A cloudy outlook for self-driving cars

Over at Gormogons, GorT muses on the prospects for self-driving (autonomous) cars, and thinks that it's more or less inevitable that these will come to dominate the market.

The Antiplanner looks at President Obama's recent call for mandatory use of this technology and delves into the downsides:

First, V2V [Vehicle-To-Vehicle communications - Borepatch] and V2I [Vehicle-To-Infrastructure - Borepatch] communications pose serious security risks for travelers and cities. With V2V communications, an automobile that suffers a fender-bender would communicate to all nearby vehicles that they ought to take a different route to avoid congestion.


That sounds good, but what happens when someone hacks the system and puts out radio signals in a hundred or a thousand critical urban intersections that effectively shut down traffic in an entire city? As one expert at the driverless vehicle symposium observed, “just think of the opportunities for chaos!”

Second, V2I communications will allow the nanny state to monitor and control when and where you travel. For example, PC Magazine observes that V2I is “so accurate a revenue-hungry town could write tickets for doing 57 in a 55 zone.”

Worse, suppose your state decides to cut per capita driving in half, which isn’t far fetched considering that in 2008 the Washington legislature passed a law mandating such a reduction by 2050. With V2I communications, the government could decide you have driven enough and simply shut off your car.

Third, what happens when all cars are dependent on V2I systems that the government can’t afford to maintain? The federal government is notorious for funding capital projects and then providing inadequate money to maintain them, and state and local governments are little better.

Finally, V2V and V2I communications will be unnecessary added expense to auto ownership.

Speaking professionally, the security risks are way, way worse than even this (pretty good) overview gives.  The idea of targeted attacks is very plausible, particularly if (as is likely) the comms systems are Internet-enabled (say, via OnStar).  Even if they are not directly Internet-enabled, they look to almost certainly be on the same comms bus with systems that are Internet-enabled.

This will be a very high value system to compromise, which tells you everything that you need to know to understand the nature of the attackers it will attract.  It ain't going to be Bobby Scriptkiddie.

And mandatory government tinkerable systems seem to be a stretch until issues like NSA metadata collection and use are resolved, which looks to be just a little before the heat death of the Universe.  Given that people's faith in the Government to be (a) competent and (b) non-malicious is asymptotically approaching zero, just how would Congress pass a law like this and survive the next election?

But it gets even worse - there will have to be controls over who can tinker with these systems, or Charlie Carowner will just disconnect the damn thing some Saturday.  So only licensed mechanics will be lawfully able to do this.  OK, how many mechanics are there in the good old US of A?

There are approximately 763700 people employed as an Automotive Service Technicians and Mechanics.

How many of these would be able to be bribed or intimidated to get access to the systems (i.e. facilitate compromise of the system)?  Consider a threat scenario when a large organization (Government Agency, Union, Special Interest Group, etc) really wanted to gather information on a particular individual.  Could you bribe a mechanic to add some code for, say, $25,000?  That's rounding error for these organization's budgets.

But GorT brings an even stronger reason why this is almost certainly not going to become widespread anytime soon:

There really is a bottom line as to why Google and other companies are looking into it: data.  The data that could be collected from the cars, passenger ids (for anti-theft / authorized usage) and travel patterns would be huge.  Toss in there the ability to market those passengers and it becomes even more enticing.

Google and Facebook are burning down any level of trust that they may once have had, because GorT is exactly right - the your data is worthless to them if they can't monetize it.  It's one thing to see pictures of your BFF's baby niece, it's another kettle of fish entirely for Google to know exactly where your car is (and if you are in it) every second of every day.

Why on earth would anyone ever want one of these cars under those circumstances?  So that you don't have to drive home after a couple too many beers at the pub?  Get a cab.

Right now you get free Internet search (and maybe email) - you get something of value from Google.  What of value do you get for letting them follow you around everywhere you go?

The potential downsides are so, well, creepy - and so hard to explain to your Mom without it sounding creepy - that these products will almost certainly be toxic in the marketplace.  And that's without adding in the Government-revenue-grabbing/busybody-in-your-bidness/NSA-stalker overlay of added creepy from the Fed.Gov.

Until the Government and Silicon Valley do a much better job of showing that they can be trusted with this sort of data, or showing that there is a much better ("killer app") benefit to the car owner, this isn't going anywhere other than the drawing board.