirely, even as a final backup system.
Eolirin
@Bruce K in ATH-GR: These kinds of AI techniques don’t work on logical deductive reasoning at all. They’re trained pattern matching systems, fed enough data to make predictions, and tuned by examining their outputs for correctness.
What they’re bad at is doing tasks that don’t have easy to determine correctness or that rely on impossible to structure data. Driving doesn’t really fall into this category.
lowtechcyclist
@Steve in the ATL:
Damn straight! I remember being pissed when my org switched to Word because Word didn’t have any equivalent at the time. (I assume it does now, just buried somewhere in its absurd menu structure.)
Bruce K in ATH-GR
@Steve in the ATL: @lowtechcyclist: As far as I can tell, Microsoft Word still doesn’t have the discrete coding that WordPerfect had. It’s a pain in the butt to try to tweak the formatting in Word – you’ve got to out-guess the machine, it feels like.
WordPerfect’s format code structure (with the open and close tags) bears a striking resemblance to the structure of HTML formatting, as far as I can tell.
twbrandt (formerly tom)
I think it’s significant that Ford is closing down its autonomous vehicle program, called Argo, that it co-owns with VW. Ford doesn’t see it working in the foreseeable future.
Citizen Alan
@Steve in the ATL: i am still both frustrated and amazed thar Word obstinately refuses to include a Reveal Codes equivalent.
lowtechcyclist
This part always boggled me, that somehow nobody was saying, “no, you can’t use our roads and highways as your test course for this untried system, there’s real people out there that you could kill.”
I think Duncan Black’s critique of self-driving cars has been pretty much spot-on. One point he kept making was that taxis were the dumbest possible application of self-drivers, because the systems would be exposed to the widest possible set of challenges, and they’d screw up a lot.
If self-drivers were going to work somewhere, the first and easiest place to try them would be on a repetitive route where they’d be on the same road every day, encountering the same shit every day. Something like a shuttle between the airport terminal and the parking garage.
I figured they weren’t going to work soon, but I’d hoped that they’d be functional by the time I got into my 80s and it was time to give up the car keys. That’s maybe 15 years away for me, but I really don’t expect that they’ll have figured out the AI to master self-driving by then.
One hope I’d have for self-driving cars is that they’d reduce the amount of car the typical family would need or want to own for themselves. I really don’t believe the theory that with self-driving cars, hardly anyone would own a car anymore: think of all the stuff you have in your car, ready and waiting for you when you go somewhere.
But the way things are set up now, when we buy a car, we have to think in terms of the max capacity we’d need on any sort of periodic basis. If you need an SUV one day every week or two, but in between, all you need is enough room for you and a briefcase or bag of groceries, you need to own an SUV. But in a world of self-driving cars, you could rent that SUV easily on those days you needed it, because the SUV could drive itself to your door. And you could get by with one of those two-seater ‘smart’ cars the rest of the time.
Another Scott
@lowtechcyclist:
Some else made what I think is a compelling argument: everyone not owning their own cars but instead calling a self-driving car whenever they need it won’t really solve the problem of too many cars – in fact it might make it worse. Think of airport car rentals – either the lot is full, or it’s empty. Cars sitting around in a lot cost the companies money, so there’s always the pressure not to have too many, but just enough for their optimum payback. That means that they’re going to err on the side of not having enough cars for the top of their demand curve (so that they can potentially charge more there). People who need a car NOW NOW NOW are going to be at the mercy of vendors who don’t have extra capacity and will charge more for the privilege.
Plus, there’s the needed parking spaces for all those self-driving cars sitting around during the work day, close enough to the users, ready to quickly meet that surge demand at “rush hour”.
Objectively, as Martin points out, if we’re smart we’ll start phasing out cars in cities and (my hope) push to advance things like enclosed electric bikes with stability and traction control and safety features and bike lanes to keep them away from necessary traffic (like delivery trucks). It’s going to be a long transition, so we need to get started.
Thanks.
Cheers,
Scott.
Eolirin
@Citizen Alan: Word has had Reveal Formatting since 2007, I believe. Shift+F1. It’s not exactly the same, but it fills the same role. And it lets you do formatting comparison between sections of selected text.
Frankensteinbeck
@Gvg:
And I was as polite as possible with my “You don’t know what you’re talking about”s. The problems Tesla is having are baked in. Machine learning can’t handle ‘edge cases’ and all you can do is rearrange what those are. Plus, some of those ‘edge cases’ are utterly, unpredictably bizarre. We have vast, extraordinary amounts of data we can feed chat bots, and we can’t pass the Turing Test. For something like driving, where the input approaches infinitely complex around the edges and mistakes are unacceptable? It won’t happen until we get some new technology.
J R in WV
@danielx:
Our CX-5 is like that too, hate that part. And the speed control slows down for curves unnecessarily, and slams on the brakes too late to help if there had actually been something to hit.
I would keep it off if I could have cruise control without it.
Jinchi
It’s a lot like fusion that way.
Jinchi
Same here. I could probably teach my 12 year old to drive reasonably well, but if I had him drive himself to school every day he’d get pulled over by the cops and they’d confiscate the car, because he doesn’t have a licence to drive.
