I agree with your overall vision.

But there's a lot being swept under the rug in saying, "These are (usually pretty minor) technical challenges, and I haven't heard one yet that we won't be able to overcome with today's technology." It comes close to a "sufficiently advanced compiler" or "just an implementation question."

Autocars are going to be held to a much higher standard than human drivers. This sucks, is irrational, and is unfair, but thems the breaks. Given that media and social environment, a single fatal car crash will cause a firestorm that could delay widespread introduction of driverless cars for years. People are terrified of new technology: see the shit show about a Tesla catching on fire, or the people who think people wearing Google Glass deserve to get assaulted.

We absolutely have to tread carefully here.

I suspect the higher standard question will be resolved in a couple of ways:

* NTSB review: Airplanes face higher scrutiny than automobiles (despite the lower overall casualty rate). Part of the reason people have grown comfortable with air travel nonetheless is because the NTSB analyzes what went wrong after each crash and recommends (mandates) certain changes. Moreover, as far as government agencies go, the NTSB does its job reasonably competently and independently.

* Gradual introduction: It's unlikely fully self-driving cars will be introduced all at once. What's more likely is the gradual automation of certain driving functions (sort of like cruise control, automatic parking, and other feature sets). I think any system that hands off control from a robot to a human introduces its own safety risks, but you might see something like a dedicated autonomous lane on a freeway with special parking areas where humans can safely resume control for driving in the city. There's also augmented driving where a car is primarily human driven, but will do things like pre-charge brakes and give advanced warning to a human diver when it believes the human is about to screw up.

* Economies of scale: Large business organizations will probably get comfortable more quickly with autonomous cars than individual consumers. If FedEx and UPS support policies like "no left turns," they'll probably be among the first to get on board with fully automated trucks. And you'll certainly see some support for the insurance business, if autonomous vehicles really do result in lower overall accident rates.

UPS employees are unionized, and driving routes (vs. working in the warehouses) is a sought-after position that is one of the benefits of seniority (as I understand it). I expect that they might have issues rolling out a computerized delivery fleet.

Fedex drivers aren't. That doesn't matter as much when there aren't real alternatives, but it can become critical if technology starts to drastically reduce the need for labor. That's part of what happened to the American auto industry during the 80's.

So they'll just get replaced. Amazon and Google do seem keen to get into that space.

Also why the airlines go bankrupt approximately once per career. The gross revenue per flight is the same as time increases, but pay raises for seniority decrease the profit as the workforce ages.

(Not making a value judgement here. Airline pilots have a tough job and do it well, and deserve to be well-paid. But the industry is not as nice as me.)

It's not like they're forced to pay their workers more as they age. The ones with unions may be forced to do so by their contract, but non-union airlines are hardly immune from bankruptcy. Airlines go bankrupt because it's a cut-throat business with extreme competition, which drives profits to zero.

An airline customer will switch carriers for as little as a $5 savings on the cost of a multi-hundred-dollar ticket. There is essentially zero customer loyalty. If you can't convince customers to pay more, and all possible cost-cutting measures have already been implemented, then you have no room.

In an ideal world, a business could survive forever in an environment like that. In the real world, business is variable, and with no margin, you can't build any resiliency. A couple of bad years is enough to screw you.

I hope they replace the workers in the warehouses first; since UPS/FedEx/Amazon own them, they could put QR codes on the floor or other craziness to automate those workers first.

There must be at least 50 million residential garages in the US alone. If we automate all cars and switch to a Cars-as-a-Service model that is 50 million large rooms suddenly available. That's a lot of reclaimed real estate.

Edit: If the average garage is 250 sq ft (pretty small), average cost of real estate is $100/sq ft (pretty low), then this is a $1.25 trillion windfall. Not bad for a small side-effect of autonomous cars.

Well, that's if you assume that Cars-as-a-Service don't need to be stored anywhere and magical disappear when they are not in use.

I know Google has magical fairy powers, but last I checked, a self-driving car is exactly the same size as a regular car, and takes the same storage space.

Part of the storage space requirements for cars in normal use is access space for human drivers and to allow random access entrance and exits. Cars as a service can eliminate much of the storage space requirement.

