Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Mercedes-Benz says Level 3 self-driving option ready to roll

thedetroitbureau.com

Post reply on HN
Print view View original

301–310 of 410 posts

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#301

Post by seanmcdirmid »

ClumsyPilot wrote: post #234
Traster wrote: post #60

Earlier quoted context omitted.

Just from a management perspective, even if you're super-confident in your engineers (which you shouldn't be), doing anything other than a very limited slow roll out would be negligent. This is a tectonic shift in liability. It's one thing if you sell 1,000,000 cars and 100,000 get in an accident. It's actually much worse to sell 10,000 cars and 1,000 of them get in an accident and the company is liable .

'This is a tectonic shift in liability. ' Its the only form of liability i consider acceptable - I woupd never use autopilot where I would be to blame if i fail to react in 50 milliseconds once the car fucks up

I think for this reason full self driving is going to be more feasible in taxi-style vehicles first. You hail the driverless taxi, like with another person driving the taxi the passengers are not liable. It’s a model we are already used to, at least. If the taxi company winds up being the car producer, then the liability is the same as if they didn’t produce but operated the car.

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#302

Post by Dave_Rosenthal »

amelius wrote: post #67

What I don't get: if deep learning can recognize a crosswalk with 99.99% accuracy, that's very impressive. Yet it is not enough.

You get it correctly. Deep learning isn't inherently good for cases where you need to be right 100% of the time. But we don't have a better tool so the industry is trying to see how far they can push it.

One big open question in self driving is: Do you have to be able to see/classify everything in the scene all the time to be a good driver? I think the answer is no (see: humans.) But, because most self driving stacks in the industry have chosen "3d classified representation of all of the objects in the scene" as a fundamental interface between perception and planning stages, there is a big challenge. (e.g. Representing uncertainty usefully is exceptionally difficult with this interface, among other things.)

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#303

Post by icelancer »

zydex wrote: post #281

A point I commonly hear made from both Tesla and Comma.ai for example is that Lidar is far too expensive with Waymo's vehicle costing a total of $200,000 and that cameras alone are sufficient for full self driving. I do think that cameras alone are probably sufficient for full self-driving but every time I hear this point I think to myself that self driving progress is moving so slowly on the camera-only front that L…

I work in a parallel industry where capturing motion at high speeds is critical, and in the last two decades, the cost of LIDAR and RADAR has barely come down - and the technical specs haven't gotten much better - while the cost of optical continues to absolutely crash through the floor and technical specs skyrocket.

Mobile phones and laptops are the obvious main reason for this. Optical tracking (cameras) have economies of scale that LIDAR and RADAR can't even come close to. LIDAR and RADAR are niche features in a specific set of high priced luxury goods while optical tracking is dominating everything from government/consumer surveillance to camera phones to sports motion capture and many, many more applications.

Optical tracking has so much pressure on the field to drive costs down and to increase feature sets (innovate) that RADAR/LIDAR don't. It's going to be this way for quite a long time, so I don't really see the costs for RADAR/LIDAR getting under control anytime soon.

EDIT: Optical tracking will be the dominant method of self-driving cars, I'm almost sure of it. The trained datasets have a huge advantage in this regard. Still, it's obvious that LIDAR/RADAR will have additive value down the line. It is just very hard to see how it becomes the primary technology. This is echoed in my field as well as many others - where RADAR dominated, machine learning / software + good enough optical tracking took over at 1-3 magnitudes of cost savings.

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#304

Post by Dave_Rosenthal »

rootusrootus wrote: post #95
everyone wrote: post #65

Earlier quoted context omitted.

