"When Volvo does sell a self-driving car, it will accept full legal responsibility for any accidents that may come. And that is reasonable, because we told the customer he could do something else while the car drives itself.”
It's fine to get on a bus or get in a taxi while drunk, I don't see any reason a 100% autonomous car should be any different... in theory. The dilemma here is can the person take control of the vehicle?
Being able to get hammered, and then pass out in the back seat of your car while it drives you home is the dream. But in the slim chance it does get into an accident, there needs to be a way to prove you really were blacked out in the backseat, and the car crashed on its own. And not that you drunkenly grabbed the wheel and tried to do a u-turn to McDonald's.
I wouldn't call a car self-driving if the occupant needs to be involved.
Unfortunately marketing has figured out that "self-driving car" is the next cool thing, so they're rebranding driving assist technologies as "self-driving".
(I am aware of the definition of different "levels" of self-driving cars; from my perspective, that's marketing. Either the car drives itself, or it does not.)
It is really fascinating to see the self-driving car space. I have been assuming that Google is way ahead of everyone else, and that other tech companies literally didn't stand a chance including Apple.
If traditional car companies are not too far behind Google, and are able to confidently compete in this space, then Apple definitely has a chance to compete in this space.
When self-driving cars become a reality over the next 10 years it's probably going to reduce the number of cars sold massively.
Many or most adults in the developed world currently own a car. If you can share them even between, say 3 people the number of cars required will plummet. There are credible predictions around as well that even say up to 10 people might be able to share a car.
There is already over capacity in car production globally. Self-driving cars are likely to drive that even further.
It's not a good market for a company that isn't in the game to be going in to. Google are different because they seem to be selling the technology to enable other manufacturers to sell self-driving cars.
Not sure why you got downvoted. While I think the number of cars sold will decrease, I don't think it will be by very much. The biggest problem is that most people go to work and go home at roughly the same times during the day, so realistically you can't have more than maybe 2-3 people sharing a car if they're both employed. Yes, some fraction of a city's residents work non-standard hours, but the majority are on something close to a 9-5 schedule.
Also keep in mind that while the number of cars sold might go down, it's also fairly likely that the margin on each car will go up, and the total number of miles driven will go up as well. Basically when you take the unwanted part of driving out of the equation (sitting in traffic wasting time) and don't need driver attention anymore, you can start adding all kinds of value-add features to vehicles like giant in-vehicle displays or massage chairs. And if you can do whatever you want while on the road, longer commutes aren't so bad, so it's easy to put more miles on a car in a year, which means more replacement parts, service visits, etc. over time.
The market will no doubt change significantly, but existing OEMs are by no means going to be excluded entirely.
I disagree. While the number of cars may plummet initially, when you share cars between multiple people, the mileage on those cars is going to increase dramatically. Shared cars may end up hitting 300,000 miles in 5 years rather than 15.
Unless self driving cars dramatically increase the lifespan of the average car, things will balance out in a few years.
Do first owners sell their cars because they become unusable or because they become unfashionable or some other reason?
In two car households does the second car primarily get purchased for use when the first car is sitting idle but just in the wrong location?
Might there be a change in ownership as a sort of "buy-to-let" system comes in where owners wouldn't be able to buy an autonomous vehicle but afford it as they send the vehicle out to work doing deliveries/taxiing people during the times they don't need it?
It seems quite hard to predict how autonomous vehicles will change things; a lot of it will depend on regulations on how they're manned - will a driver with a license still be required at all times? Will there be new licenses? Will autonomous vehicles have extra restrictions on their speed or which roads they use? Will there be protectionist laws to prevent them picking up and taxiing people? The surrounding legislation will probably be as important to the way the evolution of car use happens.
People's appetite for suburban environments is insatiable. Given the chance to work or relax 1 hour on a laptop before getting to the office and after getting to the office will exacerbate sprawl.
Car sharing is an interesting possibility, but it may turn out people will live at the farthest distance they can possibly live [due to costs], thus will be unable to afford even a five minute detour that is critical for enabling sharing.
I expect real estate to explode, car production to increase, possibly significantly, and we'll spend a whole lot more time in our private mobile offices.
Also some of the freight traffic will be shifted to nights; self-driving cars will allow door-to-door service, so number of parking spaces will be reduced; also when no driver is needed you get that extra seat.
I had this somewhat naive view as well. But there's actually some very good research and work done by traditional car manufacturers. Mercedes-Benz has done some very good work in the trucking department for example (from talking to some trustworthy AI researchers at a conference).