As well as 'redundant' storage, since each car effectively gets multiple parking spots at the moment (a dedicated garage, plus a guaranteed spot at work, etc). Once cars are automated you only need enough storage spots for the cars not on the road, which is much smaller.

Also, cars could park on big parking lots, which don't necessarily have to be near someone's home.

Just think of how little time your car is actually driving versus being parked.

Not everyone is driving at the same time, so Google can use their "magical fairy powers" (also known as statistics / data analysis) to determine how many cars are needed to support a city, maximizing the number of them that are constantly in motion, dropping one passenger and picking another. The ones that are not needed at the time can be stacked tightly in a dedicated multi-storey parking garage.

I think it's pretty unlikely that we're going to see most of the market for family vehicles be turned into a short-term rental model. Some, perhaps! Cars-as-a-service will have some advantages over, say, ZipCar today (primarily convenience, some amount of cost as well).

But owning a self-driving car sounds pretty sweet as well. Not only will you be able to get the status symbol car you want, you'll also be able to leave your stuff in it (without paying extra to rent it for times you aren't driving). And it's better than the cars of today! You don't need to worry about parking! You can drive when drunk or tired! You never need to worry about your cars-as-a-service being unable to send you a car during peak times!

Cars are popular to own today. Self-driving cars would be more valuable to own than normal cars are today. You have to imagine immense benefits to cars-as-a-service for it to displace most of the market for self-owned cars.

That $100 is only low for highly developed wealthy areas.

Square feet of livable housing is cheaper than that in most of the country.

Undeveloped land probably mostly sells for less than $0.10 a square foot (that's still high at $4000 an acre). But that's partly because large sales of relatively isolated land will dominate the acreage transferred.

Way to only see everything with rose-colored glasses.

Has it even occurred to you there might be any possible cons to your scenario?

First of all, the lot the cars come from is going to require real estate, so subtract that from your windfall. Where will they be at night or other low-use times?

Second, subtract the additional resources in gas/electricity it takes to get to my house from god-knows-where it came from.

Third, subtract the time I'm waiting in my driveway with the kids screaming and my wife in labor for a car to show up from god-knows-where. Or else the premium I'm paying to get ahead of the other guys in the queue. And while we're at it, all of those times that I wanted something from the store 5 minutes before it closes? Guess what, if I can't get a car in time, that's lost productivity. Have to buy lunch from the cafeteria b/c I couldn't get to the supermarket. Unless you're telling me that in order to accept this vision of the future, I have to be perfect and plan everything. Telling me in the future no one will be able to get to the store 5 minutes before it closes is just unrealistic, undesirable and unacceptable.

Fourth, subtract again, the additional energy needed to take the car to its next destination.

Fifth, consider all of the extra cost of adding, servicing, verifying, maintaining LIDAR and servoes to all of these millions of cars. With thousands of dollars of extra parts, they're going to require thousands in extra maintenance, inspection, training and insurance. Yeah, I know that's boring. It's not an interesting technical problem, but that's the reality of motor vehicles. If anyone thinks this won't make the price of vehicles go up significantly, they've never been in charge of anything critical.

I'm not "against the future," it's just that all of these digit-heads are seeing a grandiose signal-processing problem and doing multiplicative math like yours to make it look attractive and not even operating at a practical level whatsoever.

So, choose to have a dedicated vehicle then.

But don't be surprised when the majority of people refuse to pay 100% of the price for a vehicle they only use 3% of the time.

Interestingly, people pay 100% the price of a microwave oven, which is usually used just a few minutes per day.

Going that direction ... Did you know? A typical microwave oven uses massively more electricity simply powering the LED clock compared to the amount it uses actually cooking food.

[citation needed]. I can't find any information to verify this.

Apparently a Microwave typically uses about 1.5W on standby, so that's: 24 * 3600 * 1.5 = 129600 Joules per day

So, if it draws about a kW in use, that means that if you use it for less than 130 seconds per day, or just under 2 minutes, you'll actually use more power on standby than on cooking / heating.

