> That confidence is demonstrated by Mercedes’ decision to assume liability for the vehicle while Drive Pilot is in use.
This is a huge step forward, and will hopefully encourage other automakers to move this direction, as well. Volvo said many years ago that they would assume responsibility if autonomous systems caused collisions, but they have yet to actually implement this in their terms, as far as I know.
> We're already paying so what practical difference does it make to the owner of that car?
Vehicle accidents can extend past body damage and minor injuries and go well over the limits your auto insurance covers. As an example, if the software severely injures or kills a pedestrian in a crosswalk in a way that the driver couldn't have reacted, the liability is going to matter a _lot_ more than just paying off insurance.
Frankly, self driving cars aren't going to be worth more than slightly better (or slightly more stressful) cruise control until the company actually assumes liability -- which no one has done, until (presumably) now. So Mercedes stepping up could be a big deal.
Also... Does regular auto insurance cover the case where the driver has their hands off the wheel and software is in control?
The self-driving taxis are, like other taxis, taking responsibility for their safety during the journey.
Suppose Alice gets a Waymo to the bar, and Bob gets an Uber. On the way, bad luck they both crash into different idiots who don't think red lights apply to them. Neither Alice nor Bob are responsible. Waymo handles Alice's situation and I don't know whether Uber owns the problem or they put it on the driver, but they can't land it on Bob.
Long before Waymo was available to the general paying customers, one of their demos was a blind man gets a free journey, because of course, no problem, blind people travel by taxi all the time, you don't need to see the traffic, the taxi driver does that, Waymo is just a taxi.
A lot of taxi drivers discriminate against people with disabilities (some uber drivers do this too, they see you with your cane or walker and just drive past and cancel). Perhaps except in San Francisco where paratransit is a majority of taxi fares.
Robotaxis don't discriminate. You can get banned after the fact if you do prohibited things like snorting coke or having sex in the robotaxi, but they don't see a cane and assume you're not going to pay the fare.
There are some liabilities that cannot be assumed. For example, if law enforcement decides that a death in an accident is manslaughter, all the insurance in the world will not prevent you from being prosecuted and potentially jailed.
Your insurance is paying for you for errors when you drive. MB's effectively (and probably literally) got an insurance policy for errors when the software drives. The process is the same for any auto accident: figure out who's driving and contact their insurance (or sue if there's no insurance). Two drivers may argue for who was driving - human/human or human/software driver.
There's a fair chance they're self-insured unless that's somehow illegal in Nevada or California. Mercedes can afford whatever the sort of minimums are likely to be, because they're a large car company, not an individual.
Self-insurance is almost invariably cheaper if it's an option. I guess an insurer might give you a very attractive rate to get some exposure/ experience in an emerging market, but probably not.
> Even when you cause an accident, you don't pay. Your insurance company does. You might "pay" with increased future insurance rates.
The way I understand auto insurance is, you're always liable for the damage caused, assuming you're at fault. You have a contract with the insurance carrier to transfer the liability from you to the insurance company for a specific damage amount and causes.
> So what exactly does it mean that Mercedes "assumes liability"?
You have a contract to transfer liability in when the system is active, similar to the contract you have with your insurance.
> Are they going to reimburse insurance company?
That is possible, it is also possible Mercedes will pay first.
>We all have (or supposed to have) mandatory insurance. We're already paying so what practical difference does it make to the owner of that car?
You do pay a deductible in addition to your almost guaranteed substantial rate increase
> What is the process and what happens?
We don't know but we do know that Mercedes was required to provide proof of insurance or a $5 million bond to the CA DMV as a condition of the permit.
> Are they going to reimburse insurance company?
probably. Even in a normal crash, you contact your own insurance company and they will try to see if they can push liability to someone else, usually the other party or parties in the crash. Drive Pilot just adds another party they can attempt to push liability to. They will negotiate on your (really their) behalf and try to assign blame. In cases where your insurance and Mercedes disagree then it would likely go to court.
> We're already paying so what practical difference does it make to the owner of that car?
Even if you assume $0 deductible and $0 increase, the assumption of liability is a very strong signal that Mercedes is very confident in the system.
