The driverless car is one of those ideas that was obvious long before it was feasible. As soon as Marconi had demonstrated radio transmission of sound, televisions and videophones were on the agenda. Engineers toyed with driverless cars in the 1920s, with no success; the scheme requires massive cheap computation, which has only become available in the last decade.
I looked for an SF magazine cover with transport pods from the 1930s. In vain: the problem is not that writers hadn’t thought of it, it’s that it’s not sexy, in the white girl/BEM/blaster convention of the genre. The best I could find was this alarming robot enforcer from 1935. Contrast today’s unthreatening prototypes.
We don’t get the technical progress we need, but what comes easiest. Cures for cancer and Alzheimer’s? They will be in the post, some day soon. Multi-player role-playing video games, that we never missed before they showed up: step this way. Sometimes need and feasibility coincide, as with the smartphone: the prototype of the universal communicator terminal of science fiction, and a social revolution. Are driverless cars more like video games or smartphones?
First, the timescale. Google is only the best-known of the companies that have put driverless cars on the public road: 23 converted Lexus SUVs around San Francisco, as of today. They call them “self-driving cars”: the human backup will disappear eventually with full driverless automation. Google plan to put a self-driving car on the market in 2020, which looks a reasonable timing. They seem quite serious about it.
Google and its rivals seem to have abandoned one possible approach to driverless cars: centralising the intelligence. It has proved extremely difficult to write robust and comprehensive air traffic and train signalling software, which have to deal with a tiny fraction of the numbers of cars moving round a city. So this rejection looks sensible. They have decided to put all the intelligence in the car, with no contribution from the roadway. This requires a very fine-grained and accurate map, and a very expensive radar ($70,000). One problem not yet solved is dealing with situations not in the map: a police signal by an accident, say.
My pennyworth: put some low-level intelligence in the environment. With really cheap $5 IoT sensor chips with radio, you can scatter the mental equivalent of cicadas by the thousand along the roads. Each robot cicada wouldn’t stridulate much: “Hello world! This is Berkeley lamp-post 4843 on Main Street! It’s dark! I am switched on! … Hello world! …” Surely it would be easier for a car to navigate through a helpful chorus of low-level information of this sort, coupled of course to the map, rather than insisting on working everything out by itself, like a fallible human driver. The police stop would be part of the same system, but with a peremptory override: “Hello world! This is Berkeley police cruiser 57, temporarily at 420 Vine for an accident scene. All vehicles travelling westward on Vine are required to clear the inmost lane and keep to a 20 mph speed limit for the block. … Hello world! …”
It looks a good bet that the technology will be solved in the next decade. Mass adoption depends on the price. The $70,000 radar is a real snag, and it’s an analogue, power-hungry device. Perhaps my chorus of cicadas will come to the rescue. It is possible that driverless cars will remain a curiosity, toys for the rich like commuting helicopters, which have been inaccessibly feasible for 50 years. It is also possible that they will take over. It is not too early to start thinking about the consequences if they do.
These will be very positive.
32,719 people died in traffic accidents in the USA in 2013. The WHO estimate the world total of traffic deaths at 1.24 million in 2010. The injury toll is an order of magnitude higher. The EU Commission:
For every death on Europe’s roads there are an estimated 4 permanently disabling injuries such as damage to the brain or spinal cord, 8 serious injuries and 50 minor injuries.
These are very large numbers. Contrast reported civilian plane crashes, responsible for 1,088 deaths worldwide in 2014, a bad year. The motor car driven by humans is not safe.
A driverless car will certainly be much safer. In 1 million miles of road operation, Google’s cars have been involved in 12 very minor accidents, none of them (they say) due to failure of the automation. Production vehicles will be at least as good. Driverless cars cannot get drunk, take drugs, fall asleep, get into conflicts with other drivers, answer their mobile phones at complex freeway junctions or engage in heated arguments with fractious children in the back seat. There is a small risk of electronic failure, but car- and chip-makers now know how to design onboard computers for ruggedness. Cyberwarfare? Ming the Merciless could hack into the city police override computer, and steer the passengers trapped in their pods to slavery in his uranium mines, except for the D-cup young women who would go straight to his harem. Maybe. However, Ming could have a go today at the systems responsible today for air traffic control, the electricity grid or the Internet backbone, and he hasn’t seized them yet.
