The Robocar Wars
You have probably heard a very powerful rumor: On the basis of reports that Google is developing a ride-sharing app, tech journalists have surmised that Google and Uber are now — as in, today — competitors in the global ride sharing market. What’s more, pundits have instantly concluded — without much reflection — that Google is going to be the clear winner due to its superior technical expertise, lobbying experience, and business relationships. According to this narrative, Google will ultimately undo Uber’s hard-won global footprint in the ride-sharing space.
Let’s be crystal clear: This is not a useful way to understand the transformation that our transportation system has started to undergo. We need a better framework. To develop that framework, let’s get rid of some sloppy, poorly grounded thinking.
First, Uber and Google are not entering a head-to-head battle for the future of ride-sharing anytime soon. Uber will have breathing room for the next couple of years before it has to deal with Google as a direct competitor. However, it is virtually certain that Google is taking steps to build an Uber competitor that will launch in a few years time. Second, when Google does launch its competitor, it won’t be an ordinary ride-sharing platform of the kind we’re familiar with today. Instead, it will be a transportation platform that uses fully autonomous vehicles (FAVs or robocars) to serve individuals and groups.
What of the notion that Google will ultimately defeat Uber when vehicles become autonomous? Most technology journalists assume that Google will be the victor. For example, Doug Newcomb of Forbes argues that Google is likely to prevail because FAVs rely upon superior mapping technology, a capability that Google already possesses. Newcomb’s logic is simple: Google has technological superiority and a substantial head start, thus it would be very odd if Uber prevailed. Let’s call this the common view, as it is shared by most journalists who have weighed in.
The common view has three main flaws. First, there are economic and regulatory factors that favor slower, incremental approaches to integrating autonomy technology into transportation technology. Second, this view ignores Uber’s near term interest in ensuring that partner vehicles share a common, standard set of safety features. Third, because it is bewitched by the technical expertise of Google, the common view underestimates the strength of Uber’s long term interest in expanding beyond its beginnings as a ride-sharing company.
What’s at stake in the robocar wars? Understanding what’s wrong with the common view requires reflecting on what the robocar wars are actually about. The very term ‘robocar wars’ seems to suggest that a few companies will build autonomous vehicles and then compete with one another for market dominance. The common view, then, says these wars are conflicts between companies.
But the war really isn’t going to be between Uber and Google, or between any two companies, for that matter. The war is going to be between two different visions of autonomy in modern transportation. The first is the incremental approach. Under this approach, autonomy technologies will be slowly integrated into human-piloted cars and then carried over into FAVs. The second is the non-incremental approach. The non-incremental strategy assumes that the only way to build reliable FAVs is to ensure that human drivers are taken out of the equation.
If the robocar wars are really about two visions of technological progress, it is easier to see why the technology press has miscast the conflict between Google and Uber. In terms of corporate strategy, the tension between Google and Uber is real, but the tension is not due to any immediate plans on Google’s part to compete with Uber. The tension was catalyzed when it became clear that Uber’s strategic need to enhance mapping, autonomy and safety technology in the near term could not be met by relying on Google’s technological advancements.
While Uber’s approach is incremental, Google is shooting for the moon. Stripped of steering wheels, gas pedals and other human controls, Google’s robocar will be introduced as a fully operational four-wheeled robot. Uber does not appear to be taking this route. Nor would it be in Uber’s near and long term interest to do so, as we shall soon see.
What is Google really up to? According to Google, autonomous vehicles could hit the US market in about 5 years time. Leveraging its auto-industry partners to distribute its platform would surely make Google a leading provider of autonomous vehicle technology. That would be significant step toward evolving Google’s revenue base, which at present is dominated by advertising revenues from Google’s websites and partner sites. This would be a profound change: While Google already makes billions every quarter — $18.1B in Q4 of 2014 — moving into the transportation space would give Google a chance to make a lot more revenue.
If the opportunity is so large, why not launch an Uber competitor now? First of all, Google invested over $250 million in Uber. When you invest that kind of capital, you want the return to be stratospheric. But a stratospheric return would be seriously threatened if Google launched a direct competitor that cannibalized Uber’s core ride-sharing revenue streams prior to its IPO. Uber simply wouldn’t be able to optimize its market value in an IPO while battling a foe like Google.
