Bloomberg has a long article about Google’s lawsuit against Uber over self-driving car technology. In a nutshell, one of Google’s top engineers, Anthony Levandowski, left Google to start a new company called Otto that was then purchased by Uber for $700 million, and Google is accusing Levandowski of taking its company secrets with him and giving them to Uber.
The real story, though, is not over patent disputes but a debate in the industry over how to introduce the new technology to the market. This debate has to do with the distinction engineers are making between self-driving cars and driverless cars. Advocates of self-driving cars, meaning cars that can increasingly drive themselves but sometimes need humans to take over, argue that this stage is needed to collect as much information as possible to perfect the technology.
On the other hand are advocates of driverless cars, meaning cars that never need a human operator, who argue that not only is the self-driving phase not needed, but that it could be dangerous because a self-driving car may not be able to alert on-board humans that they need to take over in time for them to do so.
Add Intel to the list of companies working on self-driving cars. It just spent $15.3 billion purchasing Mobileye, a manufacturer of sensors used in autonomous cars. Intel’s CEO says he expects to have a complete hardware package ready for auto makers in 2024. Considering Ford’s promise to have fully autonomous cars on the road by 2021, that might be late, or it might just be more realistic.
Meanwhile, after much criticism from the industry, California has revised its proposed rules for self-driving cars. The original rules did not provide any possibility for testing of cars that did not allow a human override. This led Google and other companies to migrate their testing operations to Texas and other friendlier states.
Most states still don’t have any laws providing for self-driving cars, but because the people who wrote those laws never conceived of the possibility, most states also don’t outlaw them. Arizona, for example, has no law, and the governor “welcomes them with open arms.”
A new study estimates that self-driving cars will save the United States more than $300 billion per year. The study adds up the costs of traffic accidents and assumes that self-driving cars will reduce accidents by 90 percent. That’s optimistic, but the study doesn’t even count the savings due to congestion relief, increased productivity while traveling, and the reduced cost of delivering goods and services.
On the other hand, one analyst estimates that self-driving cars will “wipe out 4 million jobs.” A taxi- and limo-driver lobby group has already begun to lobby the New York legislature to protect jobs by banning self-driving cars.
This is where it is important to understand the difference between benefits and costs. Jobs are not a benefit; income is the benefit. Jobs are costs: if more income can be produced with fewer jobs, everybody gains. That includes the people whose jobs are lost because–at least if the society is reasonably mobile–they can find better jobs instead, paid for out of the money people saved by reducing costs.
Google has spun off its self-driving car programs into a subsidiary called Waymo (which is apparently short for “a new WAY forward for MObility”), and Forbes celebrates by claiming that this is “waymo” than just a car. In fact, the real significance is that, by moving self-driving cars out of the company’s X Lab research division, Google is signaling that its technology is sophisticated enough that it is ready to start working on sales and not just research.
Uber has gotten headlines by starting a self-driving car-sharing service in San Francisco without getting permission from the state. This was supposed to be similar to the service it has going in Pittsburgh, where it is legal. The state of California immediately ordered Uber to shut down its service. (When someone documented Uber vehicles running red lights, the company blamed it on the drivers, not the self-driving technology.)
This is ironic because California’s self-driving car law was passed at Google’s instigation to allow for experiments like this. But the state passed regulations that were stricter than Google expected, so now even Google is doing most of its experimentation in places like Texas, which hasn’t passed a self-driving car law. Legal scholars say that operating a self-driving car is legal in most states so long as a licensed driver is behind the wheel ready to take over if necessary (which is how Uber is running its trial in Pittsburgh and planned to do it in San Francisco). But the California law is much more stringent.
Retired General Motors executive Bob Lutz ruminated recently about the future of self-driving cars. He imagines “they’ll look like telephone booths laid down” and they won’t need to be streamlined “because they’ll be electronically linked in a seamless train on the freeway moving at say 200 mph.” This will happen in 15 to 25 years “depending on how quickly governments are willing to invest in the road technology needed for a fully automated. . . system to work.”
What Lutz is describing is Futurama, not the television show but the General Motors exhibit at the 1939 New York World’s Fair. That exhibit imagined that highways would have embedded infrastructure that vehicles could electronically read and follow to get to where their occupants wanted to go.
