Monday, February 23, 2015

Will driverless cars cost us more than they’re worth?

February 20, 2015

Dear Cecil:

I've been waiting for autonomous cars to become a reality. But I’m wondering how much revenue will be lost when there are no more speeding tickets, traffic violations, parking violations, or probably quite a few other kinds of fines that I haven't even considered. How much money will state and local governments lose when traffic tickets become a thing of the past?

— Jeff Grippe, White Plains, New York

Cecil replies:

What I’m wondering is why you’re even thinking about this. Driverless cars may well reshape the urban world — for one thing, autonomous car-sharing could wipe out taxis, limos, and Ubers in a single swat — and you’re focusing on parking tickets. Could you possibly have picked a more boring aspect of this development to analyze? But since you asked, yes: this particularly irritating form of revenue extraction would be mostly eliminated in the event cars became autonomous.

As it stands, issuing tickets is something governments do a lot of. New York City gave out more than a million in 2012. Roughly 23 percent were for tinted windows or seat belt violations (conjuring a rather unsavory image of what New Yorkers are doing in their cars), but the rest were for infractions that wouldn’t’t exist if cars were automated: speeding, phone use while driving, etc. Financially, this is an incredible boon for states and municipalities — the NYPD’s recent hissy-fit strike against Mayor de Blasio cost the city $10 million a week in parking-ticket money. It’s hard to find an ironclad nationwide total for ticket-fine revenue, but (for example) Virginia raked in roughly $97 million on speeding tickets in 2010; scale that up to a population of 320 million and you get a national figure of about $3.7 billion. If autonomous cars make that sum just go away, budget committees are likely to notice.

Google, the apparent front runner in the race to driverless cars, claims their increased efficiency (in part because they can travel in a tightly spaced convoy, reducing drag) will ultimately cut commuting waste by 90 percent. Forbes works the annual savings out as 1.9 billion gallons of gas and 4.8 billion commuting hours, for a total value of $101 billion. I’m not sure I completely buy the details where Google is concerned — two of the last five times I trusted Google Maps I spent a lot longer in Indiana than I’d intended — but undoubtedly the government would lose some money here too. The current federal tax on gasoline is 18.4 cents a gallon, and the average state tax is 23.5 cents a gallon, so 1.9 billion gallons saved means a $350 million annual loss in federal tax revenue and a loss of $447 million for the states.

On the other hand, the total yearly economic cost of all U.S. motor vehicle accidents dwarfs both these figures — in 2010 it was $277 billion. Driverless cars would probably have the occasional accident as well, but the most dangerous factors could be eliminated — crucially, drunk driving. Of the roughly 33,000 traffic fatalities each year in the U.S., about 10,000 result from alcohol impairment. On a pure dollars-and-cents level, that’s a total loss of something like $19 billion in future earnings that the government won’t get to collect taxes on.

Additionally, the government savings on public transportation would be huge. The Chicago Transit Authority system gets about $700 million in annual public subsidies; much of this could be eliminated if bus service, which runs up major labor costs, were replaced by privately operated fleets of driverless minivans. Other pluses on the balance sheet: the disabled and elderly would have greater taxable earnings potential because transportation would be easier, and fewer Medicaid and Medicare dollars would be spent on those involved in car accidents.

The journey from a Google engineer’s wet dream to reality is a long one, of course, and we’ve still got a ways to go before any of these considerations becomes relevant. Some don’t think we’ll ever get there: Elon Musk, founder of Tesla, is working on a driverless car too, but he doesn’t think the human element can be totally eliminated — his version would be more of an autopilot feature. Google’s autonomous car has covered 700,000 miles without incident in and around the Bay Area, but the programmers have fed it tons of data specific to local roads — it wouldn’t work if you dropped it in the middle of Tokyo. Google has preempted one obvious objection by saying it should be liable for any tickets its cars incur, but plenty of unresolved questions remain: How will the car choose in a no-win situation — say, when it has to hit either a jaywalker or another car? Is there a cheat code to get the car to drive faster? Or can you trick the software into speeding by telling it your wife's in labor?

But if you’re asking whether driverless cars are, on balance, actually worth pursuing, the answer is: duh. It’ll surely take a while for it all to play out, but if this thing winds up being half the big deal it could be, the change in traffic-ticket revenue is going to look like a rounding error.

— Cecil Adams

[The figures are for the US. For Singapore, if Autonomous Cars/Driverless Cars (AC/DC) become widely adopted and replaces the current flawed model (the one with a human driver), these will also be the loss and savings. The better safety will mean less loss of life and loss due to injury, which will be an improvement. 

The other questions Cecil suggests are intriguing - whether to hit a jaywalker or another car - this can be subject to Asimov's 3 laws of robotics. The problem is whether the computer can compute fast enough to decide.

See this: http://www.theatlantic.com/technology/archive/2013/10/the-ethics-of-autonomous-cars/280360/
And this: http://cacm.acm.org/magazines/2015/8/189836-the-moral-challenges-of-driverless-cars/fulltext
and just google "Moral Reasoning Driverless Cars"


More interesting would be communicating cars. That is, the cars are not just individually autonomous. They are also communicating with each other (need standardised communication protocol), as well as traffic signals and other traffic control and information system. So AC/DC will be seen to be "dancing" - speeding up or slowing down and moving left or right in the most efficient manner for ALL vehicles on that part of the road. So the scenario that Cecil proposes - hitting a jaywalker or another car - would not arise. As the Jaywalker is spotted, the AC/DC that would hit him would swerve and the car which it would have swerved into, would have sped up or made way for the swerving car. All other cars that need to give way, would do so.

Of course, all this will be moot if you have a "signal lost" and the cars are "dumb".]



No comments: