US safety regulators contact Tesla over erratic robotaxis | Hacker News Reader

US safety regulators contact Tesla over erratic robotaxis

bbc.com

73 pointsZeljkoS92 comments

One thing Waymo did really well was roll out their service extremely slowly. They knew they needed to build a whole lot of trust to get people to accept them. Can Tesla do the same thing? Can they accumulate a track record of being nearly perfect for several years before trying to scale up?

Short answer: no

> relying on in-car cameras rather than the radar and sensors ... It is betting that its approach will be cheaper

Is this due to reduced processing requirements, reduced sensor costs, or what?

both? they did try and get to market faster than cars with full lidar rigs etc

The didn’t actually get to market before Waymo in the robotaxi market

This decision was made at the height of the pandemic supply chain shortages, but then was never reversed when they could get sensors again. FSD will never work with pure vision and it's folly that Tesla / Musk insists that it will.

Can you cite some practical failure scenarios besides a wile e coyote billboard where camera inherently won't be able to accomplish what lidar/radar do?

I was driving in convoy over closed (to non-convoy driving) mountain pass in Norway, in winter. Everything was white, you could barely see the road. Snow was so heavy that visibility stretched as far as the car before me, you couldn't see further than that. At some point a snow plug passed on the other side of the road and completely covered my windshield with snow, to the point that for a few second I had no visibility whatsoever. Good luck to cameras then.

In case it's helpful to know, Lidar also struggles to perform well in heavy snow due to the scattering and reflection of laser beams by snowflakes, which reduces the detectable range and can create false readings.

But you are also only a camera without LIDAR or RADAR, and you apparently navigated that situation successfully...

There are advertisements on the sides of trucks. Better question is why you are willing to dismiss the wile e coyote failure demonstration.

The wile e cayote failure was using tesla's cruise control rather than FSD. Binocular vision is sufficient.

Lidar will likely be outlawed anyway as it burns your retina. Dare you to put your eye (or cellphone camera) next to a waymo sensor for 10 sec and see what happens.

FSD catapults families into highway dividers, I'm pretty sure LIDAR will be fine

Virtually all LIDARs are class 1 devices, safe for human eyes to look at. Do you have any reason to believe Waymo LIDARs aren't class 1?

The rapid rotation and your blink reflex is apparently what makes it class 1 (supposedly). I don't see how a blink occurs if it isn't in visible spectrum. The science around the safety is very shaky, many tests look at only retina and not cornea damage.

https://www.laserfocusworld.com/blogs/article/14040682/safet...

Many of the class 1 lidars do damage camera sensors, is your eye really that much more resilient?

The damage at those wavelengths is just overheating. Your eyes are mainly water with its absurdly high heat capacity compared to pretty much anything in an image sensor. I'd expect eyes to be much more resilient in short duration, high intensity exposures and significantly better at long duration, low intensity exposures. LIDARs have been shifting towards short duration exposures to get higher intensities (for range) in recent years, simply to stay under the class 1 limits.

Heat damage doesn't make it safe, eyes are very sensitive to heat and while the cornea provides some protection, not everyone has a thick cornea due to surgery, and there is also evidence that heating the cornea causes cataracts.

The short exposure is putting a lot of faith in these device manufacturers. How do they guarantee short duration, are you confident that each manufacturer has sufficient electronics to stop the laser if the spinning slows slightly due to dirt, low power or a crash? Should the exposure limits be the same during day vs. night when your pupils are dilated?

LIDAR uses wavelengths that are transparent to the receptors in the human eye (but not to the sensors in many brand of cameras).

So, safe for human eyes but deadly for a camera.

It is funny you should say that binocular vision is sufficient when Tesla Vision hardware does not have binocular vision.

HW3 has three front-facing cameras that are not only too close together to provide binocular vision with adequate disparity, but also have different fixed focal lengths making them unable to establish binocular focus even if they were far enough apart [1]. HW4 has two front-facing cameras with the same limitations.

This is of course ignoring the fact that humans with visual acuity comparable to the HW3 cameras would almost be legally blind and not meet minimum vision requirements to operate a motor vehicle in, I believe, every state. HW4 cameras are better and you would only be unable to meet minimum vision requirements to operate a motor vehicle in most states, including California and Texas.

[1] https://en.wikipedia.org/wiki/Tesla_Autopilot_hardware#Tesla...

Cameras can obviously work at least as well as a human if they're attached to a human brain. The question is whether you can put enough compute and data together into a car to be able to do it live.

Why even bother when we can make artificial eyes that see depth? The price of LIDAR has plummeted and will continue to plummet. We already know that it works really, really well for self-driving with today's available compute+data.

> Cameras can obviously work at least as well as a human if they're attached to a human brain.

Eh, I mean I think that that’s necessary, but maybe not sufficient. Eyes are _really good_.

Fair point! Is there some advantage eyes have that wouldn't be surmountable with simply more cameras (i.e. to capture different exposure ranges etc)? I actually haven't thought about this side of it super closely, but I think you're right.

It's a question of cost and technology. Cinema cameras don't cost hundreds of thousands of dollars just because studios like spending money. Humans can see differences between even the best cameras on the market.

It's also a bit of a false analogy. Cameras don't really work like human vision. We do things like mesopic vision that simply aren't possible with current sensors. We have massively high resolution. We have async "pixels" that can respond immediately without waiting for a framing signal. Our brains process color in truly weird ways.

It's not like there's some physical law preventing computer vision from being better than human vision, but it's an incredibly difficult engineering problem that we've spent the better part of a century attacking without clearly winning.

Resolution; extreme dynamic range in a single exposure; sensitivity with short exposures in low light; focal range and speed of refocusing; white point adjustments (happening to a large degree in the retina as receptors become temporarily desensitized to a given wavelength with uninterrupted exposure). I’m sure there are more.

I can't see in the dark, can't see in the fog or the rain, can't see UV, my eyes only see in the rough direction my head is facing, there's a limited ability to track objects. Bicycles coming from behind are particularly easy to miss. Speaking of easy to miss, there's a hole in your vision m that you doing notice, where the optic nerve is. Hell, there's a whole body of work for times when human vision falls short and gets fooled, called optical illusions. There's another whole field of study about failures of the lenses and other parts in the eyes themselves. Some of those failures an electronic camera system is also going to have/have a reliance on components not being broken.

Given the number of shortcomings of human vision why shouldn't our self driving cars be designed to have better than human vision, especially if the goal is to not get into crashes. Humans, with human vision and human object tracking skills, and human reaction times get into crashes all the time. Shouldn't we want better and more sensors, which would lead to few fewer crashes, simply because better sensors have better data available?

As other commenters have noted, the kind of cameras that see better than human eyes cost more than a car.

Cheaper than that and you're making significant trade-offs.

It's not a given that a camera will even work as well as a human if it's attached to a human brain. The human eye is stereo, it has a focusable lens, an iris, it's incredibly sensitive and the foveated retina has a very high resolution. Can't say the same for the cheap-ass cameras on the Tesla. I'm not sure there is a camera on Earth that is the equal to the eye (yet).

Great points! The eye (really everything attached to a biological system) is fucking amazing.

How about this? https://www.bloomberg.com/features/2025-tesla-full-self-driv...

Sure, anything that involves fog.

Shiny truck trailers have been a failure case. A reflection of the sky in a truck back looks a lot like the sky from the right angle.

Fog

Shadows. Everyone I know who's tried FSD learns very quickly that it's a random chance that the car will see an oncoming shadow (big truck, bridge, tree, etc) as a wall and either slam on the brakes or swerve.

If you have reliable depth information, ala LIDAR, you'll be able to know that there's nothing actually there.

Andrej Karpathy should be on that list.

And there should be some criminal liability since people have died.

The craziest part is that many Teslas already have these sensor installed (physically), yet their input is coded-out/ignored!

The ultrasonic parking sensors in older Teslas only have a range of a few feet. That'd be useful for automated parking, but not for self driving above 5mph.

>only have a range of a few feet

At the very least they couldn't they detect imminent collisions with pedestrians? walls?

----

My non-Tesla has a radar for cruise control and imminant crash detection — while annoying at first, it is unfailable. As I've aged (¡perfect driving record!) my thought process has gone from more horsepower! to more safety!. Now when my car beeps at me I'm more-inclined to listen cautiously.