I think I use my microwave more than 2 mins each day, but there are days I don't use it at all, so I'm not sure how it would even out.

1.5W is a crazy amount of standby power. Not that I don't believe older models may use that much, but there are now regulations and standards for that kind of thing which push that to a much lower power.

That's just the figure I found from a quick google. If you have a more up-to-date figure, feel free to share :)

You sound like you live in the suburbs. I suspect car-as-a-service wouldn't be popular there. Honestly, if you live close enough to the store that you can drive there in under 5 minutes, to me that's a sign that you don't need to drive there.

Where I live (Berkeley), and in the Sf Bay Area in general, car as a service would be remarkably effective and popular.

do people realize that all these space savings mean that builders will now make smaller houses and more of them?

That already happens in mega cities where parking spots are around $500k. I take taxis all the time so I'm already living in the future, the western developed world will catch up eventually.

The real estate requirement with autonomous vehicles is only the difference between the off-peak maximum required parks and the peak (no cars will be parked). Currently, the real estate required for non-autonomous vehicles is typically more than two spots for each car. (one at home, one at the office, and a fraction of the public parking)

Of course I'm way late to the game here, but thought I'd add one more voice of dissent:

The real estate costs can be extremely low for self-driving cars. Imagine a tower of, say, 10 cars stacked vertically with an elevator mechanism to bring them to and from ground-level. Since each autonomous car is interchangeable, you can just treat the vertical parking lot as a stack. When demand is high, pop off the stack. When demand is low, push back onto the stack. No need for ramps or anything else that existing parking garages use. And just to be totally clear, humans don't interact with this tower, it's just a storage mechanism for empty cars after they've unloaded passengers.

If demand is low traffic should be low too, why not just use empty lanes of streets for the same purpose? Or at least, some variation on that theme. Maybe things work out so well that streets usually only have 1 lane.

Theres already an abundance of parking real estate, there's no need to make more. Look into the predictive analytics used by Uber or Citibike to see how wait times can be minimized. When freed from the redundancy of having to use an overbuilt one-size-fits all for everything, the energy savings will more than make up for a car having to drive two or three blocks from the parking lot to the pick up zone. Shit, I could go on.

Maybe the store will drive to you ;)

If the average self-driving car-as-a-service costs $100,000, then replacing 50 million old cars with 25 million new cars is a $2.5 trillion dollar expense to get your "windfall".

Paying for the self-driving cars will likely be a cost savings by itself, since we won't need as many auto-cars as regular cars. Most cars are only used maybe 5-8% of the time, and sit unused otherwise. An auto-car service could probably get at least 50% usage out of a car, potentially much more. That means a 10x reduction in the number of cards needed.

For this to be true, we have to assume that, during rush hour, at least 90% of all cars are parked and unused. 90%? That's a hell of a lot, dude.

It's certainly true that most cars spend most of their time being parked and unused. But they tend to do so during times of day when there is very low demand. The vast majority of cars are unused at 4am (and, indeed, at 2pm), certainly. But even if every car that was on the roads at 4am was shared, that doesn't increase utilization of the car-pool very much. And you have to be able to handle the routine car-use spikes around rush hour.

Surge pricing equivalents will motivate people to adjust/stagger their working hours. I already arrive early and leave late to avoid peak hour.

Pod-based transport would also make public transport easier and closer to a door-to-door experience.

Or maybe surge pricing equivalents will motivate people to buy their own cars. If the advantage of a car-sharing scheme is cost savings, and then you start charging more, what's the advantage again?

In any case, I don't think that that's a feasible option for the majority of jobs.

Charging more than the cheaper default. Doubt it will be as much as the depreciation, registration, insurance and servicing of a car.

Right now, my wife and I have two cars. One of them is only used 2-3 days of the week. The other is only used for a small portion of each day.

If taxis weren't needing to cover wage costs, I'd already be better off switching from the second car to using a taxi to get to and from work. Registration, insurance and servicing for the second car is $2-3k/year.

I don't think that insurance is a good cost to include in the comparison. In a driverless car world, we imagine that insurance costs will be pretty much a constant per mile traveled -- there are no better or worse drivers -- and probably lower than they are now. And you'll have to pay for them (directly or indirectly) whether you rent or own.

