What happens when self-driving happens
A couple weeks ago I spoke to a working group of economists and climate change experts in Norway about what’s happening right now in autonomous driving: how fast it’s likely to go mainstream, the effects it might have, and the opportunities this creates both in government and in business. Below are my remarks.
Fully autonomous driving has the potential to have as large an effect on our way of living as the car itself did. So it’s worth getting past the immediate “Wow” reactions, as well as some of the initial knee-jerk fears, and start to lay the groundwork for what’s coming next.
Imagine what it must have felt like to be in a major urban center like New York in the early 1900s when, in a sea of horses and humans, the first automobile “put putted” past. The first reaction, would be pretty obvious, “Wow, I want to get one of those!” the second, perhaps more ambitious entrepreneurial type of reaction might be, “Wow, I want to build my own car company and make one too.” But I want to talk about the conversations that talk about follow-on effect, the one’s that often take years for people to realize. The second-order effect conversations such as, “that thing is going to need a place to get gas.”
Autonomy in automobiles has been in the research phase in academia and in large companies’ R&D departments for a couple decades now. On the surface not a lot about the technologies has shifted from the demos you could see at places like Stanford and MIT a decade ago. So why all the activity today and excitement about it becoming a reality?
The primary reason is a combination of three different interlocking trends that create a perfect storm for a massive disruption to urban mobility.
- Computer vision. Advances in machine learning have allowed computer vision to finally be good enough to distinguish objects on the road, build 3D maps of the surrounding area and be supported by processor speeds powerful enough to be able to operate them natively in a car.
- Ride-sharing. The self-driving technology stack is currently still quite expensive, but the advent of ride-sharing companies like Uber and Lyft has presented the possibility of amortizing that capital cost over many drivers. Projections of the cost make it look like consumers won’t be buying a mass product autonomous car for over a decade, but if you amortize the cost over a lot of drivers through a ridesharing model then you could go to mass market tomorrow if the technology got there.
- Electrification. Virtually every fully autonomous car company today is planning on an EV platform. This is only secondarily about the environment. Primarily it is because the cost of maintenance of an EV car is dramatically lower. If a single company owns the car and it has an incredibly high utilization rate, then that lower maintenance cost is easily worth the higher upfront cost. An additional benefit is that when talking to local governments about getting clearance to put potentially thousands of new cars autonomously running around your city, it’s nice to be able to talk about them helping air quality.
These three trends in concert have created a unique moment in time, but a secondary reason should not be missed either. In order to solve a problem you first need to be paying attention to it. And while autonomy has long been a love of research labs, it has only recently become an area of entrepreneurship. The catalyst here undoubtedly was the massive profits generated by Google by their dominance of the web ad business, which created such a surplus that they were able to start working on advanced projects like fully autonomous driving. Most prevailing economic theories don’t subscribe to the idea that monopolies create innovation, but it’s precisely the money machine that is Google adwords that allowed it to attempt a project like this.
It’s also the right time for the larger startup ecosystem, since the general sentiment is that all the low-hanging fruit is gone from the mobile Internet. A founder that might have built a mobile app three years ago, and a Facebook app three years before that, and an Internet web portal 10 years before that, is having to dig deeper to find truly new ideas.
Don’t think these are the same entrepreneurs? Consider the founder of Cruise, which was sold to GM this past year and where I was privileged enough to be a board member. The CEO of Cruise, Kyle Vogt, was previously co-founder of Twitch, a streaming game service which was sold to Amazon in 2014. For Kyle, Cruise represented a return to the autonomous vehicles he had loved working on while at MIT, but seemed impractical to tackle as a startup until now.
As another cogent example consider the bellwether accelerator Y Combinator. Ten years ago the 32 startups in their program were 50% consumer software, 33% enterprise software, and a handful of software developer tools. This year their request for startups included nuclear energy, healthcare, education, transportation, farming, and a healthy dose of international startups. We are going to see massive transformations across a lot of industries previously thought of as relatively stagnant precisely because attention is now shifting to them.
How fast is this going to happen?
Well, you can already hail a self-driving car in Pittsburgh through Uber, and very soon through other companies including Cruise and Waymo. But these have safety drivers sitting in front because it’s obviously not reliable enough yet. So the question is:
- How fast do safety drivers go away?
- How fast does this go from small trials to broad rollouts that would actually be seen on a macroeconomic level?
I think we should expect safety drivers to be around for a long time. Companies are going to be very conservative in their removal because there is little incentive for them to remove them until they are sure it’s going to be worth it. The ROI long term is so clear, and that makes it easy to be patient.
