“Not safe for the public’s operation.”
By Wolf Richter for WOLF STREET.
Back on August 10, amid a noisy public backlash in San Francisco from City officials on down that made its way around the global internet, the California Public Utilities Commission voted to allow GM’s Cruise and Alphabet’s Waymo to charge riders for driverless robotaxi service, day and night, anywhere in San Francisco, with no cap on fleet sizes, and with no safety driver in the vehicle.
While these driverless vehicles perform amazingly well – for example, they don’t run amok through the streets – they have been involved in some high-profile accidents, mishaps, and goofball stuff that only AI can do, with Cruise having been on the forefront with these incidents.
Then a week of accidents and incidents later – which again duly made their way around the global internet as anything with San Francisco in the headline does – the California Department of Motor Vehicles ordered Cruise to cut its fleet of robotaxis operating in San Francisco in half. That was the first strike.
Today came the second strike: The DMV announced that it has suspended, “effective immediately,” the permits of Cruise to operate autonomous vehicles in San Francisco.
Cruise can still test the vehicles in the City, but they now must again have a safety driver behind the steering wheel — sort of like being re-issued a learner’s permit after having failed the driving test.
Referring to California Code of Regulations (CCR), Title 13, paragraphs 228 and 227, the DMV cited in the statement these bases for the suspension:
- “Based upon the performance of the vehicles, the Department determines the manufacturer’s vehicles are not safe for the public’s operation.”
- “The manufacturer has misrepresented any information related to safety of the autonomous technology of its vehicles.”
The DMV said: “Public safety remains the California DMV’s top priority, and the department’s autonomous vehicle regulations provide a framework to facilitate the safe testing and deployment of this technology on California public roads. When there is an unreasonable risk to public safety, the DMV can immediately suspend or revoke permits. There is no set time for a suspension.”
The DMV also said that it “has provided Cruise with the steps needed to apply to reinstate its suspended permits, which the DMV will not approve until the company has fulfilled the requirements to the department’s satisfaction.”
Waymo’s robotaxis doing OK. The DMV’s suspension of Cruise’s permits is not a judgment on all robotaxis – just on the ones that don’t know what they’re doing? Alphabet’s Waymo can still operate its fleet of autonomous robotaxis in the City without safety driver, and fully commercializing its operation.
Our own vehicle not doing OK. Meanwhile, our vehicle got rear-ended hard in September, hard enough to get shoved into the vehicle in front, and the driver-side airbags deployed. The insurance company determined that the vehicle was totaled, but unfortunately, no robotaxis were involved, and we couldn’t blame Cruise, just human drivers doing what human drivers do: Cause accidents, lots of accidents, and 40,000 Americans get killed in a year by human drivers.
But when it’s a Cruise that runs over a firehose or gets stuck in concrete, it becomes global news? OK, maybe the DMV can teach those cowboys over at Cruise a lesson so that they can get their act together.
Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the beer and iced-tea mug to find out how:
Would you like to be notified via email when WOLF STREET publishes a new article? Sign up here.
It appears that the tech companies are going to embrace this next phase better than the car companies.
This appears to be a repeat of e-commerce: mail order companies should have adopted e-commerce, because that would be the logical order of things.
It wasn’t– instead new tech companies formed because the existing firms were run by morons.
Ditto with Blockbuster vs. Netflix.
Countdown for GM, Ford and Stellantis bankruptcies.
The UAW will be the final nail in the coffin, but the coffin was the inability to adapt to technology and changing customer trends.
Wolf– what do you think?
There could be something to it. And maybe that’s how it should be — that the automakers make the vehicles, and Waymo et al. provide the software and infrastructure that drives them.
Until an Apple equivalent comes into the picture and vertical-ize the whole thing. Oops, did I hear “Tesla” ?
Sure if Apple intends to lie and gaslight their customers consistently while delivering subpar products then sure that comparison sounds about right…
I strongly second Phoenix’s comment on crapple. And add how overpriced their products are. I’ll always give crapple credit though, for herding the sheep successfully.
Yes, companies are trying to do that. For example, Nvidia has a major self driving car project, running on Nvidia hardware, of course, and partnerships with various traditional auto makers.
Its the more logical approach especially while the tech is being developed. AI is hard, self driving is hard, making cars is hard. It makes no sense for auto manufacturers to learn AI and self driving and software companies to learn to make cars when they could simply work together. Tesla is trying to do it all and they are not close to waymo in terms of capabilities, and I expect them to slip further behind. Dropping lidar was a mistake. I understand Elons concept that “the human brain can determine depth with just cameras so computers can to!” but its going to slow things down.
Wolf,
I am wondering if there is a deeper issue at play, namely how the companies select leadership. Most of the value type companies pick MBAs or internal candidates in a type of “dynasty” system. Most tech companies are still run by founders and many never finished college and almost none have advanced degrees.
Perhaps the university system is failing to teach administrators how to add value.
If so this is a major problem and something to be evaluated when picking companies to invest in.
It is difficult for legacy companies to pivot to new capabilities. GM, Ford, and Stellantis are really good at what they do: design regulatory-compliant cars consumers (especially US) want, manufacture them efficiently with quality at scale, and sell them through their byzantine dealer network carefully managing inventory through complex revenue management (incentives) to maximize profit. Those capabilities took decades to figure out how to do.
They’re not easy – especially that middle part. Tesla is getting there, but their immaturity shows in their product reliability and ability to scale production across even a simple product portfolio. And the Big Three are doing well as late entrants to the EV market by using their manufacturing and distribution prowess to get caught back up to Tesla (and, I think, in the long run, they will be successful).