You can also peruse the regulation that they are licensed under which provides a lot of information:
It isn't a huge step forward until this is actually tested. It means they're paying insurance premiums, but nothing more than that until we actually have case law precedent around this.
You're in the driver's seat with the car in Level 3 mode, the car kills someone, their family sues you, then what? We have no idea.
> Volvo said many years ago that they would assume responsibility if autonomous systems caused collisions,
Volvo has a car model with some sort of crash avoidance technology that has so far had ZERO accidents resulting in human injury (or maybe death, I dont remember).
Don't get me wrong -- I love Volvo. I've owned a 240, a 740, an S70, and an XC70, and I tried to buy an S60 in 2021 (but the dealer promised they could get me one, strung me along for three months, and then finally admitted they didn't have any allocated).
Volvo's target of zero fatal collisions is absolutely admirable. They're the only auto company I'm aware of to set this kind of ambitious target and announce it to the public. Mercedes is probably the only other automaker with a track record of safety innovation on par with Volvo.
pretty much. Also, the limitations imposed will do an incredible amount to bring the risk down for them:
> To initially activate the system, viewing a seven-minute instructional video is also required. Operators are also responsible for keeping their vehicle in good condition. The sensors must be clean, tires must be in good condition, and the vehicle must also be properly maintained and serviced.
This probably means also if you let someone else use the system who hasn't gone through the education, they won't assume liability. It's literally gonna be an insurance-like system if theres ever an incident, the last time you got a car wash might be the difference between a claim being denied and approved. Also all the cameras recording the inside and out will help mercedes a lot, and they're confident given a good faith operator that their system won't fail.
I think I'd reserve judgement on that, I have heard of the pedestrian / cyclist detection being way behind the automobile detection. If it turns out safer, then I'd be delighted.
The only way I communicate with drivers when I'm a pedestrian, or biker is by eye contact, voice and hand gestures. When driverless cars can translate these, I'll be interested. Until then, the idea of robo-vehicles whirling around me completely unresponsive to my human communication methods gives me great concern.
While I respect what you're saying, I'd add that literally every bad interaction I've had with traffic has been down to inattentive driving. To most drivers, my signals are incidental and largely irrelevant information. Far more threatening is people not looking when leaving driveways or using bike lanes as right turn lanes, which seems like the sort of things an autonomous vehicle would catch.
Every day on my way to work I go through a 4 way stop. 1/4 of the time there's someone on the other side of me, who arrived at about the same time as me and one of us needs to decide who goes first. How do we do this? Hand gestures. We both (drivers) instantly know what the hand gesture means (after you) however with a robo driver I would just have to wait, and I guess see what it was going to do first?
On my way back from home I cross a bike trail with a stop sign for cars and bikes. Many times there are people about to enter the crosswalk. How do we communicate? Hand gestures and smiles. I can't imagine the terror I'd feel trying to cross in front of a robo car.
Oh yeah for sure, I'm not saying they're useless. I'm just saying I don't think an autonomous car is going to be a much larger threat in that situation. And even if it went at the same time you did, it can either detect that and wait for you to cross, or simply wait for you regardless.
In both cases I would assume for the bike to get the right of way and the car would simply stop and wait for the biker to go.
Either way, for the foreseeable future a human will still be sitting in the driver seat ready to make eye contact and wave you on.
What I'm more excited about is the idea that more and more cars will be able to automatically stop if they see people/obstacles in the street. I've almost gotten hit plenty of times while crossing intersections with stop signs because the car pulled up past the sidewalk too quickly.
In the 4-way stop example, isn’t that what right-of-way rules are for? Not following those rules because you saw a hand gesture seems like it puts you at risk if there is an accident.
As a pedestrian in an urban area, my main purpose for communicating with drivers is to figure out whether they're going to actually give me the right of way that I (by law) have at stop signs. This is necessary because ~30% of human drivers choose to break the law and go first anyway. I'm optimistic about autonomous vehicles because (I assume) they're programmed to give pedestrians right of way 100% of the time.
I have no clue how thoroughly implemented this is, especially for pedestrians other than police, but Waymo does have support for hand signals and body language:
Exactly. There's a reason that robots on an assembly line are behind barriers. And THOSE machines are far more controlled than a random robotic vehicle.