However, broad adoption is going to happen much faster than everyone thinks. Adoption is going to be closer to technology adoption curves than transportation adoption curves. This is largely because there is no capital cost involved. When we think about the advent of massive new technologies, like solar power, electric vehicles, or the personal computer we have to wait for a consumer to be able to get to a purchase point.
No one is going to buy a fully autonomous car, and it won’t matter
In the future laid out here remember there is very little incentive to actually own a car. This is about downloading an app for a car whenever you need. That means that the moment fully autonomous driving is available in a city you could theoretically see a majority of consumers move to that mode of transportation overnight. You don’t have to wait until a consumer is about to buy a car to increase adoption of this technology, it’s simply that they walk out their front door in the morning and decide they would rather just hail an autonomous vehicle. Their car sits in the driveway gathering dirt as they decide it’s cheaper, easier, and more convenient to get a little work done on the way to the office.
And yes, there will be regulatory barriers to the first fully autonomous driving city. But just think about what happens once the first city has 24/7 readily available autonomous driving. That city will feel like the city of the future. Overnight not having autonomous in your city is going to feel like when you lived in a place without broadband, or a city without mass transit, or electricity. Every other city or country is going to be on a mad dash to modernize to autonomous at that point.
It’s hard to say when in the next few years we will find the technology solid enough to remove safety drivers. But when we do, expect the speed of growth to be potentially faster than typical technology adoption curves, which were already accelerating. Without a massive upfront cost being borne by consumers, there is a potential for growth metrics that feel closer to Uber and Snapchat than the growth of Cars and Air travel. If GM, Toyota, Uber, or another company decides to ship enough cars to bring a city like Los Angeles online then you could see the kind of disruption in urban mobility that we haven’t seen in 100 years in a matter of months.
What will it affect and what are the opportunities.
Here’s where we get highly speculative. But I believe autonomous driving will be probably the single largest change to how we function as a society since the Internet. I’ve got a list of things here, but the effects are so pervasive it’s very hard to think about all the secondary effects.
1.The most obvious effect is accident avoidance.
Nearly 37,000 Americans die each year in car accidents, and nearly 1.3m die globally. For every death in the U.S. there are over 100 treated in emergency rooms, with an annual cost of $33B in 2012.* Self-driving here means a reduction in health care costs. It also means the car insurance industry changes completely, perhaps creating an opportunity for startups. Lastly, a traffic accident during rush hour happens roughly every day in a major American city, imagine the traffic congestion changes with that eliminated or severly curtailed. Most interestingly for human behevaior is that will normalize commute times far more than the random dice roll that a commute feels like today.
2.The marginal cost of commute times goes down. Free time goes up.
In traffic planning circles there is a principal called the fundamental law of road congestion. Typically there is a tight relationship between traffic capacity and the “pain of commuting.” So if you introduce a new traffic lane or add a tariff to commuting like they did in the city of London, it only helps temporarily because pretty quickly that capacity will be filled until you get back to roughly ~1 hour commute.
But if you start work when you get into your car and not when you get to work because of autonomous driving, then an extra five minutes doesn’t really matter as much. That probably means commute times go up as marginal cost goes down. It also means rural areas have a massive boom as people can commute longer who want more space. Inversely, cities also boom as the cost of living in a city goes down, an issue I’ll talk about in a moment.
All that would mean that the “suburban compromise” that the car created comes to an end because there’s no invisible line at 1 hour commute distance to an economic center anymore. Suburbs become a place largely for lower income workers who can’t afford to live in the city and also cannot afford the even marginal cost of a further commute.
3. Land use
5.7b square feet in the United States is dedicated to parking, it is estimated there are 105m parking spaces. Additionally, the metered parking in most city streets becomes superfluous.
What do we do with that space? Enterprising citizens and legislators can get ahead of the game and start drawing up plans. Imagine a city that passed an ordinance that any meters parking space removed due to autonomous driving effects over the next 20 years is mandated to turn into city parks, or a bike lane, instead of just additional traffic capacity. Or a city zoning that made it immediate to turn parking lots into affordable housing without any further permits.
Additionally car dealerships, independent car mechanics, and gas stations all become reclaimed territory in the landscape over time, starting in urban areas and eventually spreading.
4. Energy consumption effects.
There are a ton of things to model here, and most of the models I’ve found really only talk about first order, not second order effects. Just as a taste keep in mind that for non-policy reasons we just will have created a massive global fleet of electric vehicles. Secondly, ride sharing creates a natural incentive to carpool for even further reduced fares.
This is not to be underestimated. Already 20% of Uber’s traffic globally is UberPool, the single largest increase in pooling ever. MIT’s CS & AI lab found potentially up to 75% reduction in traffic just by broad adoption of the Uberpool/Lyft Line concept.