None of those capabilities have anything to do with making self-driving capabilities. Creating entirely new machine learning algorithms, storing, managing, and processing mass amounts of data, and creating support software and teams (e.g. remote hands)… those are things that Google and other companies are good at.
Of course Cruise is their attempt to “fix” that by having it as a mostly stand-alone and independent entity to build those capabilities. But that takes time, and Waymo has been around the longest (I mean, the original foundations started in the early 2000s).
With that said, Alphabet will never manufacture cars. So… really a JV model may work better. Keep in mind self-driving cars *add* complexity… complexity those manufacturers are really good at managing and scaling.
Name a “quality” car made by GM, Ford or Stellantis.
I was referring to the capability of manufacturing at scale with quality control. All global auto manufacturers are pretty good at that. The US is actually better at production quality control than most global competitors; it is perhaps worse at engineering quality control (or, alternatively, they set a lower design life limit – say 200K miles – than a Toyota at 300K).
Regardless, my point is that those are the big three’s core competencies. Tesla was founded 20 years ago and still struggles with production quality control despite low product complexity (they have four models with only two options per model with no customization other than color).
All the major globals produce dozens of distinct vehicle models with more local variation and customization options. One can argue that is not the right strategy, but they manage that complexity well.
Managing design-manufacturing collaboration and complex supply chain/production line management is a core competency. Coding self-driving cars is not. Hence why they are at a competitive disadvantage.
The 2024 Corvette E-Ray coupe looks like it has decent performance. Standing quarter mile in 10.6 seconds and @ 130 mph. Brembo carbon-ceramic brakes will slow her down nicely — which is just as important.
Base list price of $103k.
Years ago, my friend was the race engineer for Team BMW & Rahal Letterman Lanigan. When he was with BMW, I’d go to the IMSA races at Road America, Wisconsin. Man, when the Corvette came down the straightaway, it had the meanest and lowest rumbling sound. With your eyes closed, you knew when she was flying by.
Tbh Tesla quality does not impress.
I understand the tech and love the idea Tesla is, but I was looking at one the other day. Guy had these little toy looking hubcaps and then some bright red nicely painted brake thingys.
Like you couldn’t hardly see the cool red brake thingys for the toy looking hubcap.
Anyway, to me they just do not look like they will “go the distance” against the elements, time and human abuse.
And that is exactly what most successful cars of the past 40 years have done. Resisted aging. Given their owners high value and dependability.
Teslas cost a whole lot for what you get.
Time will tell!
67-69 Mustangs. More like art than a car. Can you afford one?
I had one (68, 3-speed stick, 289, no AC) when I was in college, and it was a POS even back then. It was only 8 years old when I bought it, and already the armrest had fallen off, the clutch linkage kept falling out, the speedometer had stopped working, the windows leaked, the carburetor leaked, the transmission was atrocious, handling was terrible, the paint was gone and the metal rusted on top of the driver-side door where you kept your sweaty arm when the window was rolled down (Texas!), it was useless if there was just a quarter inch of snow…. worst car ever. I still loved it.
Tesla is new Prius (with GM quality).
Halibut, my 2011 300C ticking along very nicely, thanks.
Ford Super Duty, RAM Trucks, and GM HD are all best in class, if you work for a living. As far as passenger vehicles, sure a Japanese car is a better purchase. My 2003 Honda Accord has 283k miles and won’t quit. But my Super Duty is a beast and I’ve seen people deeply regret buying a Japanese truck.
Mmmm – 1988 Lincoln Mark VII – Mustang engine, fast, 20 MPG!, looks that drew random compliments from strangers young and old, 5 years old when I got it, drove it for 15 years and never had a problem except the air suspension bags which had to be replaced occasionally.
Oh, and like Wolf’s Mustang it was really bad in the snow despite the weight and big tires, something I never would have suspected before owning it. Love(d) that car, but didn’t do my rustproofing (stupid) or I’d still be driving it.
I have been around (repair, drive, own) vehicles for 60+ years and I now find today’s vehicles to be loaded with crap, designed to break down and keep the money flowing. Planned obsolescence has returned and going full force. It’s though the society planners have deem autos the be the loving companion of people forever. Autos should be no more important that a dust pan. All the money and labor now producing, repairing, selling autos should be put to much better use.
That’s the way PC computers went.
It used to be a quality item with 10 parts inside that were made high quality themselves. All run by windows, which most of the time was designed well by smart people making high salaries.
Nowadays the only high quality PC are the gaming ones. The rest are trash that people literally throw out when it messes up.
Nevermind that you can fix them consumer…
todays vehicles break a lot less down than they used to. Today 150k miles is to a car what 60k miles was 40 years ago.
I’ve been building my own PCs since the early 90s. The quality and reliability has just gone up while the price has come down. I had to abandon windows for a few years because they crashed so much. Comparing Windows XP to Windows 7 is night and day.
>Nowadays the only high quality PC are the gaming ones.
sufferinsucatash, I would honestly disagree with this statement. Most “corporate-grade” laptops and desktop computers nowadays are built like tanks and with the way large companies dispose of stuff, you can pick a 3-year old off-lease Dell Latitude or Lenovo T-series laptop or a similar desktop (e.g. Dell Optiplex) for less than $200-300. Then you add another 16Gb of RAM and a new NVMe drive for another $100, do a fresh OS install straight from the manufacturer supplied image off their website and get a machine that would probably still work even when its hardware becomes obsolete.