It would be wrong to think that investment strategy is the only reason Google isn’t going to start creating a competing service in 2015. We have already seen that Google is attempting to build a near-perfect autonomous vehicle. Of course, the company’s current prototype has significant limitations. The 3-D laser mapping system cannot dynamically adapt to new traffic signs and obstacles, the density of obstacles cannot be determined (e.g., a leaf and a stone cannot be distinguished), parking is not yet possible, and Google’s robocars do not have what artificial intelligence researchers call generalized intelligence — the ability to creatively adapt to an ever-changing, unknown environment. To be sure, Google intends to solve all of these problems before launching its FAV. While this is a bold strategy, it is also a problematic one. To understand why, let’s consider an existing car manufacturer’s approach to integrating autonomy technologies into existing vehicles.
Audi’s strategy is encapsulated in its Piloted Driving program. The key objective of this program is to allow drivers to hand over control of an Audi vehicle to a complex, context-aware onboard computer. Audi calls this computer a “digital pilot,” and the company envisions a future where a driver can readily switch between being highly involved in the driving experience, on the one hand, and allowing the digital pilot to take over completely, on the other. The congestion pilot, for example, is a modular component of Audi’s digital pilot that takes over control of your vehicle during a traffic jam. Until the congestion clears up, the system takes care of steering, acceleration and braking for you.
Take a step back and notice something about Audi’s congestion pilot: The functionality is incremental. You could build this system into a car without requiring that the driving environment change radically, and without reenvisioning the relationship between vehicles and the driving environment. After all, the human driver still retains the ability to step in at any time and take over all the car’s functions. The car is sophisticated extension of its human driver, not a fully-functional autonomous robot.
Google, then, does not want to build cars that are merely sophisticated extensions of human drivers. Quite the contrary, Google wants to build pure robocars — vehicles that replace human judgment with the razor-sharp logic of a hidden computer intelligence that is not influenced by joy, anger, alcohol or drugs. You might even exaggerate the difference between Audi’s strategy and Google’s a bit to bring the point into sharp relief: While Google is fascinated with replacing humans, Audi is more interested in a cybernetic approach — one that aids human beings without removing them from the driving equation completely.
None of the above implies that Google’s autonomy advances will not make their way into the autonomous cars of the future. As one influential Silicon Valley investor and entrepreneur put it, “it is unthinkable that Google will not significantly impact the FAVs that will one day effortlessly navigate our streets.”
But three things are clear. First, Audi’s incremental strategy is consistent with how every car manufacturer approaches integrating autonomous technology. Second, Audi’s success with incremental technologies provides proof that the incremental approach actually works. Lane awareness, automated parking, and cruise control are all examples of incremental advancement. Third, Audi appreciates, as do all car manufacturers, that car ownership is going to radically decrease once FAVs become the dominant mode of transportation. If cars are robots, why would anyone need to own one?
Known unknowns Reading the tea leaves, or predicting the future, is almost impossible if your goal is to identify exactly how the future will turn out and precisely who the winners will be. One must be content to map existing trends to future possibilities without falling prey to the temptation to manufacture certainty around particular outcomes.
With this caveat in mind, consider the following. Fully autonomous vehicles (FAVs) will be widely available only when two things are accomplished. First, FAVs must be safe enough to support widespread public use. And second, the incumbents in the transportation space that have flourished under the gas-car ownership model will be forced, at a minimum, to see their revenue streams radically diminish.
There are are number of things we know we don’t know right now. Although companies like Google have been refining FAVs’ capacity to shuttle consumers around safely, serious steps to begin the regulatory and economic battle against incumbent interests have not yet begun. We do not know how long that battle will take for FAVs to become the default transpiration technology. But it is clear that very large incumbents — for example, the $198 billion dollar automobile insurance market — is not going to be happy with the arrival of robotic cars that don’t make mistakes.
Uber’s collaborative strategy Uber recently announced that it formed a partnership with Carnegie Mellon University (CMU) to create the Uber Advanced Technologies Center in Pittsburg. The Center will focus on developing new transportation technologies that will help Uber provide cutting edge solutions to customers. In addition, Uber will be facilitating collaboration between CMU and the National Robotics Engineering Center (NREC) on developing mapping and autonomous vehicle technology.