But that’s not how most auto and software manufacturers are designing their self-driving cars. Instead, as the Antiplanner has noted before, they assume that the government will provide no infrastructure other than what is already in place, and all of the electronics and software needed to guide the cars will be on board the vehicles themselves.
The Department of Transportation says that it plans to issue a series of rules for self-driving cars that will potentially preempt state laws and regulations. This comes after lobbying by Google, which was disappointed when a state law that Google had supported led the California Department of Transportation to issue rules that forbade the use of cars that didn’t allow human drivers to override. Since Google was planning cars that didn’t have steering wheels and other controls that drivers could use, the state rule conflicted with Google’s goals.
The Antiplanner has urged against federal regulation, fearing that the feds would be as likely to get it wrong as the states, whereas if the states were left to regulate, at least a few states would get it right and the others would emulate their examples. Federal regulation wouldn’t be bad if the rules were perfect, but how likely is that?
For a more detailed free-marketeer’s view of the Department of Txansportation’s proposal, see Marc Scribner’s analysis. Here, I want to focus on one thing: the debate over fully autonomous vs. semi-autonomous vehicles.
People who remain skeptical of self-driving cars simply aren’t paying attention. The biggest news in the past week is that Ford’s chief executive, Mark Fields, has pledged that his company will have “fleets” of totally self-driving cars–with no steering wheels or pedals–in American cities by 2021.
His wording makes it appear that Ford will not only sell the cars to consumers, but offer Uber-like car-sharing services itself. To help it reach this goal, Ford recently purchased SAIPS, an Israeli company specializing in machine learning and sensing.
General Motors, meanwhile, spent $1 billion acquiring Cruise Automation, a company that the Antiplanner considered to be pretty fly-by-night. This company had promised to turn any 2012 or later Audi into a self-driving car for $10,000. I think all it really did was add adaptive cruise control and lane centering, so cars could drive themselves on freeways, but not on city streets, nor could they navigate from one place to another. Yet GM appears to have been impressed.
Two years ago, the Antiplanner predicted that self-driving cars would put most transit agencies out of business. So it’s not surprising to see push-back against self-driving cars from transit supporters. What’s surprising is that it took so long.
“Cities need more public transit, not Uber and self-driving cars,” says Kevin Cashman, a policy analyst with the progressive Center for Economic and Policy Research. “We don’t need self-driving cars — we need to ditch our vehicles entirely,” argues California writer Rebecca Solnit in the Guardian.
Cashman’s argument is that self-driving cars won’t be “affordable,” while public transit is. Excuse me? In 2014, American transit agencies spent $59 billion to move people 57 billion passenger miles (see page 106). That’s more than a dollar per passenger mile.
The OECD has published a fairly realistic report about the effects of self-driving cars on urban transportation. However, the report contains some implicit assumptions that may not come true.
“Single-occupancy car use generates individual and collective benefits,” admits the report, “but these are eroded and, in some cases, obviated by environmental impacts, loss of transport system efficiency due to congestion, social inequity and exclusion, as well as road crashes and strong dependence on fossil fuels.” Except for congestion, however, most of these problems are exaggerated and even congestion is technically easy (but politically difficult) to solve with congestion pricing.
“Typically, the response to the negative impacts of car-dominated transport systems,” continues the report, “has been to promote public transport.” But this “comes at a heavy cost,” and despite many cities paying that cost, “public transport continues to lose market share to private vehicles in most developed economies.”
Tesla announced yesterday that one of its self-driving cars was involved in an accident in May that killed the occupant, who was in the driver’s seat but letting the car drive itself at the time. The car was on a Florida expressway and, instead of stopping when a large tractor trailer crossed its path, attempted to drive under the trailer. The low height of the trailer sheared off the top of the car.
Last March, Duke University roboticist Missy Cummings testified before Congress that auto companies were “rushing to market” before self-driving cars are ready, and “someone is going to die.” She didn’t mention Tesla by name, but since that is the only car company that has produced a car capable of not just avoiding collisions but passing other cars, she must have had it in mind.
The person who was killed in the Tesla crash, Joshua Brown, had posted two dozen videos on Youtube showing how his self-driving Tesla responded in various situations. One of them received 1.7 million views for showing a near collision that the car avoided when a truck pulled into the car’s lane.