Five flaws in our idealized future of self-driving cars

Autonomous vehicles are the future. These completely self-driving cars could change our lives, heralding an era of greater convenience, improved productivity and safer roads. Like any transformative technology, driverless cars are attracting huge investment — from traditional automakers, established tech giants and smaller startups. But all that promise is also fueling a great deal of hype and ungrounded speculation. Here are five persistent myths.

Recent studies from the University of Michigan and KPMG predict that the arrival of autonomous vehicles (AVs) will reduce private car ownership in the United States by 43% by 2030, and sales in urban markets by 50% by 2035. They hold that families will need fewer cars and that many people will opt for "mobility as a service" — a supposedly cheap and efficient model in which travelers will summon vehicles on demand instead of keeping one (or more) in the garage. Even the usually levelheaded BBC recently declared that "you have (probably) already bought your last car."

But people do not buy automobiles simply to get around; they own cars to get from A to B in a way that's convenient for them. This means a car at your doorstep, available at a time of your choosing (with no charge if you change your mind) and with your belongings exactly as and where you left them. Automobiles signal our values and extend our private space — things a shared service cannot offer.

Nic Lutsey of the International Council on Clean Transportation sums this up perfectly: "In reality, a lot of people will have the same inclinations as they do today, to own a private auto and use it the way they want, without compromises."

Even if the inconvenience can be overcome, a recent study from the Victoria Transport Policy Institute suggests that owning a non-autonomous car will, for many users, continue to be cheaper than buying a self-driving car or hailing a ride, for some time to come. Indeed, car sales in the U.S. are at their highest level in 40 years, and Americans are keeping their cars longer than ever. It is going to take more than a new generation of highly efficient taxis to eliminate private automobile ownership.

Automakers, forward-thinking politicians and tech leaders like Google co-founder Sergey Brin have long asserted that self-driving cars will make congested streets a bad memory. They say these vehicles will be able to travel in tight groups, known as platoons, which will pack more cars onto the road, save fuel and allow traffic to move in a mathematically optimized fashion. Science magazine recently reported that introducing even just a small number of AVs onto the roads could improve overall traffic flow and reduce trip times.

But without careful management, autonomous vehicles will make traffic worse. City-center parking is expensive, which creates an incentive to keep moving. This means self-driving cars will slowly cruise the streets, by the thousands, as they await their next ride or duty. Research from the World Economic Forum shows that as people choose driverless vehicles over public transport, traffic could increase, and parts of our cities could become more congested, not less.

It is easy to forget, however, that autonomous-vehicle software can be applied to existing hardware; both electric and gasoline-powered automobiles can be self-driven. Even if many autonomous vehicles do turn out to be electric, their environmental impact will depend on the source of that electricity: Are their batteries charged by power from coal, gas, wind or solar?

As a former chief executive of BP and co-head of the world's largest renewable-energy investment fund, I have spent more than two decades advocating for action to decarbonize the economy and update our energy infrastructure. Without such action, it is hard to say what effect self-driving cars (or indeed electric ones) will have on carbon dioxide emissions.

KPMG forecasts that total annual mileage covered by the U.S. light-vehicle fleet could rise from 2.8 trillion miles today to nearly 4.5 trillion miles by 2040, as driverless cars make road travel cheaper and more readily available to both older and younger customers. If this is right, it seems likely that self-driving vehicles could increase our environmental impact.

AVs carry the promise that human error in driving can be eliminated. This would be great news for road users — but not necessarily for auto insurers. A 2015 study by insurer Metromile estimated that, for self-driving cars, the average automobile premium could fall by $1,000. KPMG predicts that this could result in an overall contraction of the auto insurance market by 60% by 2050.

But autonomous vehicles cannot remove risk entirely; they will just reapportion it. AVs will still make mistakes, and computers will still crash, injuring or killing people and damaging property. AVs also introduce substantial new sources of risk, like the very real possibility of being hacked by malign actors. As cybersecurity expert David Barzilai told the Financial Times: "In a data center [the biggest risk] is loss of data. With the car it is loss of life. The stakes are so much higher." While it's not clear who will be responsible for insuring against these incidents, it is unrealistic to believe that insurance will no longer be necessary.

The reality is more complex. First, we do not have the right tools to verify the safety of AVs. In aviation, safety is underpinned by universal standards and exacting formal testing, which aren't yet in place for autonomous vehicles. Self-driving cars don't have enough miles on the road to compare their safety record with that of conventionally operated vehicles.

Second, even if it could be demonstrated that AVs are safer than human drivers, it may not be enough to win the public's trust. We seem to tolerate humans killing other humans in accidents, but we do not tolerate machines killing humans. That is why the crashes of two Boeing 737 Max airplanes, in which software caused tragic accidents, have provoked such fury and fear. For as long as there is a risk of an accident involving an AV, there will be a degree of mistrust.

Meanwhile, test versions of self-driving cars have already killed passengers and pedestrians. The public is understandably wary of this technology: A recent Brookings Institution survey showed that only 21% of U.S. adults would willingly ride in an autonomous vehicle. Too many more fatalities, and the development of autonomous technology could be seriously derailed, which might in turn make it harder to adopt better safety standards. Driverless cars are not yet operating in a driverless world, and many of them still seem to behave in ways that lead to collisions with human-operated vehicles. If we are scared off by such problems, we will all miss out on the potentially transformative benefits of vehicles that drive themselves.