If your costs for the second car are more than $1k in just registration and servicing are more than about $500 a year, that, first, suggests that you're not very cost-conscious today. My car's cost service and registration fees are much less than that. Second, service costs also are basically a per-mile fee -- your car's quality doesn't degrade when it sits in your garage, and you'll pay the service costs of a vehicle you use directly or otherwise.

You didn't mention fuel, but of course again that's essentially per-mile, and costs what it costs regardless of the car ownership model.

That leaves registration, which is a small cost, but would be shared in a shared ownership model.

Now, of course, the actual major expense is the cost of buying the car in the first place, defrayed over the lifetime of the car, minus any price you get for selling it. The big advantage of a car-share service would be that you could share that cost with others. If a driverless car costs $50,000 (and we have no idea how much they would cost), and you use it for only 5 years, then sell it for $20,000, you get $6k per year. In a shared ownership model where you share it with three other people, and the company that you're going through makes about 10% gross margin, you're looking at $2.2k instead of $6k, so yay, savings of $3.8k per year. That's a lot of money.

If, on the other hand, a driverless car costs $30k and can be sold after five years for $15k, then it's $3k per year to own and $1.1k per year to rent, you're saving $1.9k instead $3.8k, and that's still not peanuts, but you suggest that right now you spend more than that for a not-very-highly-used second car. The advantages of owning are considerable.

Great point, was not considering that. I'd love to see some stats on maximum car utilization as a percentage during each minute of rush hour.

You could imagine that in addition to single-serve autocab services, you could have auto "micro" buses, sitting 4 people. Your "bus" is guaranteed to stop at no more than 7 other stops. With tens of thousands of people using the service it should be possible to pick optimal batches of people for each bus, to minimize how far the bus has to deviate from an ideal path for any given rider. It would be a giant logistical problem, but I think theoretically you can get a service working that was almost as seamless as an autocab service. Would be interesting to work on.

Yeah, but how many of those garages are packed full of stuff, with the car parked out in the driveway, anyway?

If you own your garage, I'll bet you'll fill it up pretty quickly. But people who pay extra rent for garage space will reduce their use. So there will be some extra space, if not 50 million.

It's funny how none of the critics think about the human edge-cases. What happens when someone decides to drink and drive? What happens when someone falls asleep or gets distracted on the freeway?

The answer, of course, is that edge cases are just that. They don't impede general use, and though they occasionally cause some worst-case outcomes to occur, we all just decide to live with them.

I. You shift the goalposts a lot. E.g. in the same sentence you mix statements about the future and the present, e.g. "we will" and "today's technology".

> I haven't heard one yet that we won't be able to overcome with today's technology.

BTW, LIDAR doesn't work in snow and rain. Today. It's not clear that it works for multiple independent vehicles, either. We may only have self-driving car, singular.

II. Also, in your scenario, even if you do send that car over to your wife to use once you get to work, the same car will get more wear and tear than two individual cars b/c it's doing the same work plus more to go between you two.

III. You also mix up the moment when self-driving cars become as safe as humans with the moment they become safER than humans.

IV. You also overly dichotomize safer vs. not safer. What if they kill fewer auto drivers, but more pedestrians? What if they kill fewer adults and more childen? It's not that simple- our values are reflected in our choices to deploy technology.

> It's not clear that it works for multiple independent vehicles, either. We may only have self-driving car, singular.

I'm no LIDAR expert, but I do design and build synchronous detectors (think: boxcar/lock-in amplifier/gated counter). Maybe this has been tried and there's some reason it wouldn't work, but I don't see why you couldn't apply a pseudo-random modulation to the LIDAR emitter and then run the received signal through a correlator. Besides bringing some conversion gain and allowing for more accurate distance determination (compared to a simple pulsed-ToF), PRN modulation allows you to share the channel with other transmitters (CDMA, if you will).

> BTW, LIDAR doesn't work in snow and rain. Today.

Isn't this just a matter of the wavelength used by today's LIDARs? There's a huge hole in the liquid and solid water absorption spectrum in the visible range.