Your argument works both ways. Lets say a truck is driving on the wrong side of the road and is going to flatten the mercedes (not too uncommon as truck drivers seem to be crazy and will overtake each other). Lets say it's possible to swerve onto the footpath and squash some pedestrians in order to avoid the truck. A human driver wont be solving the trolley problem here and will probably just die. It sounds Mercedes…

Read what the Mercedes guy actually said, though. Swerving is bad, because it in no way guarantees a better result, and increases the risk substantially of making things worse . Any tire grip you use to swerve isn't being used to stop, and the best way to minimize damage to the car, and by extension risk to the occupant, is to get kinetic energy as low as possible and rely on the safety cage for the rest. The magazin…

I think that as car perception systems start to have understanding of the 3d scene (i.e. not just adaptive cruise longitudinal controllers) joint control of the brakes and steering will becoming standard for accident avoidance and will be a big step forwards. It probably is the case that braking ends up being the "more useful" degree of freedom (for one, braking latency is lower than swerving latency) but swerving out of the way of a car can be incredibly useful in certain cases.

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#305

Post by Dave_Rosenthal »

qeternity wrote: post #3
pojzon wrote: post #2

If tesla still has issues with self-driving cars and they are like a decade if not more ahead of everyone else. I dont believe this PR stunt.

Wow. Or Tesla are the real masters of the PR stunt, and are not a decade ahead of everyone. EDIT I watched this comment drop from 15 pts to 5 pts. The Elon apologists are out in full force.

I believe that Tesla is ahead in the self-driving industry from an overall technology perspective. What they have failed to do, however, is to tackle the problem of when and how much the user can actually trust their system. This leaves users to struggle to build their own mental models of the same ("Oh, it likes this road", "I don't trust it at high speeds", etc.) This is terrible UX.

Mercedes seems to be stepping up to the plate to actually take a stab at this problem. Inevitably the answers to these questions are not super impressive compared to Tesla's 'hey, it might work in any situation, you never know' approach. But I suspect users will really like it, even though the underlying tech is likely far behind Tesla's.

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#306

Post by ehnto »

gota wrote: post #275
cortesoft wrote: post #269

Earlier quoted context omitted.

You are already going to be to blamed if you get in an accident for most hardware failures in a car. Why is this one so different?

I guess because appropriate usage and maitenance makes that kind of hardware failure extremely unlikely, no? I can be "warned" to check for issues if the car is older, creaking, seems to be "driving" different etc A mistake in a mostly completely uninterpretable bundle of software is definetely much sneakier. You can't prepare for it So I guess that's one argument. The other aspect of the same argument would be that…

I don't think the GP is 100% correct anyway, it depends a great deal on what happened to cause the accident. Brakes fail because you haven't changed the pads? Your fault. Steering shaft u-joint comes loose on your 2020 Ford? You were never expected to maintain that item in that time frame so is it your fault?

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#307

Post by TaylorAlexander »

fh973 wrote: post #299
TaylorAlexander wrote: post #290

Earlier quoted context omitted.

I would say the race isn’t over yet. One thing to consider is the training dataset. By using only vision, Tesla was able to start collecting data with their entire fleet starting many years ago. They probably have more edge case data than anyone else, in more diverse driving situations. They would not have been able to do this if they needed to use expensive lidar on company owned training vehicles. What we will like…

This point about Tesla has been made frequently in this discussion. I still wonder: how does Tesla store any meaningful video data in customer vehicles and transfer it? I know that other vendors have a hard time putting the necessary fast storage in the trunk and get the data off with 100G cables in the garage.

Their AI system has something called “shadow mode” where it can make observations without having any effect on the car. When Tesla needs to collect a dataset of something, like short videos of cars that put on their blinker but did not change lanes, or partially obscured stop signs, they can train a net to run in shadow mode on all their cars and collect more of this edge case. I presume that they also collect a lot of data around disengagements. And then they send the data presumably over Wifi when the customer gets home, or otherwise over the car’s 3G connection. But they collect targeted data, not mass dumps of drives.

Andrej Karpathy has described this system in various talks including on Tesla AI day.

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#308

Post by zaroth »

jhugo wrote: post #282
zaroth wrote: post #274

Earlier quoted context omitted.