I do feel that Google (and Apple) have an edge. Traditional car companies (at least in Germany) are driven by engineers and usually decision makers tend to be ex-engineers. They have a subtly different attitude towards programming and software (often only working with Matlab and the like) and I often feel like software isn't all that important to them. Alas software is eating the world is especially true in this segment.
I don't know how to succinctly say it without sounding like "ZOMG programmers > engineers haha" elitism but I think "software first" thinking and not "engineering first" thinking is the principle that will succeed in this space.
Generally I would say that a company like Google is researching self-driving, while a company like Volvo is building self-driving cars. It's very likely that Google has the edge when it comes to something like autonomous taxis (i.e. pods), but that is still some years ahead and who knows how much of that will have been a relevant effort then. I guess they will have a lot of patents though.
Android demonstrates Google is happy to have someone else build the physical hardware. The current auto industry is predicated on many independent suppliers and Google could easily become yet another one.
Yes, but the technology they are researching (full autonomous) is still further down the road than something like highway autonomy only. Which would still give you most of the benefits of an autonomous car in boredom, accident and unproductive time reduction. While not requiring as expensive parts and as much research.
If you do a self-driving pod then it's another story because then you share the cost between passengers and you can engineer it to only go city speeds for instance.
What I'm essentially saying is that at the point Googles research becomes "a reality" the car becomes the weak part of the equation. When we have full autonomy, I might almost just take a train (that will probably always be much faster than cars because of physics) and catch another autonomous pod at the other station. This is of course what could/should happen in theory, not what will happen.
A Smart car is already fairly close to the pod concept in terms of form factor. Google Pods biggest difference is moving the seats further back. IMO, if Google get's the software down most company's could put out a self driving car within 1 or 2 model years. However, people are going to take much longer before people are going to take much longer before buying cars that can't be driven.
Apple? I wonder why you say this? Thy haven't publicly announced anything to do with cars until now as far as I am aware. Admittedly they are very secretive about things like that, but I would be surprised if they had any sort of lead over the competition other than with the brand name.
To me Google had an edge because Thrun came from winning the DARPA challenge, after years of 'nobody finished the race'. Even if traditional car brands had unused research for SDV, I assumed it wouldn't be as advanced as the DARPA project.
It's easier for BMW to implement Mobileye in existing model than Apple to build self driving software, self driving sensors, good car and efficient production line for it all.
Out of the car companies out there, Volvo is one of the better ones as far as sw goes. I've heard good words about the way things work in Daimler, Volvo and WV but not many good things about the way other companies treated software.
It's actually quite amazing how much the self-driving car hype is focused around Google and Tesla. This thread has been up for four hours and only has 6 comments, meanwhile the submission about Tesla has over 200 comments (!)
"Plan to" as in they will do it if nothing very strange happens. The cars are currently driving around here in Gothenburg with Volvo employees behind the wheels. The cars will however only work for some well tested roads, eg. the big transit roads around the city.
To be fair, the Tesla submission covers an recent rare event while this here is more of general information about things that will happen in the future. Just this is enough to cover for the amount of comments. There is a lot to discuss.
Based on what I saw at CES this year and some other automotive tech conferences and demos, I would say Google is 1-2 years ahead of everyone else with their automated driving capabilities. Not a huge lead, but they have far more data to use for training and far more cars in the wild than any OEM. Most competitors have a few cars that they field test occasionally and trot out for press events.
What's interesting to me is that the technology gap seems to be closing with the rapid advancement of sensors and embedded processors, and more standardized approaches to computer vision, machine learning, etc. making their way into automotive applications. Volvo is obviously taking advantage of this in a great way.
Depending how you look at it this isn't entirely true for a few reasons, but also not entirely wrong either. I'll do "why its not true" first but also "why you're not wrong"...
The usage of this data is relatively unknown, but based on how autopilot is improved week to week it is assumed that it is being used to learn the speed to take turns, where lanes are, etc. The provider of Tesla's autopilot hardware (partially) calls this something like "Active Mapping". Or something.
The reason you're not wrong though is that Google has more sensor-rich data. Tesla vehicles currently just don't have the same level of sensor package onboard. So Google has a lot of data Tesla can't get.
> Take the company’s latest active safety feature, an emergency steering system in the 2017 S90 that senses if the car’s about to leave the road and takes control.
I wonder what happens if I am actually stearing for the ditch?
The source says "light steering inputs if the system senses the vehicle is about to veer off the road." So it sounds like if you're full-on aiming for the ditch it won't really affect you.
I hope no car will ever unexpectedly turn on autopilot in the middle of driving. Either eliminate human control entirely, or let the human control it until they decide not to.