I’m typing this on an Optiplex 5070 that I use as my main home station hooked to an amazing Dell u2415 monitor, both got used for ~$100 each online. I used to build PCs from components from late 90s and at times it was a very wild experience quality-wise. I’d say, if someone is not into heavy gaming on a PC, corporate refurbs are a way to go now.
Actually FOR ”in”
Ditto here , on my 4th,,, after my techie wonder guy turned me on the the corp refurbs.
Nothing at all wrong with the first two except going out of date of the SW after some years.
Current two are dell and lenovo and currently running very well with all the bells and whistles I could possibly want in every regard…
Long long time ago to reference hardware that could not keep up with upgrades to SW,,,
but would not be surprised to see ”planned obsolescence” become more again,,, rather than less as it has been last couple decades.
Some friends have basic laptops now approaching 20+ years old that they claim are just as good as new with their personal skills/ability to upgrade, with manual intervention, to current SWs.
There’s a website called Discount Electronics that sells used and refurbished office computers and accessories, and you can get amazing deals on laptop and desktop computers, particularly the Dell computers.
When there is an accident involving a human – one can sue the living **** out of em.
It would be nice to see the same for driverless OEMs and their sw suppliers.
This is possibly the single biggest hurdle for FSD that has yet to be figured out. We LOVE suing each other in this country. So when the potential costs get into the tens of thousands or even 6 figures in damages and medical bills, we need to know who to sue. Is it the manufacturer? The operator? The DMV for allowing it?
Even if self driving vehicles are getting closer to large-scale prime time, they clearly still mess up. If they never messed up, CA DMV wouldn’t be pulling unmanned Cruise vehicles off the road for safety concerns. So when they do mess up, who’s on the hook for damages? It’s a huge grey area right now because it’s all so new. Case histories will take years to build. And even once it’s perfected in the coming years, machines will still fail sometimes. If robots didn’t poo themselves regularly, I wouldn’t have a job! So there will still need to be some entity to carry responsibility when the machine fails. Could be because an on-board computer or sensors fail. Could be a programming fault. A design flaw. Could be component failure due to poor maintenance (ownership). Not everbody is going to be able to deflect blame.
We live in interesting times.
“we need to know who to sue”
You’re making up problems that don’t exist. We know who to sue. Nothing changed.
The OWNER of the vehicle gets sued. That has been the case for a very long time. The only exception is if the vehicle was stolen or operated without permission of the owner. By definition, a robotaxi operates with the permission of the owner unless the owner can prove that it had been taken over by a hacker.
In addition, other entities can be added to the lawsuit, such as the manufacturer if there are allegations of a defect. This is why lawsuits often start out with a long list of defendants. Defendants can then request to be removed from the suit if they can prove that they have nothing to do with it.
This is all very well established.
Class action suits against EV OEMs and sw suppliers are certainly in the cards. There is no human agency in a driver-less car. This is a nice funding opportunity waiting to happen! No Pfizer-like EUA protections for the barons of auto!
And I do remember we were promised self-driving taxis in 2016. Bay Area hype never gets old.
Hey, we have them now, 7 years behind schedule, not too bad, LOL
Sounds right to me.
GM has already gone bankrupt twice, and their management seems to be as clueless as they were in 1970.
“They have forgotten nothing, and they have learned nothing.”
I am glad they are beta-testing their driverless vehicles in San Francisco, and not where I live. I was surprised DMV approved them so quickly. Must be some big bucks flowing.
I have one of those locked in the garage. It keeps trying to get out, but I lasso it back in.
My experience with Cruise test vehicles near the Texas Capitol in Austin:
-In the past they were hesitant and unpleasant to drive behind because of semi-random braking for no reason. (This seems to have been fixed.)
-I saw a Cruise stop at a stop sign on West 11th more than one car length behind the line. For no reason. Annoying to following cars.
-I saw a Cruise stop at a red light in front of the Capitol, but in the crosswalk and straddling the two traffic lanes. It appears to have been confused by sunlight filtered through the trees making strange shapes on the asphalt that the computer couldn’t interpret.
-A few minutes later, the same Cruise passed a school bus that was pulled over with its hazards (but not safety lights) blinking by changing lanes without signaling. The Cruise did not slow down but zipped on by the school bus.
-The Cruise does not appear to make right turns on red–not unsafe but annoying to other drivers.
Basically, a garden-variety bad driver which I try to steer clear of.
“…a garden-variety bad driver”
To me, bad drivers do the following:
Those are not the problems Cruise has, but those are the problems human drivers have that kill 40,000 people in the US a year.
Reading your parade of horribles (callback to previous discussion) I am reminded of my 18 year old niece, who I would classify as an “unusually dangerous” driver.
But not wanting to argue semantics, I will let you define “garden variety” and instead designate Cruise as a “soft core” bad driver.
I encounter a handful of soft core bad human drivers every day, and I steer clear of them to avoid mayhem. Every Cruise self-driving car I encounter falls into this category.
So are you saying driverless vehicles wont kill any
other drivers, pedestrians, or cyclists?
And if they do the exact figures wont be made
public.
No, what I’m saying is that human drivers are absolutely the worst drivers and need to be replaced. Humans are incapable of safely operating a vehicle — that’s what 40,000 traffic deaths a year and millions of severe injuries a year are proofing beyond a reasonable doubt.
