Hacker News Reader

Comment by paulmd | Hacker News Reader

The other bubble that driverless cars are going to pop is ticket revenue. There are a lot of police departments and cities who have structured a lot of their budget on heavy-handed enforcement of motorists.

That's going to decline significantly once driverless cars are scrupulously following the rules of the road. And any tickets that get written will be going against a giant corporation like Google, with detailed information about what the car was doing at the time of the supposed infraction, a lot more financial ability to fight the ticket, and a lot more incentive to protect their reputation.

View on HN

Luckily the poors won't be able to afford to the new cars, so we can continue our campaign of oppression via infraction ticketing against them for quite some time.

The cost of owning a car (maintenance, financing, registration, license, insurance, tickets, parking, cleaning) will totally eclipse the cost of Lyft/Ubering a self-driving car, even (especially?) for the poor.

We're already to the point where if your commute is less than 8 miles or so, using Uber daily is cheaper than the cost of ownership on a $30k vehicle. Just imagine what self-driving cars will do to that equation.

it's a good point, but just a reminder that for most people the price for comparison would be public transit, bicycle, or scooter, not a $30k car.

I guess that's true. I'm not saying an 8-mile commute and a $30k car is where the line is either, though. That was just my situation roughly, and Uber wasn't a little less expensive, it was a LOT less expensive. Like, roughly 60% less expensive than owning the car when you factor in things like tread wear, depreciation, and the value of your time. And that's with no lease, no loan, and minimal insurance.

The true line right now (where car ownership becomes more expensive than Uber) based on my experience is probably closer to an 8-mile commute and a $15k car. That line gets much lower when you add in things like financing, overpriced insurance, and leasing, which do affect a majority of the population.

I should actually do a writeup about this. I have access to all my numbers on it but I've only ever done rough estimates (which were clear enough that choosing Uber/Lyft was obvious). Now I'm curious to see exactly how the costs compared.

Plenty of people in the valley also commute in a brand new M4, Tesla, Porsche GT3, etc. >$70k cars used for commuting might not be the norm, but look around some BigCo parking lots....I'm sure that the average is much higher than $30k depending on the company.

> We're already to the point where if your commute is less than 8 miles or so, using Uber daily is cheaper than the cost of ownership on a $30k vehicle. Just imagine what self-driving cars will do to that equation.

Which brings up the question: why the hell would anyone buy a $30k vehicle for 16 miles of commuting?! You can get a perfectly reliable car that's even reasonably safe and comfortable for regional road trips @ well under $10k, without even looking for sales or deals.

Until true driverless (no driver in vehicle, no insurance policy needed) happens, ownership will always be cheaper than daily taxi rides.

I did. I bought a LEAF for around $30K (minus $10K in government cheese) for a commute that's about 18 miles RT.

I did it because I wanted an electric car, and specifically a 2014 or later and there were very few 2014s for sale and those that were for sale were overpriced, IMO (in some cases, more than the net price of a new one).

Our other car is indeed a sub-$10K 2005 CR-V with 180+K miles on it.

Traffic laws are notoriously regional in the United States. The fines will adapt. There some potential infractions.

Such as keeping your car in a minimum condition. Or having legacy system support. It needs to get patches to update to newer network types.

Completely unattended driving (enter destination and taking a nap) is still up for debate. Requiring a machine to be watched seems like a pretty easy-to-pass regulation.

Current traffic laws are written for human drivers. If there are robotic drivers on the road other traffic laws will be passed for those.

> Completely unattended driving (enter destination and taking a nap) is still up for debate. Requiring a machine to be watched seems like a pretty easy-to-pass regulation.

Requiring a human to watch implies a means for the human to intervene and take over. Having a human second-guess the machine will increase accidents. And having the controls available for the human to do so will increase complexity and decrease reliability.

Many regulations are of questionable validity, but do you really want to be in a machine that you have no control over? Would robotic drivers be better at things like merging, keeping distance, not over correcting? Yes. But people will still want to intervene when it doing something dumb like driving into a pond, or gets stuck on a median and cannot figure out how to get off.

Driverless cars will probably not be magic pods that whisk you around. It's still a vehicle that needs to navigate terrain. And roads are not perfect everywhere.

> do you really want to be in a machine that you have no control over

Yes. I regularly fly in planes, which I have no control over, and which spend most of their time on autopilot. I also regularly deal with machines that do their jobs far better than I could, and I feel no desire to take over and do those jobs worse. Why should that answer change when the stakes get higher and the cost of a mistake goes up?

(That doesn't mean I want to beta-test an experimental driverless car; it does mean I'm eager for a non-experimental one.)

> But people will still want to intervene when it doing something dumb like driving into a pond, or gets stuck on a median and cannot figure out how to get off.