I don't know the system, but I'm sure it's already synchronous in that there are numerous receiver channels and their acquisition values are correlated with the illumination angle.

If the system has to deallocate time channels for other vehicles to use, it loses bandwidth, proportional to the number of vehicles in the vicinity.

Sharing the data between vehicles is possible, however, it invokes geographical/networking problems, such as how to partition the acquisition, how to synchronize the acquisition and how to share it, including the necessary bandwidth, which I'm sure is quite high.

As soon as multiple vehicles sharing data enters the picture, security must also get factored in. Besides the link control concerns, the system will be sensitive to jamming with that much multiplexed sampling taking place.

PP is correct in hinting that the leap to multiple self-driving vehicles from one self-driving vehicle will be large indeed.

> I don't know the system, but I'm sure it's already synchronous in that there are numerous receiver channels and their acquisition values are correlated with the illumination angle.

This is not "synchronous detection" in the way I meant it. I don't mean this as a put-down, but it may be instructive to Google "lock-in amplifier," a term I mentioned in my previous post.

All of your other concerns are addressed by appropriate choice of PRN code(s). Additional vehicles, which operate without the need for coordination, merely raise the noise floor. They do not "jam" each other. It should be obvious that a minimum S/N ratio is required for the LIDAR system to work and further that an arbitrarily-high number of LIDAR transmitters therefore cannot coexist. However, it is far from obvious (to me, at least without learning more about LIDAR and plugging in some numbers to a model) that a congested freeway of driverless cars would have "too many" LIDAR transmitters.

No, snow scatters the emitted light and scarcely any is returned at all. A wall covered in ice looks like an air gap.

(I know this because my very expensive 800kg autonomous vehicle almost drove into a lake under my software control due to this effect...)

Aaaah, a domain-knowledge expert, thaaaank god!

Actual information on this topic is sorely lacking....

....While we have you on the line, would you please contribute some more information on LIDAR w/r/t autonomous vehicles?

For example, how well does LIDAR work with rain? fog? How does LIDAR interact with other nearby LIDARs? How much power does a typical system emit? How sensitive is it to jamming? How many frames per second can be acquired? How much computing power is used to re-assemble a scene?

Generally LIDAR "works" reasonably in light rain because the rain drops scatter most of the emitted beams, and you get no returns. Occasionally you'll hit a raindrop straight on and get a reflection back to the receiver, but your algorithms should probably filter this out.

Fog is so much denser that you get heaps of reflected returns, and naive algorithms would treat this as there being lots of solid stuff in the environment. Your best bet is to combine sensors that don't share the same EM bands; e.g. lidar + camera, or + radar, sonar, etc. It's not just redundancy, but the ability to perceive across different frequencies so that things that scatter or absorb in one band don't do so in another, allowing you to distinguish.

I don't think one LIDAR would interact with another to any great extent. Even if off-the-shelf models did, there'd ultimately be some way to uniquely identify or polarise the beams such that this wasn't a problem. I suspect its reasonably easy to engineer a solution around this.

Power emission I'm not entirely sure about, but almost all are at least class 2 laser devices. You shouldn't point an SLR camera at Google's vehicles for example, as you can destroy the CCD. The Velodyne they use draws 4-6amps at 12V, but a lot of that power goes to heating the emitter, motors, etc, and isn't all emitted by any means.

Frames-per-second isn't really the right measure for Velodynes, but their max rotation is 3-4 revs per second IIRC. It's about a million points per second. (For something like the Kinect For Windows V2, which is a flash lidar, it should run at 30fps, but with lower depth resolution.)

I can't think how you could 'jam' a lidar, but you can certainly confuse the crap out of it easily enough. Scatter some corner reflectors on the roads for example, dust grenades, fog cannons, etc.