No, this reasoning is flawed. The baseline safety of cars is an absolute shit show. 30,000 people dying every year and trillion dollars of damages and lost productivity. Car A can enable FSD in all cases, be safer in all cases, but still not be a good deal economically for the manufacturer to accept liability. Car B could be making drivers less safe overall by preventing them from enabling FSD in most cases, but at l…

Your comment is predicated on the assumption that FSD (the actual system installed in the car, not a future theoretical perfect system) is safer than the average driver in the situations where Mercedes currently disables it. I'm not sure we have data to support that? We know Tesla's autopilot is safer on average, but most of those miles will have been driven in the situations where Mercedes allows it to be used.

I am saying such a scenario may possibly exist, not necessarily that it does exist.

Mercedes could be increasing the number of overall deaths by limiting the availability of the feature and still be reducing their liability for when the system is in use.

Let’s say with FSD on all the time that instead of 30,000 people a year dying that only 20,000 people a year die. Would a company accept the liability?

What if the death rate was 10,000? 1,000? 100?

If FSD could prevent 29,900 deaths a year but still see deadly failures 100 times a year, would a company say “I accept the liability”?

So you see, perhaps it’s crucial that companies not be able to be sued out of existence even if a few hundred people a year are dying in exceptional cases under their software, in order to prevent over a quarter million deaths and untold number of maimings every decade.

Also consider in this ethical and legal liability dilemma that these populations are not necessarily subsets, but could be disjoint populations.

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#309

Post by Dave_Rosenthal »

zydex wrote: post #281

A point I commonly hear made from both Tesla and Comma.ai for example is that Lidar is far too expensive with Waymo's vehicle costing a total of $200,000 and that cameras alone are sufficient for full self driving. I do think that cameras alone are probably sufficient for full self-driving but every time I hear this point I think to myself that self driving progress is moving so slowly on the camera-only front that L…

The conversation is complicated because LIDAR is used for many purposes in self driving. It's used for localization, it's used for objection detection, it's used for classifications, it's used for understanding occlusions, it's used to get accurate 3D positions and speeds of objects, etc. For each of these cases there is a way to use just cameras (or camera augmented with radars), but often with some significant performance penalty.

One of the big challenges is that most self driving stacks have an interface between the perception and planning stages that is specified to be a 3D model of the world. LIDARs are particularly helpful at creating 3D models because that is essentially their native data product. So, for "traditional" AV stacks that use this interface, LIDARs are bound to improve performance a lot.

If you use a different approach, say pure imitation learning off of sensor data, you might find that LIDARs are not as important. (intuition: Humans drive well without understanding super accurate positions or velocities of objects.) Though Tesla isn't taking a pure imitation learning approach (yet), they are more in line with this strategy.

All this said, I don't think that having or not having LIDARs is a major factor in the progress of self driving. It's just a way to use money to improve perception performance (and reduce data labeling cost!) but neither is the major blocker for the industry. If we extrapolate from the last 10 years of progress, it seems like high-level self driving is going to take a while and I think that it's likely that LIDAR prices will have fallen dramatically by then and we'll see them as part of the overall sensor constellation on most vehicles.

Re: Mercedes-Benz says Level 3 self-driving option ready to roll

#310

Post by TylerH »

seanmcdirmid wrote: post #301
ClumsyPilot wrote: post #234

Earlier quoted context omitted.

'This is a tectonic shift in liability. ' Its the only form of liability i consider acceptable - I woupd never use autopilot where I would be to blame if i fail to react in 50 milliseconds once the car fucks up

I think for this reason full self driving is going to be more feasible in taxi-style vehicles first. You hail the driverless taxi, like with another person driving the taxi the passengers are not liable. It’s a model we are already used to, at least. If the taxi company winds up being the car producer, then the liability is the same as if they didn’t produce but operated the car.

That's not really "full self driving" if you are referring to the SAE scale. Level 5, the top, is fully autonomous driving, everywhere. What you are referring to is level 4; fully autonomous driving in specific geofenced areas on predetermined routes. Think Delamain in Cyberpunk 2077.
Post reply on HN