A vehicle that did any of those things would be considered broken in a deadly way. The solution isn't to hand people broken vehicles that they sometimes need to take over from; the solution is to not call a car driverless until it handles those situations (generally by not getting into such situations in the first place).

> Driverless cars will probably not be magic pods that whisk you around.

Yet.

> Yes. I regularly fly in planes, which I have no control over, and which spend most of their time on autopilot.

Those planes with their auto-pilot system still have a means for the pilots to intervene. That's not quite the same as being in a vehicle for which no human has the possibility of intervening. (It's also a bit disingenuous to imply that it is.)

First, those pilots are heavily trained and certified, far more than car drivers. And second, far more accidents occur due to humans than due to mechanical failure (and of "mechanical failure", only a subset of those involve incorrect autopilot, though I haven't found that broken out). http://www.planecrashinfo.com/cause.htm puts total human causes at 67% (53% of which is pilot error), and mechanical failure at 20%.

Some accidents also occur specifically when pilots ignore/override the automatic systems; I'd be interested to see statistics on that, too, but I haven't found any.

Do you have statistics on the number of times that the pilot has to take control because the auto-pilot is unable to cope or is doing the wrong thing? Focusing just human-caused accidents vs. machine-caused accidents ignores this.

I don't have the stats, but I have a bit over 1000 hours as a pilot of light aircraft.

"What's it doing now?" is a common question in the industry, and the subject of this excellent American Airlines training seminar video: https://www.youtube.com/watch?v=pN41LvuSz10&t=21m41s

(The whole video is good, but that's the specific automation confusion question.)

I really hate this association of aircraft and car autopilots.

Aircraft operate in essentially empty space, and barring major mechanical failure the pilots typically have tens of seconds up to many minutes to act to avoid disaster. In a car you often have less than a second to react, and certainly never more than 4-5.

Aircraft hardware and software is tested, retested, tested again, and certified. Auto hardware and software is pushed out the door with reckless disregard for reliability.

There is really no comparison.

People forget planes, the kinds with sophisticated autopilot, don't come out with a new model revision every two to three years. Their cycles are very long, decades long.

Autolanding doesn't have the empty space, yet the first autolanding of a commercial flight with passengers aboard was done in '65.

Errr... yes it does. A runway is empty space. There are no other planes on it, it's a straight, well-maintained stretch of pavement, kept clear of ice and snow, with clear markings. Those are conditions so easy that non-self-driving cars can manage to navigate them for non-trivial periods of time without hitting anything! The Mythbusters have had a couple of cars go entirely out of control a couple times under those circumstances and the damage was a fence, once, IIRC.

There's more to landing than taxiing on the runway, like approaching the ground at a minimum of 250 km/h while getting hit by winds, often with very little visibility.

Hike dem goalposts!

They're just talking about reality of every day landings. If you don't like it, don't make up counterfactual claims in the first place.

> but do you really want to be in a machine that you have no control over?

Yes, it would almost definitely be safer than a lot of drivers I've been in the car with, if not all of them. And it's be far less nerve-wracking than with some drivers.

> do you really want to be in a machine that you have no control over?

Do you ever ride in taxis? Or as a passenger in any car? How is that situation any different?

Some traffic systems, such as red light timing, actually have been shown to increase accidents. But they also increase revenue

edit: source that I haven't read, but there are numerous sources from a simple web-search: http://www.illinoisherald.com/articles/2014/q4/chicago-red-l...

http://www.washingtonpost.com/wp-dyn/content/article/2011/02...

They also save pedestrian and motorist lives.

This is purely opinion-based, but I think the problem is primarily local/city-level. That's where most of the current problem with ticket farming happens, and once driverless cars are commercially available I think it will be relatively unpalatable to ban them at a state level or higher.

You can pass traffic laws that target driverless cars without outright banning them. The cronies at a local/city-level that pass laws to turn traffic into a source of income have friends in state and federal. Or move to the state/federal level. Is the Federal government immune from passing similar laws?

If this sort of thing becomes common enough at the local-/start-level, I can see it becoming a legal battlefield for the companies that stand to benefit from driverless cars.

And two more bubbles: long haul trucking and taxis.

Getting rid of drivers is a high priority for that industry. Human drivers are expensive and they can only drive 70 hours a week. [1]

Oh, and Uber is dying to get rid of all its drivers too. [2]

1. https://www.transportation.gov/briefing-room/new-hours-servi...

2. http://sanfrancisco.cbslocal.com/2015/01/27/how-ubers-autono...

In a sense, the damage Uber and Lyft have already done to Taxis was a boon for them. They get some extra heads up that their industry is probably going to become defunct.