Not really. Stopping distance at 20mph is 20ft on dry pavement, and it goes up quadratically the faster you're going.

Ultrasonic isn’t particularly useful for fast reactions. If it’s navigating around a parking lot slowly, sure, but if you’re going any reasonable speed it’s basically useless.

>if you’re going any reasonable speed it’s basically useless

My active cruise control works even when set to 99mph (e.g. it will slow me down upon approaching another vehicle). I'm sure it'd work at higher speeds, but won't "set" above 99.

I’m willing to believe that machine vision would eventually become good enough to match or exceed human visual perception given the same inputs.

But humans in a car have a massive advantage over little cameras that no one seems to discuss much: we have two sensors (eyeballs) mounted on a servo (our head) that can move around and is looking through a truly monstrous aperture (the windshield), and that aperture is equipped with fancy cleaning devices (wipers and cleaning fluid spray), and the car’s operator is motivated to clean the windshield and maintain the windshield, wipers, and spray system to be able to see.

A Tesla car has little tiny camera lenses that are every bit as exposed as the windshield but don’t have all the compensating machinery.

Go stick a pair of nice cameras on a three-axis servo mount with a range of motion of a whole foot (or a camera array and no servo), stick that two feet behind the windshield, train it well (use that massive parallax!) and I’d believe the result would be competitive in performance but definitely not cost. Also the car would lose an entire seat.

Or use radar and lidar and achieve super-human performance.

Fir what it’s worth, the military was and is fully aware that lidar and similar tech can outperform human eyeballs in “battlefield conditions”, and I’m aware of old DARPA projects to do things like pulsed laser range-gated imaging to see through fog and such. (You still get attenuation and scattering, but you can mostly disambiguate the additive signal from fog from the stuff behind it.) Lidar can do something similar. Humans can move their head to acquire more data. Little cameras are at the mercy of the fog and can only use fancy image processing to try to compensate.

Watching this from the outside, cameras only seems to be a religious decision based on how strongly people react to the question.

5 years ago I agreed that you'd need the other sensors. ML vision has improved so quickly now I'm really not sure you do. From what I've seen the system available to consumers also performs well IRL.

I’ve been using Tesla FSD since the original beta rollout. Collision avoidance has never been a real problem. Phantom braking has gotten a lot better, though still not 100.0% fixed.

Most of the problems I have now are things like lane selection or turning into the wrong parking lot, which seem solvable in software given enough time, and the Robotaxi project should ramp up the urgency on that front

I think the correct answer here is that when you are producing vaporware to run your hype machine, you avoid hardware costs like the plague.

Lidar units suitable for this sort of thing used to be extremely expensive, but they’ve come down a good bit and will likely continue to. At this point hard to read it as anything other than obstinacy.

I always thought it was because of patent licensing. Basically, extra unit costs.

Costs. Lidar is extra sensors (and also extra signal integration but not having a lidar requires a lot of extra video processing to get information the lidar would straight up give you).

Allegedly, they believe a system with inputs similar to human vision is best suited for interpreting signals on roads designed for human eyes, and conflicting signals from LIDAR makes disambiguation challenging when combining sensor types. Per a recent Musk interview.

How long before the NHTSA gets defunded?

Will robotaxis be a commodity by the time Tesla ships a viable product?

Even if Tesla gets it working, it will never be popular enough to justify their valuation. It's a niche product that will only compete with traditional taxis/ubers in urban areas, it has no chance of competing with car ownership at large, which is what Tesla's investors think it will do.

I'm not sure why anyone thinks it will compete with car ownership. You know what's better than an autonomous taxi, a private autonomous taxi that's just for you.

Kinda, but do also consider that most people can only afford (and have space for) one car, perhaps two if they're upper-middle class or have really old second hand ones, which means what they do buy has to cover most or all needs over the year.

On the other hand, if there were a way to rent any vehicle for about the same price as ownership (not that I think Musk will deliver this, but in principle someone else might), you can rent the one with the big boot once a week for shopping, the truck once every two months for a trip to the hardware store, and the tiny cheap city car for your commute.

>also consider that most people can only afford (and have space for) one car, perhaps two if they're upper-middle class or have really old second hand ones

In the U.S., something like 59% of households own two or more cars. 22% have three or more.

https://shunauto.com/article/how-many-people-have-a-car-in-a...