My parents were killed in 1976 by a series of willfully idiotic decisions made by pilots of a Boeing in what is the largest airplane disaster in Turkey. Only humans can be that stupid. One of the idiotic decisions was that they turned off the autopilot and flew the plane themselves, at night, right into a mountain. You can read all about their multiple idiotic decisions, one after the other, right here in Wikipedia. Humans make stupid decisions as a matter of routine, and should be banned from operating a motorized vehicle on public streets and highways (they can drive it in their back yards, as far as I’m concerned):
https://en.wikipedia.org/wiki/Turkish_Airlines_Flight_452
…again, heartfelt condolences to you, Wolf, and a plea to all to work diligently at reducing our own pilot errors as we attempt to navigate an ever-challenging world (…one may have ‘the right of way’, but don’t make it ‘the dead-right of way’…).
may we all find a better day.
This year I will log in about 5,000 miles on the road. I alternate between three different bicycles. The two newer bikes have disc brakes.
On Friday, 8 September, riding my Bianchi with hydraulic disc brakes, the ability to brake and control into a sideways slide as a mini-van pulled a U-turn into me causing a collision saved me from harm.
The dumb-ass driver never looked or signaled. He slowed to the right side of the road and I moved to pass on the left of his vehicle at about 25 mph (50×14 @ 90 rpm cadence). Just as I was on his left-rear quarter, he accelerated and spun left. The impact happened in just over a second after his turn, and was spread through my thigh, hip, forearm and knuckles.
But because I had disc brakes, I was in control of how I slowed, and how I rotated the rear wheel by locking it up to slide. I maintained an upright position on the bike. The collision left me with only a slight abrasion to my right hand (that was on the brake lever). Oh yeah, I left a nice dent in the driver’s side door and took out the rear view mirror. I body-checked a Chrysler Town & Country as my final stopping force!
Young driver, right in front of Roosevelt High School in south Minneapolis. He apologized, and no, I did not punch his lights out, but his window was down, and I thought about it . . .
Had I been riding my 13 year-old bike with Campy Super Record rim brakes, things would have turned out much different — and worse.
Bottom line: always wear a helmet; always be on alert and assume the worst from others as you ride.
“The Motorbikes’ & Bikes’ rules for riding.”
vecchio gatto veloce,
So glad to hear you’re alright. Glad you have a trained body of steel that can dent some sheet metal. Glad you had the equipment and decades of experience that gave you a chance. I gotta add “U-Turn in front of a bicyclist” to my humans-are-terrible-drivers list.
DanRo – well-practiced and played! Glad to hear the repercussions were (relatively) light.
To reprise an incident of my own:
About a decade ago, I was motoring the NX250 back home on the narrow, paved-over stage route that serves as the main road for our area. Coming downhill on a blind left-hander I was confronted in my ‘lane’ by another rider on what I dimly recall as an R1100s. To the right was a nearly-sheer 40′ drop and not a good ‘out’. Flicked it left, but not hard enough as the other rider flicked his hard right upon seeing me. My front wheel hit his right cylinder, bending the rim and forks and launching me over the bars to land on my head (thank you, Shoei!). My bell was rung pretty good, and I struggled with staying conscious, getting my bike out of the road, and dealing with the remonstrations of the other rider about ‘neither of us suffering that much damage’ before he quickly remounted and rode off. Fortunately, I was close enough to somehow wobble home, concussion, damage and all, though I was ‘foggy’ for the next two weeks…
It would be easy to assign complete blame here, the Beemer rider being on the left side of the road on an uphill blind right, but that would also make it too-easy to skip over my own pilot error, that being the one of never having visualized a moto in my lane in that situation. I’m confident that were it an auto, with its slower control inputs, my evasive maneuver would have been sufficient. Being a moto, however, made it more akin to a wild-west showdown on the main street, his panicked ‘draw’ carrying him well back into his lane and me, where my reaction probably allowed for a car, but NOT a bike…
Please, always, always work on your piloting chops, and never assume you’ve seen/know it all (…and, as my good friend NBay says: “…learning is fun!…”, or, at least, might help one live to fight another day…).
may we all find a better one.
91B20,
Wow. Sometimes, things happen quickly, and reaction becomes instinctive & reflexive.
As Terence Hill’s character, ‘Nobody’ repeats to Henry Fonda’s character, ‘Jack Beauregard,’ “The secret to a long life is you try not to shorten it.”
Glad that we both understand this, eh? Take care . .
Corruption rules. C’mon man, were dealing with humans.
I imagine Tesla’s entry into this enterprise is going to be a cakewalk for them since they produce the vehicles and all the software
Tesla’s software maturity (and use of sensor data that is actually good enough to make Level 4 and eventually Level 5 driving possible) is lacking. If they are serious about entering into this market, they are not demonstrating the investment to make it happen. They’ve made significant cuts to their autonomous driving investments even as the rest of their business grows.
The money Waymo (and Cruise) are pumping into autonomous driving is at least an order of magnitude more than Tesla.
“If they are serious about entering into this market, they are not demonstrating the investment to make it happen.”
Oh they are serious alright…serious in taking your money and promise you the moon and let the general public be their crash test dummies, something the general public didn’t sign up for…cherry on top of that, gaslight you to believe their Autopilot is already level 5…
but if that seriousness is about perfecting the technology to be flawless (or close to it) do what’s it is advertised to do an not overselling..then no, the head freak is not a serious person by any means…Dickapedia anyone?
It would be interesting to learn how tech savvy the politicos who approved initial fielding truly are/were.
Except Tesla doesn’t have an autonomous system. It has a driving-assist system that’s constantly trying to kill people. Tesla still has ways to go to catch up even with Cruise.
But if Tesla ever gets this worked out, it would be hard to beat.
Musk watched a little too much westworld on HBO during the coding phase. ;)
/s
I think we are seeing in real time why QQQ has been so resilient in the face of interest rate hikes, inflation, etc.
It also explains the underperformance of value stocks over the last few decades.