So there are a lot of effects here, and it’s early, but a paper from the Centre for Integrated Energy Research tried to quantify all the first order effects, from congestion mitigation to platooning, and found the possibility of up to a 45% reduction in energy demand across all road transportation.
Even more dramatic is a study from my neighborhood, at Lawrence National Lab at Berkeley. They looked at Autonomous Electric Taxis, this combination of autonomy and electric we’ve been talking about, and model some extremely high benefits — Jeffrey Greenblatt projected as high as 87–94% reduction in greenhouse gas in cities by 2030.
5. Work, part 1, the fear and the change
Every major technological change has come with worries about what displaced workers are going to do, and prognostications that the economy will collapse. John Maynard Keynes famously worried that current technology trends would give us only 15 hours to work every week. That was in 1930.
But that’s not to be callous to the displacement that is going to happen. There are 3.5m truck drivers in the United States. It is the most common job for males in this country. Truck drivers and cab drivers are the most obviously displaced, but the effects will be far-reaching.
Consider policing. 4 out of every 10 police encounters (42%) are traffic related (Bureau of Justice Statistics 2011), and traffic stops are the most common single way that the police engage with their community. Cops will go from grossly underfunded to suddenly having 30–40% more time. What does that mean for police staffing and work? Or perhaps their funding will be an issue; over $50m in revenue is generated from parking meters in the city of San Francisco alone.
And it will keep going: local mechanics, car dealers, consumer car washes, auto parts stores, and gas station attendees all get impacted in ways that will cause society to be forced to tackle complicated questions. Every politician for the last 40 years has been talking about re-education, and now perhaps we will get serious about it.
6. Work, part 2, the opportunities
With change also comes opportunity. It’s incredibly hard this early to see all the new opportunities that might be created, especially when you are trying to consider secondary effects of autonomous driving. For instance, at the advent of the smartphone you could have predicted a product like Google Maps or Yelp perhaps, but it would be highly unlikely you would predict Uber.
Still here are some ideas, in no particular order.
- As mentioned earlier, we’ll need a new insurance model for cars that are autonomous and shared. Perhaps the incumbents will create new divisions and hire people to manage these new models, or perhaps there will be new startups that take advantage of the opportunity.
- There’s a massive new opportunity created for new transportation technologies. Every previous technology, from fuel cells to EVs, has always had a network effect problem. Without infrastructure like gas stations they were stuck trying to gain adoption. But if a single provider owns the vehicles then it suddenly becomes very easy to develop a new energy technology that might be more efficient than current Lithium-Ion EVs and actually get it to market quickly. We’ve largely ceded battery innovation to Asia currently, but perhaps now is the time to re-invest in this sector.
- The large car companies are good at mass-producing what we’ve got today, but if you can order anything to show up in 10 minutes at the touch of a button, perhaps startups will invent new modes of transportation. Vehicles like the MIT City Car, or delivery robots the size of a trashcan, become a startup opportunity since you just need a small set of captive customers to get startups instead of the massive dealer networks required by the incumbents.
- Wireless networks will get potentially overloaded as significant amounts of commuters are working from the road. Perhaps the incumbents adapt, or perhaps there are new startups in wireless mesh networks that start to take their place.
- We won’t be working all the time, so what do we do when we are relaxing? The current mobile gaming revolution is often bound to 4–5 minute session increments largely because of affordances to how we use mobile phones. So what are the affordances when everyone has an hour each day on the road, with no interruptions from co-workers? The car as entertainment platform will likely require new affordances we haven’t considered yet.
- If distance matters less in driving, then perhaps the tourism industry changes. Air travel tourism goes down since you can go an extra couple of hours by car easier, with all the employment and real estate knock-on effects that this would cause for those towns suddenly bustling with explorers. That ski resort that was twice as good, but just a bit too far from the nearest airport now looks like prime real estate.
And there is reason to believe entirely new markets could be created. The creation of massive amounts of new free time in the past has had this effect. For instance tourism all but didn’t exist a few hundred years ago, but advances in transportation combined with a new bounty of free time has created a multi-trillion dollar industry employing millions around the world.
It can be creatively invigorating to think about everything that may happen, but it can be productive to do so now as well. We can pass civic policies to ensure that the transition creates the cities of the future we want. We can begin to think about the startups and new industries that could potentially be created. We can begin the process of rethinking education today to prepare the workforce of the future. It’s hard to understand fully what will happen, but it’s worth trying to put ourselves in that position now and lay the foundation to be ready when it does.