Car drivership cannot be universal. It is limited to able-bodied adults. The very old, children, and many people with disabilities or chronic illness have to rely on somebody else to drive them where they want to go. By my heroic guesswork, that’s around a third of the population (footnote). You may say: they can use taxis today. But taxi regulations were designed for horse-drawn vehicles, and taxis are often kept expensive through restrictive licensing (to limit the volume of horse manure, perhaps?). See Mark’s findings on the gains from Uber. Assuming driverless cars replace driven ones at similar cost, that would expand and equalize access to mobility.
Lower numbers from rental
The third gain is more speculative. Driverless car boosters suggest that the vehicles would not typically be owned by individuals but by rental fleets. Fleets can get very much higher usage from vehicles, so you need fewer of them: 30%, according to Stanford computer scientist and former Google project leader Sebastian Thrun. He’s conservative; drawing on experience with car-sharing pools, other studies for Lisbon and Austin suggest up to 90% fewer vehicles, on paper. The change translates to lower costs per trip, since the capital cost is independent of usage.
There is no reason to think traffic would go down: probably the reverse, through the access effect. The problem of congestion, and the political difficulty of solving it by congestion charging, would remain intact. However, you would get an enormous reduction in the total number of vehicles, and the need for parking spaces, which tie up a lot of valuable city-centre real estate.
The effect on climate change would be positive, though not huge. The heavy lifting is done here by the shift to electric traction, not automation. But fewer cars require less in the way of steel, plastics and batteries. It also means that the transition could go much quicker – you might only need to replace one in three of the world ICE car fleet, 258 million not 773 million. That’s 3½ year’s annual production of cars, not 9½ years. The trucks and buses are used much more intensively and you would not get much of a reduction from a parallel switch.
Is this scenario plausible? People are very attached to their cars – though less than formerly. My parents gave names to their cars, like their pets; I never have. But to me the boosters look to me to have the
psychology right. The motor car seamlessly replaced the horse as a symbol of status and masculinity. Learning to drive is gaining mastery (sic) over a powerful, exciting and potentially dangerous machine, formerly animal. In Rio de Janeiro, I’m struck by the absurd priority given by the city fathers to drivers over pedestrians as users of the public highway: an atavistic survival of the feudal priority of the seigneur on his horse over the peasant on foot.
Driverless cars completely break this psychosocial complex. The car or pod will become a mere convenience, with no more affective importance than a dishwasher or TV. By design or regulation, they will all travel at the same speed, and choose their optimal route without human intervention. There will be few mental barriers to switching to rental if it’s cheaper and more practical. City ordinances will steadily make life harder for the holdouts who insist on driving their own car, especially an ICE one. Life got difficult for horse-drawn vehicles in American towns in the 1920s.
Footnote: Guesstimate of the UK population excluded from driving
UK 2014, official data:
Total population 64.51 m
Age 8-16 6.47 m (10.0%)
Age 75+ 10.68 m (16.6%)
Total number disabled persons : 11 m
Remove double counting: all disabled children (6% prevalence of disability:
0.39 m 0.77 m) and disabled old people (45% prevalence: 2.35 m), and you get a total excluded population of 19.95 19.05 million or 30.9% 29.53%. [Corrected to remove the disabled children 0-8 from the disabled total as well as 8-16]
This is very rough of course. A good number of 80-year-olds and disabled people are fit to drive. On the other hand not every teenager gets a driving licence at 17, and I’ve left out illness, fear, dislike and test failure as additional obstacles to driving.
* * * *
Thanks to Keith for pointing me at this.