The CEOs of value stock type companies simply have not been good at incorporating technology and it is other companies which have reaped the rewards.
Blockbuster vs. Netflix is the most glaring example, but newspapers missed the boat on offering a Craigslist type service, even though they already had the classified listings nailed down. All the newspapers had to do was set up a website with the classifieds and have a communication portal! But they couldn’t figure it out.
Same thing with Sears– which had run mail order for over a hundred years– they couldn’t pivot to e-commerce even though they had the infrastructure in place to do so!
Traditionally value stocks have outperformed growth, but something appears to have fundamentally shifted.
Maybe the MBA education system is completely deficient, or there is some other fundamental flaw, but even simple adjustments that existing companies could do to adapt to the environment aren’t happening.
Does anyone have insight into this?
Understanding this dynamic is likely to be critical to generating alpha moving forward.
Why do established companies miss technology/business inflection points?
It is akin to an immune reaction. Andy Grove called it the “Creosote Bush” effect. The creosote bush prevents anything from growing around it. So does a high-margin established business.
It does sound earlier as it is. Think about a traditional car manufacturer. He buys components from a supplier. Such components come with their own small computer. The same applies to their EVs.
On the other site e.g. Tesla comes with a single computer as it is designed from the ground. Tesla can easily update the software while traditional manufacturers need to update the software for each component and need to test that all pieces are fitting together.
The same applies to many traditional companies. They are spending large amounts of resources to link things to new technologies. Very often at the same cost as doing it from scratch. However, they benefit from their recognition in the marketplace. Just consider how many millions e.g. Walmart has spent to buy and make its online store. Not much less than other online stores have spend
Walmart is a good example.
When you consider how dreadful the whole Walmart-in-store experience is, they do an amazingly good job online.
(I suspect that the online folks have NOTHING to do with the bricks-and-mortar folks, except the name.)
So…what are the success ratios for tech/high tech transition to full scale industrialization?
I have worked tech/high tech transition from concept to to innovation to at scale production. It’s not the easy, certain path many think it to be. Especially with software and networks.
It aint cheap either.
Tech innovation is easy but changing existing patterns of human behavior takes an enormous amount of work. Especially when you are changing monetary reward models. Constant and focused effort which is usually more than you are paid to do.
Well, around here the newspapers figured out how make their classifieds an internet service. But it was only space for one such service so those newspapers not in the group taking the market lost out. Not that much better for the newspapers in the system, the internet classified have grovn to a larger, dominating business…
The QQQ big tech stock bubble is inflated by printed money, interest rate suppression, and naïve or careless buyer sentiment. It’s performance is like a drugged athlete, aided and fake, primed for an inevitable downfall. Historical valuation data by Hussman shows high statistical likelihood of negative to zero returns over next decade.
I do wish one day we will get to a point where Level 5 is fully trusted and remove unsafe drivers off the road. So many douchebag or morons driving around SoCal that are clotting up the road that should not be there. Hopefully AI will solve the problem of seeing Dodge Charger/Challenger douchebag drivers on the freeway…well I mean AI can still drive those stupid Charger/Challenger but hopefully they won’t program in the douchebag personality of the drivers these vehicles are known for…
And you’re going to protect all those cars with thousands of chips, sensors and connections to back-end databases, operators and other connections from becoming vulnerabilities?
You’re probably familiar with the famous UC San Diego automated vehicle hack.
Tech advancement can be immensely value-adding. It, however, always brings its own set of new complications/externalities.
Nah, I don’t think that far ahead…I only want self driving to get rid of douchebag Charger/challenger off the road…lol
Would be nice but those said drivers would not voluntarily reliquish their aggressive driving behavior in favor of automagical driving. We have the AI technology now to enforce bad driving with cameras at intersections and along highways but we don’t seem to have the will to do it.
There are lots of douchebags in all walks of life who directly/indirectly add risk!
Make this equal opportunity societal pruning and I’m all for it!
>And you’re going to protect all those cars with thousands of chips, sensors and connections to back-end databases, operators and other connections from becoming vulnerabilities?
Honest question – how this is conceptually different from a bad actor stealing a car and, e.g. then killing someone in it and fleeing the scene?
In both cases a breach of a “security system” was performed by an unauthorized party.
In both cases the identity of this bad actor is unknown.
In both cases there is some substantial damage from these unauthorized actions.
How a cyber-car-jacking would be different from a very-real-car-jacking?
You missed the point: risk.
Classify, cost and add probability as you wish!
My thrust is simple – no free rides.
I don’t care which prevails in the end: ICE or EV; whichever is cheaper for me is fine. I see a limited use case, in today’s context, for EVs -> low monthly mileage drivers. Otherwise, the huge subsidies tell me all I need to know (independent of recent assessments of problematic EV issues).
I’ll confess another personal criterion: personally, I am not one who favors loss control.
Various *experts* are now saying no fully automated cars until 2035. We’ll see. The challenges are immense. Like most newer cars, my mini van can steer itself on well marked roads, with clear lines. But on a typical day I will be on side streets and dirt/gravel roads. It gives up on those. I doubt computers will ever be able to do an adequate job where things get confusing. They’ll never recognize a cop directing traffic, or road crews/detours with all the cones and diversions that are not standardized.
Long before this tech was hyped, there was an article saying future roads would need to have sensors embedded in the concrete that cars can follow. There would need to be stardardized sensors for that cop in the intersection, or emegency vehicles that need to wind around traffic. Then, of course, it’s all based on wireless phone data that is subject to delays. That has to be solved as well.
“Various *experts* are now saying no fully automated cars until 2035.”
