A Google self-driving care encounters two cyclists (red objects) at night. That retro arcade game look is charming but I imagined in my techno naivete that it'd all be much more...advanced

NYT columnist Joe Nocera is enthusiastic about the potential of driverless cars to eliminate the 1.2 million deaths on roads worldwide each year. In an article republished by Fairfax on the weekend (Google’s driverless cars will make roads safer), Mr Nocera says “the sooner they are a reality, the safer we’ll all be”.

Driverless cars have enormous potential to change the way we travel, at least in theory. They could indeed virtually eliminate road casualties, as well as multiply road capacity by as much as eight times according to some estimates. They could also “create” productive in-vehicle time (including sleep), reduce the size of the national vehicle fleet, and significantly lower the cost of private travel (see Are driverless cars coming?).

But while I think governments should already be thinking about how to deal with them in the future, I don’t think driverless vehicles are going to take over the roads of our cities for a long time yet. I certainly don’t see them having much impact within the 5 year and 10 year time horizons I commonly see cited by the industry.

Most observers seem to think getting the technology right will be the easiest obstacle to overcome; the major problems will be how social institutions adapt to the new technology.

One line of thought argues that the complicated legal and social adaptations required by driverless vehicles, especially during the transition period when there’s a mix of driverless and (much more dangerous) human driven vehicles on the road, will be extraordinarily difficult.

Tyler Cowen notes that almost all States in the US assume the existence of a driver who is legally responsible for the conduct of the vehicle. Politicians will find it difficult, he says, to avoid clamping down on autonomous vehicles in the early years when the inevitable mishap occurs (almost certainly not caused by the driverless vehicle, though).

There could be demands to shut down the cars until just about every problem is solved. The lives saved by the cars would not be as visible as the lives lost, and therefore the law might thwart or delay what could be a very beneficial innovation.

Joe Nocera notes the difficult privacy issues associated with the collection of data. He says consumer advocates are right to press Google – and all the big tech companies – on privacy issues. The profligate use of personal data has become a big concern for many Americans, he says.

Another issue is more straightforward but no less difficult. Even if the technical and socio-legal barriers can be overcome, driverless cars won’t reach a tipping point until they’re affordable.

Building and maintaining maps is currently expensive and in the early years at least the vehicles will be too. While it’s reasonable to think the cost issues will be addressed with scale (and theoretically driverless vehicles can be lighter) that will take time.

But another reason why driverless vehicles might be some time coming goes against the accepted wisdom. Despite what the boosters claim, the technology is actually a long way from being viable. The constraints are lucidly explained in this article published in October last year on Slate, Driving in circles.

The author, Lee Gomes, says Google’s driverless cars can’t yet operate in heavy rain and snow, avoid potholes, find a space in a multi-level garage, or “tell the difference between a big rock and a crumbled-up piece of newspaper” on the road.

As of last month, Google’s 23 Lexus RX450h SUVs have clocked up just over a 1 million miles in autonomous mode on public streets. Sounds very impressive, but Lee Gomes says those streets are the same ones in Mountain View California that’ve been mapped in copious detail and are “driven over and over again”.

Google has developed a clever and expensive 3D mapping system to control the cars; they can’t move without them.

These maps contain the exact three-dimensional location of streetlights, stop signs, crosswalks, lane markings, and every other crucial aspect of a roadway. That might not seem like such a tough job for the company that gave us Google Earth and Google Maps. But the maps necessary for the Google car are an order of magnitude more complicated.

The company’s technology requires that entire cities and nations are mapped in high detail and kept up to date on the fly before driverless cars could be universally deployed. That can very probably be done but it will take time, political support, and money.

Unlike with the early history of the car, driverless cars will have to be technologically near-perfect before they can fully reap the promised benefits. In the meantime, manufacturers will loudly proclaim various technological advances in conventional vehicles, but the theoretical benefits won’t be fully realised until human drivers are completely unnecessary (and probably banned).

I think the most prospective application at this time might be in large trucks because a lot of freight traffic can be confined to a small number of major, limited access routes that can feasibly be mapped, regulated and monitored e.g. interstate transport on motorways. (1)

There’s a separate issue about whether the likely changes wrought by a world of driverless vehicles would all be positive. That’s an interesting and important issue I’ll discuss shortly.


  1. But then there are a number of driverless trains already in operation (e.g. this is just subways), so it can’t be assumed that driverless trucks operating on limited routes would necessarily be competitive in the future.