> Drive Pilot cannot be retrofitted to existing EQS or S-Class sedans, though. The system must be ordered at the time of purchase, but the cost is a surprise. Rather than charge for all the equipment upfront, Drive Pilot will initially be rolled out as a $2,500 annual subscription.
Interesting that it's a subscription but can't be added after-the-fact.
> it's a subscription but can't be added after-the-fact
It’s financing for the equipment, additional insurance and a subscription to updates rolled together. Curious to see how those break down across the $2,500 price.
It'll also be interesting to see if/how they improve the capabilities over time. I'd be more likely to consider something like this if I were guaranteed to lock in the price (even if the tech improves over time, and you can use it at faster speeds or on more roads).
> For Drive Pilot to activate, the vehicle must be traveling no faster than 40 mph (65 km/h), it must have a vehicle in front to follow, road conditions must be dry and clear, the lane markings must be detectable, and the route must be pre-mapped by the system.
At $2500 per year, I would expect that this would cost on average $100 per trip when actually usable. Maybe more.
This is true, but then each "trip" is reduced to the part of the commute that has sub-40 mph traffic. I guess that could be much of your commute if you're a poor schlub, but then what are the odds you can afford an S-Class Benz?
That's an interesting way to think about it. For my city (Edmonton) it seems like if you're well off you're going to be living in a community in the city. It's the middle-class who are paying for a much larger home for a less convenient location.
This is in line with what I was trying to say. Basically, you might have a lousy long commute where this would be very valuable. But then you probably can't afford this car.
Ah, I got confused by the "I guess that could be much of your commute if you're a poor schlub" part, because I would assume they'd get up to highway speeds to end up wherever they live, whereas someone more well off wouldn't hit that.
Where on earth do you sit in traffic that doesn't burst up above 40mph every now and then? It's called stop and go, not go consistently below 40mph. Sounds like this wasn't well thought out. In LA, you'd be in traffic going 80, then screech to a halt, then back up to 80. Even on my commute into Downtown Seattle where people drive slow as fuck, there were always openings between onramps where it was no problem to hit 40. That's such a low bar.
Yeah, I agree. This seems like it would be useful for stretches on 405, but not likely the bulk of a trip. Where I live, I could see it being useful for a decent chunk of 101, which could amount to 30 mins each way, each day.
I wonder if it drops down into a less-capable driver-assist mode when the speed goes higher, such that you can still just monitor the system for nearly all of your trip. Or does it drop you out into full human-driving mode?
It depends on the climate where you live though, right? Where I am there's plenty of congestion. There's also plenty of rain, snow, slush, and road construction. The snow removal and construction really take a toll on the lane markings, so detectable lane markings isn't a guarantee either.
This would be usable during rush hour on 101 going through SV, and in many parts of LA. Some people could use it 2x/day. However, those people are not likely to be able to afford an expensive new Mercedes, since super-high earners probably aren't the people going into the office every day. Also, people wealthy enough for an S-Class Benz probably don't live super far from the office, since they can afford to live pretty much anywhere.
Some wealthy people like driving. An S-class is one reasonable thing you might purchase if you are wealthy but enjoy driving and would prefer to do more of it than (likely) is necessary or ordinarily appropriate for your station in life. It's a nice car, it's too expensive if you need to justify the price, but if you don't care about the price then why not?
So if you like driving, then you might choose to live somewhere that's a "longer commute" even though you needn't - as this is part of your day which you enjoy. Now, that's probably not going to be bumper-to-bumper freeway traffic, but it absolutely might include country roads at limited speed - I said enjoys driving not racing, if you like track racing and have $$$ you can indulge that without owning a road car at all.
I do. I have a Bolt EUV and turn on SuperCruise (hands-off but only on divided freeways) when I can, adaptive cruise control when I can't.
The car handles speeding up and slowing down for me based on what the car in front is doing. I just have to steer. I know that jumping lanes is not going to get me there much sooner but it's hard to resist "competing" to get there first.
My car with radar cruise and camera based lane centering: can go any speed, doesn't need a leader, handles terrible conditions (notably it sees better than I can during night time rain with glare from on-coming traffic), comes to a stop, starts from a stop, and handles some pretty crappy lane markings.