AND public transport, particularly buses, but I'm sure trams/trains will eventually be automated too.

A lot of trains already are...

Maritime shipping...

Yeah, good point, ships are pretty automated. But ships really do need crews for constant maintenance and supervision. You can put the helm on autopilot but that's about all it can do.

Now, let's talk about this again when someone makes a decent R2 unit.

Shipping is already tremendously automated. The remnants are most difficult, and the environment is difficult and has some particularly bad failure modes.

There are numerous tasks shipboard which aren't readily automated, and once you have one person aboard, you tend to need more (cooks, maintenance, multiple shifts, mechanics, etc.).

The reduction in employment in shipping since the 1950s, particularly relative to cargo handled, is already immense, much of that from containerisation though.

It just means that the bulk of the revenue will come from people too poor to afford self-driving cars. Ratcheting up the injustice once again.

They'll also be able to automatically track cars speeding so maybe they'll just be the equivalent of a tax if you want to go faster.

Pay per use becomes practical with the greater tracking granularity available. Driving on rainy days, hot days, cold days, heavier cars (trucks), all increase road wear. If we all paid the actual direct cost of driving, public transit and bikes would be a lot more popular in the U.S. (maybe everywhere).

What makes you think Google will fight my ticket on my behalf? Surely, it would benefit them to have traffic laws written to clarify that the occupant is still liable for any laws his self-driving car breaks.

Also, as the other commenter said, the laws will adapt. The purpose of traffic laws is to maintain $X level of revenue, so if revenue starts to drop, local governments will increase the number and complexity of traffic laws until revenue is back to the expected level.

The end state for driverless cars is most likely not individual ownership. Large companies will probably own fleets of cars available for hail. In this world, the fleet owners would probably develop a rigorous process to cost effectively fight traffic tickets.

Even if local governments tried to increase the complexity of traffic laws, the fleet owners would increase spending on fighting tickets; this would result in an arms race which would be costly for the local government. It's much easier to extract rent from the little guys.

The worry I have is of increased enforcement of local non-traffic laws to make up revenue.

I'd expect it would be cheaper in the long run for those large companies to influence laws to move ticket liability over to their customers rather than take on the expense of fighting traffic tickets themselves.

Possibly, but that's assuming no aspect of consumer choice in the survival of ride hailing systems. If customers had to worry about tickets, they would quickly migrate to a competitor which assumed liability.

The same way that competition ensures things like Net Neutrality because consumers vote with their wallets?

Most people don't care about net neutrality. Most people will care about having to pay tickets in self-driving car fleets.

I envision a concierge service, with a low monthly fee or high one-time usage fee, for road side legal and technical counsel.

If the car is already driving, why should a a passenger know which rule the car was following at the time an officer said a moving violation occurred.

The concierge could access the car's driving records on the fly and provide specific reference to traffic laws at that specific jurisdiction. The officer and the concierge could hash out the entire interaction roadside and it could recorded.

Meanwhile, the passenger/"driver" would be sitting there mouth shut, hands on wheel, and hopefully not attracting any attention beyond the moving violation.

>local governments will increase the number and complexity of traffic laws until revenue is back to the expected level

creating a competitive advantage for least ticketed cars.

why wouldnt google just constantly update the software to comply with all applicable laws at all times? or Ford or Tesla

certainly Uber would have an interest in software that would follow all applicable laws, given the impression your robot driver getting pulled over would give to a passenger.

I expect eventually the software won't allow speeding when in autonomous modes. Why would it allow any illegal behavior? Stopping in the middle of the road at a green light? Turning right from the left hand lane? Driving without headlamps on 30 minutes after dusk, etc? These are all really rudimentary things to deal with in programming, simply disallow it. If you want to speed, you're on your own, minimal features like brake assist maybe still work. But making it possible for the car itself to break the law? I don't see it even in the medium term, and yes I'm aware the recent Tesla crash was speeding while Autopilot was on.

Driverless cars are excellent witnesses of the behavior of other drivers. Speed, initiate an illegal lane change, roll through a stop sign, etc, and it wouldn't be a technological stretch for a driverless car to send a report to the police on your behavior. Police could even pay the owners of driverless cars that report bad behavior a reward, if one of their reports results in a ticket.

Local authorities will have to recoup costs via registration fees or maybe usage tax (miles driven times some decimal times an income bracket). Unless GM, Ford, Tesla, Google are going to maintain roads and pay for policing --although FB and co. could assess who have more perp risk and assess taxes accordingly...

"The car was obeying the rules of the road too well. It was suspicious."

"Your honor, witness testimony is notoriously inaccurate. Will you believe the officer? Or the cryptographically signed and verified vehicle logs?"

Reply on news.ycombinator.com