So tell your experts this: You can hail a ride TODAY with a Waymo in San Francisco, and the vehicle will only have you in it, and it will take you to your destination through the chaotic congested traffic, day or night, with steep hills and hordes of pedestrians. It’s actually quite amazing. These things work a lot better than I thought a few years ago they would ever work.
It’s still geofenced. Tell us precisely where are it’s limits. I doubt you know that.
To be impressed by this is to be impressed by those Bill Gates presentations that were scripted for him… and even then the software often failed.
As per the permit, they have to stay within the city limits of San Francisco. That’s the only geofence. So they’re taking paying customers; when you order the vehicle via the app, you enter your pickup point (must be within city limits), and you enter your destination (must be within city limits), and the vehicle shows up without driver at the appointed time and takes you to your destination. There are lots of these rides now taking place. People aren’t even turning their heads anymore when an empty vehicle or a vehicle with passengers in the back but no driver rolls by. SF isn’t the only city where robotaxis now operate.
Dragging the script of a Bill Gates presentation into here as proof that these vehicles don’t work? Good lordie.
This has been weird to me over the past couple of years. There is such a huge disconnect between reality (open-access driverless hailing in parts of Phoenix for YEARS with recent expansion into far more challenging environments like SF and parts of SoCal) and what people say (“driverless cars are decades away!”).
There are some unaddressed challenges (certain weather conditions for example) but the issue of driverless vehicles seems to be more an economic one than a technology one at this point. It appears to still be cheaper to pay Uber drivers less than $10/hour after expenses than pay for the costs associated with driverless taxis (the sensor packs are stupid expensive, the on-board compute isn’t cheap, maintaining the detailed mapping isn’t cheap, the remote hands support network isn’t cheap, ….).
I am very interested in the next 1-2 years. Given that it is primarily an economics problem at this point, scale is key to making it cost competitive with rideshare. So Waymo (or Cruise, or whoever) needs to decide to “go big” (or, decide to go home)… otherwise they will keep burning money.
So, who knows. Maybe we’ll wake up in 6 months with an announcement that Waymo is partnering with Tata to mass produce their autonomous taxi Jaguars… or with GM to mass-produce autonomous Equinox taxis or something.
The experience of the tech industry over the last 80 years or so is that:
-Stuff gets cheaper over time.
-Once you have an algorithm which works, making more stuff which uses that algorithm is (relatively) easy.
-Adding features, including safety features and reliability features also gets cheaper over time.
The expectation is therefore that this stuff will be cheaper than human taxi drivers in the near future. Right now, the SF robo-taxis are more a proof of concept than prototypes, but if you wait 20 years, I suspect that they will be quite practical.
@Dave – 20 year time horizons for Waymo and Cruise are irrelevant. They need to figure out how to commercialize the tech at scale within the next 3-5 years. They’re burning billions in cash. Maybe the parent companies could deal with a longer horizon (7+ years) if money was still free but it’s not anymore.
Small scale incremental doesn’t work because at the scale they operate, each incremental vehicle is a loss. So, either they go big in the next few years, or it is 10+ years off.
Unfortunately some of the “tech always gets cheaper” stuff doesn’t fully apply here. Yes, the onboard compute gets a bit cheaper. The cost of simulation and so on gets a bit cheaper as datacenter compute goes down. But without scale, the required lidar, camera, and radar systems don’t inherently get cheaper. Now, some of that is being addressed… partially.. by Level 3 cars starting to get Lidar. Maybe that bends down the cost curve. But you get my point.
The cost of a Waymo car is estimated to be $200K+ today – with $75K+ of that being the sensor pack, a much smaller portion being the compute pack (probably $10-20K), and the residual cost above a regular car ($30K+) being the assembly/integration costs.
When you add in the overhead of maintaining the service (mapping data, algorithm upkeep, remote hands, garaging/storage, car maintenance, …) – if you want to compete with $10/hour human drivers in $30K cars you need the total car package cost to be <$100K. The only way to get there is via scale. Scale will drive down the sensor pack cost and integration/assembly cost (direct compute won't go down materially from scale since it is almost all commodity as is).
If I get rear-ended by an autonomous vehicle, whose insurance pays for the damage?
The owner of the vehicle at fault (or their insurance) pays. Nothing changes that.
Only in at-fault states.
That’s not what no fault means, in any state. It’s a very common misconception. No fault just means that your own insurance pays for medical expenses, and no attempt is made to apportion costs between the insurance companies to the at risk party. It has nothing to do with damage to the cars at all.
My 1936 Cord does a nice job of getting me around, except for the horde of onlookers. My old 1967 Mustang is less notable. My piece of shit, 2002 525I wagon suffices for most hauling jobs, but if needed, I use my 1999 GMC, even with the failing transmission.
You lace pants boys are hilarious.
DM: Elon Musk saved Tesla from ruin by promising self-driving cars. The problem? 700 crashes, 19 deaths and claims of a FAKED promo video. So TOM LEONARD investigates: Is his empire speeding towards disaster?
A new investigation has shockingly claimed that, since it was introduced in 2014, Tesla’s self-driving software ‘Autopilot’ has been involved in more than 700 crashes, at least 19 of them fatal. As a result, Tesla currently faces 10 known lawsuits, each alleging crucial flaws in Autopilot which they say the company failed to deal with while grossly exaggerating its capabilities to the public. In similar cases that have already come to court, Tesla lawyers have argued that the driver is ultimately in control of the vehicle and must always be paying attention. Last month, Tesla attorney Michael Carey told an on-going California trial over the death of 37-year-old Tesla-driver Micah Lee that Autopilot is ‘basically just fancy cruise control’. Nonetheless, this blizzard of legal action now not only risks potentially denting Tesla’s reputation and bank balance – but, after his recent troubles with Twitter, adds to the mounting questions Musk faces about his stewardship of major global brands…
Tesla doesn’t have self-driving cars. It has driving-assist systems.