But if pull over a driverless car for a moving violation, it…. just drives away again.
“No citation was issued during the traffic stop,” the police department said.
Curtis
@Goku (aka Amerikan Baka): They’ve been saying self-driving cars are 5-10 years away for the last 25 years.
dnfree
My 2018 Mazda has that cruise control that responds to cars around you. It’s very helpful 99% of the time, and you don’t even notice that the car ahead of you slowed and your car responded. But you can’t take your hands off the wheel or your eyes off the road. When the car sees a situation it doesn’t recognize, it turns the cruise control off, and you suddenly find you’re coasting. The situation I have most often encountered that in is when I’m in the left (fastest) lane of the expressway, there’s nothing ahead of me, there’s nothing beside me in the lane to the right, but in the SECOND lane to the right, there’s a semi truck that’s casting a shadow into my lane. The car doesn’t know what to do, so it slows. As far as I’m concerned, that’s what it SHOULD do, rather than keep barreling ahead. But this is the fast lane and cars are coming up behind me that have no idea why I’m slowing down. I have to quickly accelerate. It’s a convenience, but I have no illusion that it’s all-knowing.
dnfree
@J R in WV:
I have a 2018 CX-5. If your car is like mine, you can turn off the adaptive cruise completely, or you can set the interval you want maintained between you and the vehicle ahead of you. Check the manual or ask your dealer.
Steve in the ATL
@Citizen Alan: I concur.
@Eolirin: I dissent. It’s a pale imitation!
azlib
As a retired IT professional with a Masters Degree in Computer Science, I have always been sceptical of all the self-driving claims. Most of the algorithms are based on neural networks and nobody really knows how they make decisions. A lot of the results depend on what datsets they “learned” from and as researchers have discovered, biases in a dataset can have a profound effect on the results.
While automation can handle common cases very well, it is those edge conditions (like pigeons in the roadway) that have proven to be very hard to master. I can see self-driving cars or trucks doing very well on Interstate highways, but handling day to day city driving seems to be a pretty daunting challenge.
JaneE
My last several cars have come with various “driver assist” features. About the first thing I do is try all of them. None of the features on any of the cars will do exactly what I would do if the car had nothing at all.
The feature that gives me the least problems is the cruise control, but if you are following a car and the cruise control is maintaining the distance, a sharp curve will move the car out of position enough to make my car think it doesn’t have anyone in front of it, and it will speed up until it gets a good position on the car ahead again and then it slows down again. Depending on the road, that ranges from annoying to frightening. Easy to fix, if it gets too close for my comfort a tap on the break will disengage and I can resume when the road straightens out.
Lane departure 1) wants me closer to the center than I normally drive. After a few years I learned to drive more in the center of the lane. Left to my own, I drive to the left on city streets where pedestrians or animals might be and to the right on two lane roads. It works well on well maintained streets with clearly painted lanes and shoulders. On the roads with no lane markings, or lots and lots of reflective tar fill-ins it can’t always tell where the lane is. Reflective tar is often taken for a lane marking. And it seems to wait until I have left the lane to squawk. Too late, IMO.
Lane Centering works great, if the lanes are well marked. Otherwise no. And it wants me to hold the wheel, and complains if I am too loose or if I am fighting it too much. Curves can be problematic. If I turn myself as I normally would, it complains, assumes I want full control, and turns itself off. If I don’t it depends on the sharpness of the turn whether it maintains the centering or just gives up and tells me to take over. I do use it on straight roads, or if there is a lot of wind buffeting to fight, because it does that pretty well.
The automatic rear breaking has never engaged. Once I thought it should have, but I don’t know how close it needs to be.
And any time it thinks I am inattentive or drowsy or have my eyes off the road, it keeps beeping at me. There are certain angles of sunlight that make it think I am not looking at the road constantly. And if my glasses transition to darker even a little bit, it just says it can’t tell any more and turns off. And of course if the sun glare is such that I am being blinded, so is it, and it tells me. So much for the help when I really want some.
If I read the manual that came with my cars (two brands, four cars) every one of them has pages of bold face print telling you about the limitations of the safety features and reemphasizing that the driver should maintain control of the car at all times under all conditions. I have not seen a Tesla manual, but it would not surprise me if it did the same. CYA, but the other brands don’t have Musk telling people off the cuff that the system will drive itself.
Matt McIrvin
@lowtechcyclist:
Those already exist–driverless individual “pod” systems that go on a dedicated roadway have been installed at a couple of airports. The reason it hasn’t been more widespread is just that it’s a dumb application: it doesn’t have the volume needed to be practical, compared to some other kind of people mover on a track (which can be automated more easily), or a big articulated bus, where there’s less reason to not have a driver. Usually it’s something like an expensive showpiece installed at an international terminal for well-heeled customers.
Matt McIrvin
@JaneE: The steering assistance in my car complains at you if it thinks your hands aren’t on the wheel, but the way it detects that seems to be that it wants to feel you struggling with the automatic steering a little, giving it nontrivial input. Which makes me wonder what the point is.
Also, it’s possible for its collision-avoidance radar to give up and shut itself off in very heavy rain.