Do you consider automatic braking part of autopilot? Several cars already do this if they think you're going to hit something in front of you at low speed.
If you are actually pressing the throttle the car will chime several warnings but won't stop you - any car that I know of that has the automatic collision avoidance feature only brakes for you if the foot is off the throttle.
I don't think that's accurate, at least on the luxury cars where I've seen it implemented. Take a look at this list [1]. In many cases you could overpower the automatic braking by applying more throttle, but the brakes will definitely apply automatically to some degree if they think you're going to hit something.
Hmmmm yeah, I knew you could definitely do that in Nissans, and wiki offers an explanation:
"it alerts the driver using both a screen display and sound, then generates a force that pushes the accelerator pedal up" - so I guess it's like you said, you can just press the throttle harder and ignore the warning if you really know what you are doing.
This is a great point. Imagine if you're driving manually going 20 on slippery ice, and a child slips and falls off the pavement into the road in front of you. The car, for whatever reason, does not "see" the child. You decide you don't have enough stopping distance, so your response is to steer the car off the road or into another car. But then some "safety feature" decides to stop you from doing that, and hits the child instead.
The fallout would probably result in a multi-decade setback for (semi-)autonomous cars.
Volvo already has this feature in some cars, and from my testing it appears to only work on freeways where there are clear lines to show the lanes, not just any road. As well, it only seems to do adjustments, not full on overtake, so if you're gunning for the ditch I don't think this feature will help much. (Although, for obvious reasons I didn't actually try that.)
It is not the same. This is a full overtake system where the driver are supposed to not pay attention to the driving at all. What you are talking about is a driver assistant system.
Of course, the chances of such a clear-cut moral dilemma are likely worse than the chances of winning the lottery. Besides, if the ice is so slippery that you can't stop, it's likely too slippery to steer off the road.
There are very few legitimate reasons for someone to purposely steer off the road at highway speeds, but many reasons for the car to prevent it.
Well, lottery is won several times per week :) it doesn't have to affect you directly to cause a general population concern, multimillion dollar suits, and a setback in the autonomous driving industry.
Presumably, stepping on the brake or doing manual steering input above some treshold would take the car off autopilot immediately.
The opposite problem that the autopilot would have to drive the car off the road is equally hard, or even harder: if a concious driver would have made that decision (ditch the car over hitting someone) then so would the autopilot have to, or it would be vastly inferior to a human driver.
That means that the autopilot has to make that judgement and make it very well, since it will involve potentially harming the car and its passengers on purpose. Driving the car into a tree because of misidentifying something that isn't a human for a human would be very bad news.
It's the same deal with wildlife: if a moose is on the road you may have to drive off the road if you can't break, but if it's just a deer you'd likely rather hit it than drive off the road. Autopilot would have to do the same.
> but if it's just a deer you'd likely rather hit it than drive off the road.
A deer will seriously screw up your car and might kill you. Deer do kill more people than sharks in the US. In a lot of areas the ditch is a much safer play.
Yeah that was perhaps a bit locally biased: Deer around where I live are of either the Bambi or Rudolf variety and pretty harmless to cars (No risk of being tall enough to enter through the windshield) while our elks are lethally tall and either end up in the lap of the driver or flatten the top of the car.
Obviously if a deer is of a kind that approaches elk/moose size then it's in that category wrt. ditch vs. hit too.
I've only seen one moose on the road. We passed by each other on an icy morning on Hwy 2 west of Lakota, ND. He was just standing on the road on a foggy morning. It did scare the crud out of me.
We've had buffalo. One killed a motorcycle rider and annoyed the buffalo. The other I personally witnessed on a very foggy morning. Seems a very young buffalo escaped the local wildlife reserve. I was standing on the porch of the building where I worked trying to clear my head with a cup of coffee. Out of the fog and down the street ran a small buffalo pursued by a military vehicle. I think they figured they could get it into the back without it hurting them or it[1]. I walked back into the building not having cleared my head.
1) I did learn two things: 1) buffalo do not turn very well going down a hill 2) the door to a Pontiac Firebird does not hold up very well when a small buffalo hits it square after running downhill.
Yeah, its not as rare as many folks would imagine. You can steer a lot better than you can break on suddenly slippery roads. The choice of a ditch versus an accident is pretty straightforward.
I'm really impressed of what the team has achieved. I worked with Erik back in 2006-2007 and he's a very smart guy. Back then the team working on "active safety" was quite small, and the company (under Ford) was not given a lot of R&D budget, so I did not expect them reaching this point within 10 years. But many things have happened since.
Yes.