Computing power to reconstruct is significant (many DARPA Grand Challenge teams had problems containing their power budget for CPUs/GPUs) but manageable. It depends on the algorithms used, and in many cases the amount of "history" you infer over. Google's approach is to log everything, post-process into static world maps, then upload those maps back to the vehicles. When they're actually driving for real, those maps effectively let them take the delta between what they currently see and what the static map says there should be, and they only really have to handle the differences (i.e. people, cars, bikes, etc). This is still hard, but it's much easier than e.g. the Mars Rover problem (more my area of experience)

There's the simple issue of legal liability, though. If you crash into someone, it's your fault and you can be taken to court for it. If a Google driverless car crashes into someone then it's Google's fault. That's an extremely large number of potential crashes and legal challenges.

Absent any higher concern, Google has to be very careful about this out of their own self-interest.

I hear people fretting about liability as a "problem to overcome," but it's actually much simpler than the technical challenges.

People have to drive the same distance regardless. Whether it's human or computer error leading to the accident is irrelevant to your premium. To a car insurance company it's all about rates and risks.

Assuming driverless cars are shown to be safer than humans - and they'll have to be in order to get approval - your car insurance company would be stupid not to cover them. From their perspective it's simple economics. The accident rate is lower, thus the potential liability is lower.

If your insurer isn't willing to insure your driverless car, you'll switch to someone who does. If no established provider insures driverless, that's a no-brainer company to start. Similar payments in a high-margin industry with lower average payouts? Yes, please.

Liability is the easy part.

The problem is your insurance company. If you get in an accident, they're not going to want to pay for it, and they're going to go after Google for the damages. Now we're right back where we started.

No. Current driver-based insurance policies might be based on the ability to hold the driver personally responsible if they are negligent.

Insuring a vehicle is just figuring out what the likely payout ratio across a cohort of those vehicles is going to be, then working backwards from there to come up with a premium that will cover the claim amounts, administrative costs and provide a pool of funds to invest.

Lots of things are insured today with no interest in who is operating them. Automated cars are no different to that.

Essentially you're imaging a current situation (where a driver is insured) and transferring that to a situation with no driver. That's wrong.

Insurance companies don't care - they just want to be able to predict what the likely payout rate is. And that's how much it will cost. If that includes cover for buggy code, it just means a higher premium. Essentially younger drivers have buggier driving code, and that's why they cost more to insure.

Also, in terms of regulatory capture, it's almost certain that the legal requirement to hold insurance for your car will continue. But the combination of the fact that payouts will probably move towards class action against software bugs (where class action suits usually have a lower per-person payout, and will be paid out by massive policies held by the car companies) and the fact that in general fewer of those people will ever need a payout thanks to safety improvements (especially as we move into most cars on city roads being driverless) will mean insurance companies will have to pay out less.

I will never understand why people think insurance companies will be a barrier to adoption. Insurance companies are going to love this. I think the tipping point will be when insurance rates on driving your own car skyrocket.

>I will never understand why people think insurance companies will be a barrier to adoption. Insurance companies are going to love this. I think the tipping point will be when insurance rates on driving your own car skyrocket.

Yes. The logic goes like this. 'Currently I'm an insured driver'. I will have a car with no driver. It therefore cannot be insured.

Where I live cars are already insured for third-party personal damages, regardless of who is driving them. It's a fixed-cost premium attached and paid at vehicle registration time. Thus, if you are injured in a motor vehicle accident, regardless of who is at fault, you (or your heirs) will be paid compensation in a regulated environment. This is a functioning market which doesn't adjust for driver age or experience, but instead collectively insures all vehicles on the road, and pays out when a claim is made. Periodically, when more than one insurer is involved, they will negotiate an 'at fault' percentage in order to work out which ratio at which the claims are attributed to each insurer. This may involve discussion of the drivers actions (to determine fault), but the payout is always made - because the vehicle itself is the thing that is insured.

In this system, the only way the driver can become liable for the claim is if they specifically and deliberately engaged in a criminal act, such as driving an unregistered vehicle, or driving under influence. Mere speeding or red-light running doesn't void the insurance (as these are traffic infringements rather than criminal acts). And in these cases, the insurance company still pays out the compensation, but may proceed to recover the compensation from the driver that broke the law.