Ok, but (1) worldwide (see e.g. my own profile); (2) I said people, not households.

Working class American households have one car per adult and often a third beater for their teenager. In America, factory workers drive pickup trucks instead of more affordable cars, not because they need to haul things, but simply because they prefer the aesthetic.

This notion of Americans not being able to into car ownership is divorced from reality.

With all the maintenance and other hassles? No thanks.

EVs are virtually maintenance free. Versus the chances of getting into a car that was mistreated by some other POS?

Yes please.

The opportunity cost of having a vehicle sit unused in a parking lot 90+% of the time is really high. We've only tolerated it for so long because human labor is and has been more expensive than the vehicle's depreciation. Autonomous taxis potentially flip the script on that assumption.

Sure, rich people will always have enough money to afford luxuries like privacy and comfort, but most people are price sensitive and will opt for the cheapest transportation solution that delivers them where they need to go in the fastest amount of time. I would wager this will be a Waymo type autonomous taxi service but time will tell.

We're talking about personal vehicle ownership, there's no opportunity cost because you weren't going to make money on it anyway. My car sitting in my garage isn't costing me money nor is it losing value to me as a means of taking me and my stuff places. If you're one of the people that cares about the deprecation of your car's resale value then you're already doing the math on this and living a depressing life. But if you're like most normal car owners you bought a car to drive it and will replace it once you get your 200k miles out of it when its value is less than the $2000 minimum trade-in the dealership gives you.

The other problem is that I can't haul shit in a robo-taxi. Or drive long distances. People use their vehicles for more than just transporting bodies to and from work. And if I kill my neighbor I probably can't take his body in the trunk either. I would love to know what kind of life you and the other sibling commenters are living where a taxi service encompasses the totality of your transportation needs.

Assuming that you're keeping the registration up to date, your car does cost some money to sit in your garage even it if you'd consider it a negligible amount

> I'm not sure why anyone thinks it will compete with car ownership.

Maybe not 100% of cars, but a very good chance that it will eat into owning a 3rd or 2nd car in a household.

Not sure why this is downvoted. The whole premise of this robotaxi shit was started on the fact that your own, personal car, could be used as a taxi when you're not using it.

And also, the same reason people don't use taxis exclusively today is because you have to plan ahead or sit around and wait for a vehicle to come get you. Who is going to want to do that, when your empty driveway could house a car that pays for itself, makes you money, and is available whenever you need it?

I also don't think that taxis will be cheaper just because there is no driver. The savings is never passed down to the customer, never has been, and never will.

People who don't have cars today already manage to get around just fine, mostly due to where they live. This won't really change anything for them.

> I'm not sure why anyone thinks it will compete with car ownership

As you can see in your replies, there are a lot of techies who think it will. And furthermore, Tesla's valuation can only be made sense of if Tesla investors believe Tesla robotaxis will eat whole car industry, not just the whole taxi industry.

The majority of people lives in urban areas, especially in high income countries.

And the market for taxis is much smaller than the market for selling cars to urban people. Robotaxis aren't going to replace car ownership, but that's the delusion propping up Tesla's stock.

Well, public transportation is a thing in a lot of cities. Autonomous vehicles can fulfill the role of buses/shuttles/people movers with more flexible routes and without having to cost of a driver. Maybe allow a bit more cargo for people who want to get their groceries home. But yeah, taxis don't seem like the optimal shape here.

45% of households nominally in New York City have a car and that is by far the lowest of any American city. In the ideal American city for not owning a car, almost half of the households own one anyway! This is about the same in Europe too. Pick a major European city and you'll probably find that between a third to a half of the households in that city own a vehicle, despite the famous European public transit and habit for walking and cycling that online anticar activists love to brag about.

In every other American city besides NYC, most households have a vehicle. And besides virtually every American city save one being built for cars, you also have to keep in mind that the "most people live in urban areas" statistic counts all the people living in the suburbs around cities, e.g. people who have chosen a lifestyle literally built around car ownership.

"percent of households that have at least one car" isn't a particularly informative metric here. If that car is tied up by a commuter then the other household members could still use transportation. And it also doesn't tell us how many households have more than one car and could get rid of a second one if public transportation provided more flexibility.