It don't handle sharp corners and doesn't auto-drive. It does, however, take 90% of the effort out of the place I spend most of my driving time - highways.
I thought it was odd that an independent journalist wouldn't even mention Tesla's Autopilot in an article about self driving. They did mention several other glorified lane assist systems (Bluecruise, SuperCruise, Lane Assist), but not Autopilot? Bluecruise is just about to hit 100 million cumulative miles compared to Autopilot's 3.5 billion cumulative miles.
Honestly, with the high price and relatively severe limitations this system seems like more of a gimmick than anything else. If you commute in California every day in heavy traffic I can see maybe considering it, but that price point is discouraging.
It is important to note that they are using the UNECE Level 3 certification loophole which technically demands that the company takes responsibility, but sets the minimum intervention interval for the driver at zero seconds for the under 60km/h case. That means, as long as the car exits its autonomous mode before a crash, the driver is instantly fully responsible. It's level 2 in all but name.
If true, I’m not sure I’d feel comfortable using this feature. This is separate from the fact that the EQS line all look like beans.
I doubt that's true. I don't know why it wasn't in the actual article, but the byline of the title is: "PREPARE TO RESUME CONTROL IN 10 SECONDS...", which I think is the minimum time that it takes to exit autonomous mode. the ars technica article is just kinda garbage honestly.
> However, if the system senses a pedestrian or a bicyclist on the highway, detects an emergency vehicle or enters a construction zone, Drive Pilot will hand control back over to the driver. There are audible, visual and tactile warnings — the seatbelt tightens — and during that time the car is still actively driving.
> If after 10 seconds I do not respond, Drive Pilot brings the car to a stop in the lane,
> Drive Pilot is currently legal on California and Nevada freeways. Of course, all states are different in terms of traffic liability, but Mercedes-Benz says it stands behind its product as long as the driver operates the system correctly in a properly maintained vehicle.
I haven't seen any evidence that Mercedes has published anything about the specific intervention interval, but even so liability is discovered in court and Mercedes can't just wave a wand and say "we aren't liable because we deactivated 0.001 seconds before the crash!".
It would definitely be very disappointing if they tried to exploit the loophole but if they did its very possible that a judge or jury would look at Mercedes encouraging the driver to watch a movie and tell them and their loophole to pound sand.
That’s a good point. And that’s not even to mention the hit to their brand reputation, which historically they seem to care about more than some other car companies.
> The system will also deactivate when other requisite conditions are not met, as well as when an emergency vehicle is detected, when approaching toll roads, in construction or special regulation zones, or _if a pedestrian or cyclist is detected_
Wait for the abuse other road users will get for causing the expensive autopilot to switch off.
This may not sound like much, until you realise that that other self-driving vehicle is currently in over 100 court battles and police investigations due to serious faults and in some cases injuries and even deaths.
66 comments
[ 0.20 ms ] story [ 104 ms ] threadThis is a huge step forward, and will hopefully encourage other automakers to move this direction, as well. Volvo said many years ago that they would assume responsibility if autonomous systems caused collisions, but they have yet to actually implement this in their terms, as far as I know.
Even when you cause an accident, you don't pay. Your insurance company does. You might "pay" with increased future insurance rates.
So what exactly does it mean that Mercedes "assumes liability"?
What is the process and what happens?
If that system causes an accident, how do you contact Mercedes? How do your prove you were running it within desired parameters?
Are they going to reimburse insurance company?
We all have (or supposed to have) mandatory insurance. We're already paying so what practical difference does it make to the owner of that car?
Those are really basic question that no one seems to address.
Everyone just repeats that vague marketing line like it means something.
Vehicle accidents can extend past body damage and minor injuries and go well over the limits your auto insurance covers. As an example, if the software severely injures or kills a pedestrian in a crosswalk in a way that the driver couldn't have reacted, the liability is going to matter a _lot_ more than just paying off insurance.
Frankly, self driving cars aren't going to be worth more than slightly better (or slightly more stressful) cruise control until the company actually assumes liability -- which no one has done, until (presumably) now. So Mercedes stepping up could be a big deal.