This type of system would easily be translated to driverless vehicles. The owner of the vehicle would pay a registration fee, and a portion of that fee would be passed onto the owners choice of insurer (or a valid certificate of insurance would need to be presented for registration, which is the same thing). As long as the insurance was paid, then third-party personal and property damage would be paid out to any accident claimants. If the insurance company then decided someone was directly liable for causing the claim (such as tampering with vehicle systems or negligent coding) then they could recover the claim money.

Such a system would be a vast improvement on the current situation, whereby it's the driver that is insured, and because of the various risk factors, insurance premiums are all over the place. As a result, the riskiest drivers also tend to be the ones lacking insurance, which is the worst case for someone who needs to make a claim.

But then insurance is one of the most misunderstood products around, so I guess it's not surprising that ignorance of how it works abounds.

I don't understand why you think this. Yes, for small incidents, it won't be an issue, but if there's ever a claim over $X (where X is a figure that outweighs the cost of the time to hire lawyers to seek compensatory damages for that claim), you can bet they'll go right after the maker.

Why couldn't Google just offer their own insurance plan? They could bundle the cost with a monthly or yearly service/maintenance fee. If the margins are so high on insurance products, I'm sure they would be more than happy to keep the profits themselves.

It depends on how they decide to hedge their own risk.

  Buy from an insurance agency = safe but more expensive.

  Self-insure = cheaper but riskier.

But what would they insure against? Individual accidents or class actions?

Is this actually a real problem? There is a robust insurance environment around cars, with a huge amount of both pricing experience, case law, and experience with payouts for people injured in accidents. There is also plenty of experience with recalls and fatal manufacturing flaws.

Is there any reason to believe that insurance companies won't just happily write policies priced for automated cars, and everything will proceed pretty much like it does now? You drive around, accidents happen, insurance companies pay out to the injured parties, and adjust the rates on automatic cars to match the real cost to operate?

Legal liability seems to be a red herring IMHO. At the end of the day, it's an insurance company that pays for it all anyway, regardless of whose fault it is. And car insurance companies already have the capability to adjudicate and litigate a large number of crash-related legal challenges. If anything, Google could use the data collected by its autonomous vehicles to streamline that process.

If I were Google, I would just make some kind of deal with an insurance company to handle all scenarios where a crash occurs when a vehicle is in autonomous mode and bundle said insurance in with the cost of the vehicle.

Exactly. Who's at fault in a self-driving car crash? The car model+software combination (unless the driver explicitly did something negligent or reckless). But the insurance company will cover it, and simply adjust the cost of insurance each year based on how often that model+software combination crashed in the previous year.

Should you step in front of a Googlecar and get hurt, remember before you sue that it bristles with sensors and recording devices, and that it'll all be admissible in court.

Is there precedent in courts for how that kind of data is handled?

Liability should be pretty simple. Just like now, the owner of the car is liable. In my state this is the case even if someone else is driving; not sure this is the case elsewhere but that would seem to apply to a computer driver as well as a human one. The logic is, if you allow someone else to use your property and they cause damage with it, you are liable.

Insurance (assuming the owner has it in force) may pay, but the liability still rests with the owner.

I'm thinking there could be an agreement you sign when the car is purchased that states that the self-driving abilities are not to be fully relied on and the driver must be fully aware and ready to take over at all times. Thus if the car ever crashed by fault of Google they can say the driver should have been paying attention. This would have a lot of downsides. For instance, DWI's would still be a thing.

Getting humans to participate only in emergencies is almost the worst of all approaches. With reduced practice the human is deskilled and they are less likely to be alert and aware of the situation if they have to take over suddenly.

Signing an agreement will not make people alert and ready.

Even ignoring the issue of the driver's skill reducing as they practice less, switching from the thing with microsecond-range reaction times to the thing with second-range reaction time when an emergency has already started will almost never result in an optimal outcome.

This is not simple issue at all. Who's at fault if anti-lock brakes don't work as expected? What about cruise control? The new cruise control that auto-follows? What about lane-following technology?

We are not going to jump directly to self-driving cars. There will be a dozen steps along the way. When does liability shift from the driver to something else? This is an interesting question for sure.

> We are not going to jump directly to self-driving cars. There will be a dozen steps along the way. When does liability shift from the driver to something else? This is an interesting question for sure.