I don't think that autonomous buses would be as cost-effective as private vehicles, since you'd want someone to be on board for security on a shared vehicle even if they're not in the driver's seat. Labor is the most expensive component of public transit operating costs, at least in my area. That said, in low-density areas it's not uncommon to see just one or two people on a big bus that comes once every 40 minutes in the middle of a weekday, and I'm sure running autonomous taxis instead would provide better service at a lower cost, given the same subsidies.

If Tesla only did trucking, that alone is a $1T industry. Now imagine they take a bite out of that, a bite out of Uber and Lyft, food delivery, transportation for an increasingly aging population, continue to make large investments in energy, robotics, etc. It's not that crazy of a valuation.

A valuation based on imagination?

Ah yes, the Tesla moon math. Just pull TAMs out your behind and give Tesla a % of it, that is how a true stonk "investor" does it sir! Did you come here straight from Wallstreetbets?

I wasn't aware that a car could fit in an elevator and deliver to your door...

People pay for Uber and Lyft because they're too lazy to leave their apartment. Tesla would have to be significantly cheaper than its competitors in order to justify the front-door-tax.

Uber and Lyft necessitate leaving the apartment. How else are you going to take the rid somewhere?

I assume you mean food delivery, though. In my experience, it isn't about being lazy to leave the apartment. It's that the infrastructure where I live is so spread out that it'll be at least 30 minutes total of driving to get that meal, and that isn't accounting for cooking time. A 5 minute walk would be much more tolerable.

In the context of this subtopic, Uber refers to UberEats, which is a larger business than the rideshare.

There's no way people will pay more money just to not walk to their door or lobby. Pricing always wins in mass market consumer products. The human option will be the luxury

They already did that. They released Tesla Semi 8 YEARS ago to revolutionaize the trucking industry.

Search for "Tesla" on HN, it will come up near the top. As the second most upvoted comment rightly note "Anyone else find it odd they don't talk about the battery?". What's really funny though is that 8 YEARS later, they still don't talk about the battery !

Anyway, Tesla is a scam. The company is real, and worth maybe $50B at the very optimistic valuation - their only real business is selling cars and getting subsides directly or indirectly. The sales are falling like a rock (thanks to Elon and the competition showing up) and subsides are going away soon pretty much everywhere. They are cooked.

Related: Elon Musk doesn't want you to know Tesla's response to the NHTSA's Robotaxi questions | https://news.ycombinator.com/item?id=44366731

Is this happening now because Trump doesn't like Elon any more?

Do you think the trump admin is faking the footage of the tesla taxis breaking the law?

No, but I do expect Elon to make some nice tweets about Trump and the next thing you know, murder by robotaxis will be legal and any states attempting to regulate them will be prosecuted.

i think it just means he can't just doge the regulators out of existence this time.

> i think it just means he can't just doge the regulators out of existence this time.

I don't believe this "contact" by regulators has shown that to be the case. However ultimately my point is that this "contact" happened because Tesla's taxis are failing at obeying the law and believing the regulators are out to "get" Tesla (or any company for that matter) is partly why America is facing its current challenges.

Maybe not, but if Trump were still a fan of Elon's, that footage would never have surfaced.

Regulators about to be fired and decimated by DOGE

2024: Waymo and Zoox are under federal investigation as self-driving cars allegedly behave erratically

https://www.cnn.com/2024/05/14/business/self-driving-cars-wa...

I expect that Tesla's FSD progress will accelerate now.

They are now operating a service where Tesla themselves is responsible for every mistake the cars make. Operating in public. Having the media and the authorities breathing down their neck.

Plus they have a clear benchmark: Growing faster than Waymo.

This should motivate everybody at Tesla to work their asses off. Including Musk.

You're assuming they weren't already operating in public before, and were just sitting on their laurels, rather than self driving being a really hard problem regardless of how hard you work on it. From what has been seen it's the latter, so I don't expect this to cause any meaningful change to Tesla's FSD progress.

"bad news is good news, actually"

> I expect that Tesla's FSD progress will accelerate now.

Just two more weeks guys, two more weeks! By 2014 all teslas will ship without a steering wheel, and by 2022 we'll be on Mars!

Reply on news.ycombinator.com