Also... Does regular auto insurance cover the case where the driver has their hands off the wheel and software is in control?
Suppose Alice gets a Waymo to the bar, and Bob gets an Uber. On the way, bad luck they both crash into different idiots who don't think red lights apply to them. Neither Alice nor Bob are responsible. Waymo handles Alice's situation and I don't know whether Uber owns the problem or they put it on the driver, but they can't land it on Bob.
Long before Waymo was available to the general paying customers, one of their demos was a blind man gets a free journey, because of course, no problem, blind people travel by taxi all the time, you don't need to see the traffic, the taxi driver does that, Waymo is just a taxi.
Robotaxis don't discriminate. You can get banned after the fact if you do prohibited things like snorting coke or having sex in the robotaxi, but they don't see a cane and assume you're not going to pay the fare.
Yes. If you're driving a Tesla, you're covered by your auto insurance even if it is in FSD mode.
Self-insurance is almost invariably cheaper if it's an option. I guess an insurer might give you a very attractive rate to get some exposure/ experience in an emerging market, but probably not.
The way I understand auto insurance is, you're always liable for the damage caused, assuming you're at fault. You have a contract with the insurance carrier to transfer the liability from you to the insurance company for a specific damage amount and causes.
> So what exactly does it mean that Mercedes "assumes liability"?
You have a contract to transfer liability in when the system is active, similar to the contract you have with your insurance.
> Are they going to reimburse insurance company?
That is possible, it is also possible Mercedes will pay first.
>We all have (or supposed to have) mandatory insurance. We're already paying so what practical difference does it make to the owner of that car?
Potentially lower insurance costs.
You do pay a deductible in addition to your almost guaranteed substantial rate increase
> What is the process and what happens?
We don't know but we do know that Mercedes was required to provide proof of insurance or a $5 million bond to the CA DMV as a condition of the permit.
> Are they going to reimburse insurance company?
probably. Even in a normal crash, you contact your own insurance company and they will try to see if they can push liability to someone else, usually the other party or parties in the crash. Drive Pilot just adds another party they can attempt to push liability to. They will negotiate on your (really their) behalf and try to assign blame. In cases where your insurance and Mercedes disagree then it would likely go to court.
> We're already paying so what practical difference does it make to the owner of that car?
Even if you assume $0 deductible and $0 increase, the assumption of liability is a very strong signal that Mercedes is very confident in the system.
You can also peruse the regulation that they are licensed under which provides a lot of information:
https://www.dmv.ca.gov/portal/vehicle-industry-services/auto...
https://www.dmv.ca.gov/portal/news-and-media/california-dmv-...
You're in the driver's seat with the car in Level 3 mode, the car kills someone, their family sues you, then what? We have no idea.
Volvo has a car model with some sort of crash avoidance technology that has so far had ZERO accidents resulting in human injury (or maybe death, I dont remember).
They are slowly getting there.
Volvo's target of zero fatal collisions is absolutely admirable. They're the only auto company I'm aware of to set this kind of ambitious target and announce it to the public. Mercedes is probably the only other automaker with a track record of safety innovation on par with Volvo.
> To initially activate the system, viewing a seven-minute instructional video is also required. Operators are also responsible for keeping their vehicle in good condition. The sensors must be clean, tires must be in good condition, and the vehicle must also be properly maintained and serviced.
This probably means also if you let someone else use the system who hasn't gone through the education, they won't assume liability. It's literally gonna be an insurance-like system if theres ever an incident, the last time you got a car wash might be the difference between a claim being denied and approved. Also all the cameras recording the inside and out will help mercedes a lot, and they're confident given a good faith operator that their system won't fail.
I think that this will have a big calming effect on traffic and the increased safety may encourage more people to cycle.
Examples:
https://www.forbes.com/sites/carltonreid/2022/05/16/semi-aut...
https://arstechnica.com/cars/2023/05/teslas-full-self-drivin...
Every day on my way to work I go through a 4 way stop. 1/4 of the time there's someone on the other side of me, who arrived at about the same time as me and one of us needs to decide who goes first. How do we do this? Hand gestures. We both (drivers) instantly know what the hand gesture means (after you) however with a robo driver I would just have to wait, and I guess see what it was going to do first?
