In the first of two recent articles discussing the emergent autonomous vehicle market, From Auto Sales to Mobility Miles, attempts to nail down how large of an autonomous fleet various car companies will need to amass in order to complement their existing business. The second, Carmageddon is Coming, paints an aggressive picture of the impending upending of transportation as we know it. Upon careful review, between both articles there are the following assumptions:
– That the shared vehicle economy will be competitively priced against car
– That the adoption of electric and autonomous vehicles will be rapid like
– That autonomous vehicles will be much safer than human drivers.
These assumptions are naive, at best, and paint an unrealistic picture of our autonomous future. Make no mistake – autonomous vehicles will be disruptive and will enable things that were previously considered impossible. But if you think that humans will wake up one day and come to the conclusion that sharing and trusting a robot is better than owning and trusting yourself, you’re mistaken. Let’s take a look at the arguments made: economic, social, and safety.
Industry projectors seem to be settling on a target of $1.25/mile driven in revenue for shared autonomous vehicles. The idea behind this number is to take the total amount taken in (revenue), from fees, fares, etc., and divide by the total miles driven. They combine this with Uber and Lyft’s 85% utilization rate, where utilization is the percentage of time that a vehicle is operating and collecting fees (in other words utilization is what percent of your total driving time are you making money), to imply that sharing will generate sufficient revenue for car manufacturers at 1-2% of total global miles driven. As $1.25/mi is around the estimate for car ownership ($0.90/mi), they feel this is a reasonable adoption rate. All of these numbers are wrong.
To begin, $1.25/mi is revenue – not the price people are paying. Even at 85% utilization, this implies a cost to the consumer of $1.47/mi. 85% utilization is wrong, too. It stems from Uber and Lyft’s metrics, but Uber and Lyft have human drivers who will stop driving once business becomes too slow. A fleet of autonomous cars operating all day will only find 85% utilization if they are small enough to meet 85% of demand during off-peak hours, however, you then cannot capture business during peak hours. Now, you have a maxi/min problem with the amount of cars you need to capitalize on business during peak, while maintaining a low-enough off-peak utilization. Some of this can be addressed with peak pricing, essentially increasing price during periods of high demand, but there’s always a floor and the reality is that humans just don’t drive enough outside of rush hours to make up for those down periods. Public transportation has been trying to solve this problem for ages, and the only way to do it is to subsidize transit. I’m doubtful that robot cars will fare better.
Human beings drive an astonishing 19.5 trillion miles every year. This is enough to drive to the sun and back around 100,000 times. The numbers in the mobility miles article add up to 1.1 trillion miles across eight leading car manufacturers. This amounts to about 60% of all driving done in the United States every year. Evenly spread across all of humanity, it seems highly unlikely that any company will get to 1%, let alone many in the same amount of time. It certainly will not happen if the cost per mile to the consumer ($1.47/mi) is 60% higher than the cost of alternatives ($0.90/mi).
It’s even worse than that for mobility miles, because included in that $0.90/mi is depreciation and financing. Once you own your car outright, the cost per mile becomes $0.56/mi. Why would anyone choose to spend three times as much getting around in a car they don’t own than they do now? That’s like saying people are going to stop buying houses and all live in hotels because the maid service is free.
Beyond the economics, let’s talk about the social aspect of transitioning towards both electric vehicles and autonomous vehicles. The incredibly rapid adoption of smartphones has skewed our perception of how technologies are adopted. Smartphones were able to capitalize on the mobile phone market, which turned over every two years due to the structure of phone contracts. Cars don’t turnover like that. A much better analog to the world of vehicles is the TV market. Innovation in TV happens and people adopt it, but it’s rare to see anything that causes a large shift. It’s very likely the last TV you purchased is a flat screen, it’s also very likely that that had little to do with your decision-making process. So, too, with EVs and AVs from a purchasing standpoint. If you’re in the market for a new car, you may explore those technologies, but it’s doubtful that you will rush out to upgrade. It’s simply too large of a purchase. My 2000 Camry still runs just fine and is paid off. The transition from ICEs to EVs and to AVs will happen inline with the market upgrade cycle for cars, and that cycle is slow and plodding.
Steve Jobs had some choice words on the second part of the TV subject. This is the bigger part of entrants into the autonomous vehicle market: fragmentation of the driving ecosystem. It’s one thing to put sensors on a car for forward warning of collisions; it’s quite another to make a system that can navigate a large east coast city built 200 years ago when people still rode horses. How do you deal with this as a car company? Do you make the Washington D.C. version of a car? Do you put a cap on its generalization (say, stay out of complex downtown areas)? And, ultimately, how do you do this in a way that provides drivers with greater utility than their own decision-making? At moovel, we have spent years toiling to digitalize the myriad fare systems for public transportation that our apps are in. Extending that to the plethora of arcane rules around driving will be no easy task.
Which brings us to safety. I’ll concede that early data seems promising, and to the extent that sensors and early warning help keep us all safe I am all for them. Before we go too far, though, I would like everyone to read the story of the Therac-25. While I am worried about drunk drivers, in an autonomous world I would be much more worried about the developer on a rushed deadline who pushes lousy code and accidentally kills thousands. You see, when human beings fail, we do so in isolation – but when machines fail, they do so at scale and that scares the hell out of me.
I don’t worry much about my car breaking down, because it was engineered by certified professionals working with the benefit of the learnings of the last 100 years of mechanical engineering. One can’t even be called a mechanical engineer without a rigorous certification process. Software is much less mature, and while the fact that anyone can buy a book on programming and go on to invent the next great app is one of the beauties of the industry, it doesn’t translate all that well to the rigorous attention to detail necessary when human safety is at stake. Let me assure you that I will never work on your robot car, because I don’t trust myself enough to entrust your safety to me.
I’d be much less concerned if the people promising our autonomous future seemed to realize the gravity of that responsibility, but, unfortunately, they don’t. The risky movements of Travis Kalanick and his ilk may be good at garnering funds from VCs, but it’s hardly the personality you want from someone who wants to build robot cars to drive your kids to soccer. Every time one of these mavericks starts promising a world devoid of human suffering from automobile accidents, all I can see is some M. Night Shyamalan-nightmare, where half the cars in the world start accidentally driving into buildings because a developer forgot that Arizona doesn’t follow daylight savings time.
I am bullish on the possibilities of automation, but I urge all of us to be cautious of grandiose claims. All it’ll take is a Tesla plowing through your block party because it didn’t recognize the rope your neighbor put up as a barrier for us to slow down and make sure we approach the question of safety appropriately. The FDA takes years to approve drugs; I don’t see any reason why we shouldn’t be equally cautious about autonomous driving. Traffic accidents are the leading non-biological cause of death, and I’m not sure that’s the right place to launch our autonomous future. Perhaps we should start with robots that make breakfast.
The future of electric vehicles and autonomous vehicles is upon us. They will open up new possibilities, revenue streams, and will create new efficiencies for us. If everyone involved can make a future where transportation is cheaper, safer, and ubiquitous, I am 100% in. At its core, that’s the promise of public transportation, and it’s why I enjoy working in this space. And, no matter what happens to cars, there will still be buses and trains, and we’ll help you get on board.