Well "shift" implies that it must be one place or the other; in reality, liability is not zero-sum, so one party having reduced liability doesn't imply another party has more, and vice versa.

This is particularly the caseWe are not going to jump directly to self-driving cars. There will be a dozen steps along the way. When does liability shift from the driver to something else? This is an interesting question for sure.

In the case of automotive systems, there's no reason the driver, owner, and manufacturer all can't be liable, and no reason that any changes in the degree of liability for one have to be reflected at all in the liabilities of the others.

I think this will eventually be resolved the way vaccines were with liability shifted to government. There's just too much societal good to come out of self driving cars for liability to be the deal breaker.

"a full network of communicating cars and fail-safes" is a long way away, though. And to get there you have to go through a long period of some self-driving cars, sharing the road with old-fashioned human piloted ones. As a human driver, I spend a lot of my time anticipating the behavior of my fellow road-users. That car coming up from the side street - did they see me? They're on their phone, maybe they didn't see me... they're going to pull out... better brake...; now I'm coming up to an exit; the car in the lane to my left just changed speed, probably going to try to cut across in front of me, better give them some room...; cyclist coming up on the right, am I going to get past him before I need to slow for my right turn, or should I slow up behind him?

Understanding and anticipating all those human behaviors seems to me to be well beyond the capabilities of current AI.

I don't know, it seems to me that a computer is particularly suited to this problem. I think a computer would be much more accurate in judging the speed of the approaching bike or the car coming up the side street or in the left lane. The essentially negligible reaction time compared with a human would give significant power to a defensive driving algorithm. And I'd guess probabalistic models are a better predictor of behavior than most people's intuition.

Self driving cars see 360 around themselves at all times. Real world data is used to train self tuning algorithms that drive the cars. These cars see everything and they know everything about how irrationally other human drivers are, and they don't take any chances. They drive slowly when they identify a danger, and they react almost immediately when there is a surprise, compared to the relatively sluggish reactions of humans. And humans make mistakes, lapses in judgement, are tired, and take stupid chances all the time. Self driving cars drive perfectly all the time. Sure they can't do much if someone deliberately drives into them, but it won't be their fault.

How about making eye contact? When I'm driving, eye contact & facial analysis ("did that person see me?") is significant in my risk assessment of other vehicles. That is, if I know that the other driver knows, then I don't have to worry as much. I will still be careful, but I don't have to be quite as wary.

It sounds like the equivalent of eye contact for automated traffic is the "full network of communicating cars", but there must be a middle ground with mixed traffic, where automated cars are extra-careful about manual vehicles. Will that be annoying or reassuring to other drivers?

There's always a more cautious option in each of those scenarios. If the self-driving car always errs towards that, hopefully it'll mitigate disaster.

Following good driving principles to the letter - accounting for stopping distances etc, like in the Highway Code (UK) - will hopefully allow time for those unpredictable road users. Humans are more risky in that we give ourselves less time to react to begin with.

Some of the exact things you mention are covered in the video in the post. I think you are underestimating current AI.

> We won't need to waste 4-10%+ of our entire lives staring at the road doing nothing.

No you don't. That's why many people use public transport or combine travelling with exercise (bike). Self driving cars doesn't solve the real problem which is that people have the idea that cars are actually sustainable.

Self-driving Teslas charged with nuclear-generated electricity seems very sustainable for me. Make them rented on-demand instead of owned and you have a PRT system, combining the convenience of a car with efficiency of public transport.

You must not live any where in the entire center of the United States

Parking structures might become high-security, automated bldgs that people don't enter. Further, local parking spaces may be unnecessary as your car (or the car you rent) will pick you up.

I don't understand why we're so hung up on trying to proactively create legislation around this. Why don't we let someone get maimed/killed first, and then go from there?

In the world of things, you can't foresee every possible outcome, and it's silly to draw up legislation, which has a nasty habit of becoming rather sticky, for something so potentially earth shattering.

Let it fail, turn it off remotely, and then we'll figure out who to blame and what to do next. For now, this tech is so beyond us, we don't even have a frame of reference for how to properly legislate it.