If it is made by man and AI, is there will always be a difficulty. Fact!!!!!!
Was this written by AI?
If yes, AI has a long way to go
If no, humans have a long way to 🤣
I think the main purpose of self driving cars will be to extend the days of happy motoring in the the face of the rapidly increasing cost of car ownership, as the last of affordable used vehicles wears out the only way to keep the average joe off a life on the bus will be a way to run cars 24 hours a day without the labor cost of a driver. Then even if cars cost $200,000 and insurance is $20,000 a year an average joe can still get a ride to the ballpark or swap meet.
Couldn’t have anything to do with people using self driving taxis as a motel for 20 minutes.
lol!
Wolf
Did you ever find out why the car rear ended your wife? Was the driver on a cell phone? That’s happening a lot around here. Ins COs routinely get cell phone records of the offending vehicle.
The insurance companies are struggling with this right now. I don’t know the details.
My neighbor recently purchase a Tesla. He was telling me all about it’s great features, especially bragging about it’s self drive ability. I said, “That sounds pretty cool, can I see it? He replied, “Well I would be glad to show you but it took off three days ago and I haven’t seen it since”.
Eloped
🤣
Driverless cars MAY be the way to divert the public’s attention, time and money to better things in life than CARS. I think gov should become the provider of transportation just like they do for roads, harbors, water, garbage collection, waste removal and many other things. The current system of transportation is nothing more than a whirlwind of cash, debt, and wasted time.
I think the gov should provide for everything.
Communism is wonderful.
Everybody can have everything they want, the government provide it for free!!!! I love it.
/s -just in case.
I remember years ago late 70s / early 80s where it was not uncommon to see in the local newspaper used car classified private party ads saying something like “for sale, 1975 Buick Electra. 100,000+ miles, some rust, everything works, good drunk car, $900 or best offer”. So the car knew its own way home.
Had a Tennessee Walker that could do the 60 years ago!
I am now imagining all the lives that will be saved from drunk drivers not being a thing when AI vehicles are really everywhere. This will be a good thing, and it will increase accessibility to transportation. Of course Google gets the AI model right.
I wonder what it will do to the hedonistic inflation adjustments.
It will do well for the bar business like Uber does now.
What? Tom Cruise will drive you up and down the hills of beautiful San Fran?!
Wait a sec, I’m going to go reread this article because SURELY I have just skimmed it and come to an illegitimate conclusion.
Ahhh yes as I anticipated.
Twas’ not the case.
And yep we had some deaths here in Raleigh via cars. Seems to always dominate the headlines.
DOJ now part of probe into Tesla’s claim and sales of ‘full self- driving’
along with DOT, which some time ago ordered an engineering review often the prelude to a recall.
PS: the DOJ role seems to be about exaggeration of range.
From a risk management perspective, it makes sense. Less humans behind the wheel = less auto accidents. I’m not sure that an endless stream of autonomous vehicles in my neighborhood will add value so I’m glad to see that others are willing to give it a go. Looking forward to hearing how the locals feel after full implementation.
Anecdotally it seems like 90% of the viral bad self driving vids are Cruse vehicles. Does anyone know if there’s actual statistics on Cruise vs. Waymo incidents?
There was an excellent PBS NOVA entitled “Look Who’s Driving” about autonomous vehicles back in 2019. The problems are far greater than most people realize.
That was four years ago. Waymo vehicles are doing great without driver.
So are the subsidies! (which are enormous)
It’s too bad we don’t have a true democracy (one where everyone can vote on every issue and has full veto authority) and true capitalism (nothing “state” owned). Then it wouldn’t be up to some DMV goons to decide these issues.
“…too bad we don’t have a true democracy (one where everyone can vote on every issue and has full veto authority)…”
California has a fairly direct democracy with dozens of issues on state and local ballots at each election. It’s kind of fun because it’s such a huge and crazy mess. Few people even try to understand the long and complicated legislative texts they’re voting for or against, with sometimes astonishing results.
According to your proposal, there would be thousands of such propositions at monthly elections as each administrative decision would have to be decided by ballot. That would be a hoot.
Actually seems like a good or very good idea to have EVERY,,ONE,,, or at least every ”driver” fully ”required” to VOTE EVERY week on every issue AKA ”problem”…
Those who chose NOT to vote and thereby lose their driving license(s),, etc., etc., including maybe losing their ability to ”charge” any and every thing may more quickly see the light…
While this is certainly whimsical at the least,,, just think about the condition where EVERY ONE is required to participate in EVERY single ”election”???
While WE, in this case the VAST majority of humanity WE sit and endure the atrocities currently seen and proven,,, in order to continue to increase at least the first, and hopefully the second derivative of the line of increase of actual ”democracy”,,,
IMO the more involvement of EVERYONE the better, as it will almost certainly smooth out those lines and influence upward the general and long term trend to democracy rather than the opposite.
No reason citizens couldn’t band together and collectively hire expertise – kinda like the non-expert lobbyists who control federal and state governments….even in one party states like CA!
Please! Read the propositions! In Nance’s state no less!
(one where everyone can vote on every issue and has full veto authority)
Full veto authority? Since out of millions, SOMEONE won’t like it, you are advocating for society without law.