On my way back from home I cross a bike trail with a stop sign for cars and bikes. Many times there are people about to enter the crosswalk. How do we communicate? Hand gestures and smiles. I can't imagine the terror I'd feel trying to cross in front of a robo car.
Either way, for the foreseeable future a human will still be sitting in the driver seat ready to make eye contact and wave you on.
What I'm more excited about is the idea that more and more cars will be able to automatically stop if they see people/obstacles in the street. I've almost gotten hit plenty of times while crossing intersections with stop signs because the car pulled up past the sidewalk too quickly.
Maybe I’m misunderstanding your example.
https://topdriver.com/education-blog/4-rules-4-way-stops/
https://www.cnet.com/roadshow/news/waymo-self-driving-cars-p...
https://medium.com/waymo/safely-sharing-the-road-with-cyclis...
Interesting that it's a subscription but can't be added after-the-fact.
It’s financing for the equipment, additional insurance and a subscription to updates rolled together. Curious to see how those break down across the $2,500 price.
I've decided that I have no interest in ever trying features in a car that will disappear once the subscription ends. Just not worth it.
At $2500 per year, I would expect that this would cost on average $100 per trip when actually usable. Maybe more.
If you live somewhere where you only experience highway congestion 25 times a year, you probably aren't the target market.
I wonder if it drops down into a less-capable driver-assist mode when the speed goes higher, such that you can still just monitor the system for nearly all of your trip. Or does it drop you out into full human-driving mode?
So if you like driving, then you might choose to live somewhere that's a "longer commute" even though you needn't - as this is part of your day which you enjoy. Now, that's probably not going to be bumper-to-bumper freeway traffic, but it absolutely might include country roads at limited speed - I said enjoys driving not racing, if you like track racing and have $$$ you can indulge that without owning a road car at all.
The car handles speeding up and slowing down for me based on what the car in front is doing. I just have to steer. I know that jumping lanes is not going to get me there much sooner but it's hard to resist "competing" to get there first.
My car with radar cruise and camera based lane centering: can go any speed, doesn't need a leader, handles terrible conditions (notably it sees better than I can during night time rain with glare from on-coming traffic), comes to a stop, starts from a stop, and handles some pretty crappy lane markings.
It don't handle sharp corners and doesn't auto-drive. It does, however, take 90% of the effort out of the place I spend most of my driving time - highways.
Putting lives at risk with a "beta" self-driving vehicle is insane.
> The incremental rollout will build consumer confidence along with its assumption of liability.
That sounds like it was writen by Mercedes-Benz.
Honestly, with the high price and relatively severe limitations this system seems like more of a gimmick than anything else. If you commute in California every day in heavy traffic I can see maybe considering it, but that price point is discouraging.
this is like Google Maps saying you’re finally allowed to rely on it for directions now
It is important to note that they are using the UNECE Level 3 certification loophole which technically demands that the company takes responsibility, but sets the minimum intervention interval for the driver at zero seconds for the under 60km/h case. That means, as long as the car exits its autonomous mode before a crash, the driver is instantly fully responsible. It's level 2 in all but name.
If true, I’m not sure I’d feel comfortable using this feature. This is separate from the fact that the EQS line all look like beans.
https://techcrunch.com/2023/09/27/first-drive-navigating-la-...
> However, if the system senses a pedestrian or a bicyclist on the highway, detects an emergency vehicle or enters a construction zone, Drive Pilot will hand control back over to the driver. There are audible, visual and tactile warnings — the seatbelt tightens — and during that time the car is still actively driving.
> If after 10 seconds I do not respond, Drive Pilot brings the car to a stop in the lane,
> Drive Pilot is currently legal on California and Nevada freeways. Of course, all states are different in terms of traffic liability, but Mercedes-Benz says it stands behind its product as long as the driver operates the system correctly in a properly maintained vehicle.
It would definitely be very disappointing if they tried to exploit the loophole but if they did its very possible that a judge or jury would look at Mercedes encouraging the driver to watch a movie and tell them and their loophole to pound sand.
Wait for the abuse other road users will get for causing the expensive autopilot to switch off.