Minnesota’s DMV will financially punish, year after year, those citizens who chose to buy and drive a newer four wheel motorized vehicle.
Now, Minneapolis has a sales tax rate on 9.025%. Out-state it is 7.625%; with many local add-ons in various places. But the most egregious thing is the license plate renewal fee structure. Not only do you pay the nearly ten per cent tax at purchase, but there is a sliding fee scale based on vehicle age and initial cost basis.
You want that 2024 Corvette for $103K? You pay $9,296 extra to drive it off the showroom floor where I live. But the two little plastic stickers needed to put on the plates in one year will set you back $1,324 plus a few additional “registration and wheelage tax” fees.
This extra tax rate slides down until, at year twelve, tab fees are equal for most all vehicles.
At least with motorbikes, this does not happen . . yet, anyway.
EV tabs have a $75 per year surcharge added.
The reason we do not have such systems is because they are unrealistic. They sound great to naive goobers, but they don’t work in the real world.
I see this action as being the result of a systems problem.
With a normal human driver, if a police officer, firefighter or even just a construction worker waves to get your attention and then uses gestures to tell you what to do, you’ll likely figure it out and follow instructions. The current crop of self-driving cars do not know how to identify a human figure that has the authority to tell them what to do, or even how to follow those gestured directions. This make the self-driving car a multi-thousand-pound hazard that cannot be reasonably controlled by emergency personnel. At best, they can try to stop the self-driving car by standing in front of it. I can easily understand why people are not wild about this plan. If the car doesn’t stop, your new name is “Hood Ornament”.
Solving this is COMPLICATED. If you’re going to do it visually, the computer had better be able to tell the difference between a police officer in a black uniform and a thug in a black hoodie. If you are going to give out a control device to “authorized users”, you have to make sure that those are only in the possession of the right people and that you have the ability to disable control units that have been “lost.”
Current self-driving cars are not “smart”, or “intelligent”, but rather an idiot that can follow an incredibly complex ruleset very, very rapidly. But first, somebody has to write the rules.
Tesla has the advantage in FSD because they have millions of cars on the road collecting data to feed into their neural net training data sets. Even if those cars are not using auto pilot, they still collect the data.
Unfortunately its a very hard problem to solve as the 80/20 rule applies. 80% of the effort must go into handling 20% (or less) of the edge cases. So a premium exists in collecting lots of data on edge cases and thats very hard to get.
Does being successful in FSD mean the robot drivers are intelligent or simply clever at the somewhat narrow task of driving a car safely ?
Can there be real general intelligence without consciousness ?
I have never found an entry into Tesla but I am still watching. I am probably too conservative as an tech/growth investor. I consider myself too ignorant in those areas.
I know shorts in general have learned to walk very lightly around Elon. He is not a man to bet against.
Elon/TSLA probably has more real road data than all of the rest of the self driving crowd put together. That seems to matter in AIs teaching themselves and Elon was a founding investor in ChatGDP/OpenAI.
I doubt I will ever short Elon and I had my best year ever in 2022 almost completely short. I admire Elon. More than the efforts of any other fairly large handpicked group that could be chosen he is responsible for turning transportation green/electric and he is only ~52.
I encourage all of the defamers of Elon and TSLA here to take out shorts on him.👹
The principle of safety first should have been in place before authorizing these things. Autonomous AI docs are not allowed and would not be allowed without a real doc. Simply because, while they perform beautifully overall, they are subject to spectacular particular failure: in both cases with the same potential result: death.
That’s not acceptable. Great technology: unsafe.
What really happened was this:
This hit-and-run incident is still being investigated. According to Cruise, its autonomous vehicle (AV) detected the collision and stopped on top of the pedestrian, then veered off the road, dragging the pedestrian about 20 feet. When the AV finally stopped, it appeared to pin the pedestrian’s leg beneath a tire while videos showed the pedestrian was screaming for help.
This is in keeping with the Uber self driving car that drove over a homeless person in Phoenix a few years ago. IIRC, Tesla SFD ran over a child mannequin *twice*.
You left out the most important part:
The pedestrian was hit by an effing HUMAN driver and got thrown into the path of the EV. That’s the thing that really matters. An effing HUMAN driver hit the pedestrian hard and threw that person into the path of another vehicle. Human drivers are the WORST. ALL human drivers should be banned from driving. Human drivers kill 40,000 people a year.
A Cruise would have NEVER hit that pedestrian in the first place. DUH.
The biggest problem for Cruise is that Cruise mislead everyone about it for weeks. Cruise has a history of doing that. And now it got whacked.
I am with you Wolf, of course I am biased, I am a full time cyclist. No carbon spouting metal boxes for me.
Someday in very old age electric vehicles driven by well trained AI’s will deliver me to doctor appointments and have their robot assistant carry me in.
The true test of driving is being able to respond to situations one never trains for. On that front Cruise failed just like the 40000 drivers you mentioned. Those 40000 must have been accident free until they weren’t. After all, they got their driving license by passing a test overseen by a human in DMV.
And now they are a statistic. By that logic, SFD from Cruise, Uber, and Teslas need to be too.
As for your claim about SFD not hitting others under normal circumstances, neither do human drivers. That is why the number of accidents is so low considering the number of cars in rush hour across the country.
For SFD to be the panacea you expect, you will need to be around till the end of time because human behavior changes and the software needs to have seen a few instances to know the correct response.
Human drivers kill because they knowingly do stupid things, such as indicated in my list, which I posted above and repost now:
To me, bad drivers do the following:
Those are not the problems Cruise has, but those are the problems human drivers have that kill 40,000 people in the US a year.