Although Uber has not publicly shared its approach to developing mapping, autonomy and safety technologies, its near and long term strategic interests favor the incremental approach.
Consider what Uber needs to accomplish in the near term. Besieged in 2014 and early 2015 with regulatory challenges and lawsuits, the company has a clear interest in making its ride-sharing service safer. These regulatory challenges and lawsuits have a common lesson: Uber has to make its ride-sharing service as safe and reliable as possible in every world city it serves. Because these markets aren’t homogenous — their laws, technology infrastructure and insurance markets all differ — Uber has to find ways to increase safety in all of them even though it controls none of them. That, my friends, is precisely why mapping, safety and autonomy comprise such a critical piece of Uber’s future. This technology trio is something Uber can exert concrete influence over, and that’s precisely why the company partnered with CMU.
With respect to Uber’s long term interests, consider the company’s overall trajectory. Uber’s first act is replacing the byzantine taxi regime, but Travis Kalanick, Uber’s CEO, is clearly interested in delivering a powerful second act. It is during this performance that Uber will expand into transportation technology by aligning Uber’s interests with those of powerful car manufacturers. Those companies know that FAVs will reduce car ownership, so naturally they are exploring ways to introduce autonomy technology that will enhance safety and convenience but have the smallest possible impact on ownership. After all, in the United States alone almost 17 million new cars are sold every year. With smart dispatch software powering robocars, those numbers would collapse. Car manufacturers know this, and have strong incentives to protect the appeal of car ownership in some form.
The big car manufacturers are interested in developing autonomy technologies and integrating them into their newest models. Uber wants that as well, but it also needs to find a way of creating technology assets that will assist the company in its mission. Enter the partnership with CMU. The minute the partnership was forged, it changed the dynamic between Uber and its principal competitor, Lyft. Uber can now claim to be developing a set of forward-looking technologies than Lyft does not possess. That not only makes Lyft a weaker IPO candidate, it also makes it harder for Lyft to raise money on favorable terms. Through its partnership with CMU, Uber just radically reduced one existential worry.
Second, by partnering with CMU to create safety, mapping and autonomy technologies, Uber is able to move more quickly toward its goal of ensuring that its vehicles benefit from standardized safety technology. Exactly how that will happen is anyone’s guess. Uber and CMU could develop a competitor to the Cruise aftermarket autopilot kit, which turns ordinary Audis into semi-autonomous vehicles. Alternatively, Uber could work privately on better safety and auto-pilot technologies with the intention of providing a la carte access to car manufacturers. Whatever specific steps Uber takes, the objective will be to help the company carefully manage a transition from a world where cars are shiny consumption goods to a future where cars are merely utilities controlled through smartphones.
Third, this partnership allows Uber to think big about the future of semi- and fully-automated commercial transport. To take one speculative example, CMU, along with the National Robotics Engineering Center (NREC), could help Uber develop technologies to facilitate automated trucking and sea transport. Moreover, NREC is already developing expertise in robotics that may one day help enhance the safety of autonomous vehicles. For those who haven’t heard of NREC, it was recently in the news in connection with its co-development (with Boston Dynamics, a Google company) of a robotic dog that can adapt to environmental threats and follow a human leader through difficult, unknown terrain.
Robotics may seem irrelevant to Uber’s future. But that’s not true. Take collision avoidance, for example. A self-driving car cannot avoid accidents unless it can reflexively respond to unforeseeable dangers in the environment. For example, imagine it’s 2028 and you have been picked up by a FAV to be whisked from San Francisco to Palo Alto along the scenic 280 highway. As you’re speeding along at 65 mph, a rogue soccer ball leaps into the roadway immediately in front of you, and your FAV has to react. Abstractly, this is a common problem for robotics and autonomous systems: there will always be unforeseeable input. In the soccer ball case, the FAV has to decide what to do in the blink of an eye. Should it drive straight or swerve, accelerate, retain or reduce speed? NREC’s robotic dogs represent a bold attempt to address these very challenges: These robots actively test the environment, choose optimal routes to a destination, and adapt to a range of avoidable threats. The self-driving cars of the future will need to be able to do all these things.
Mr. Kalanick sees what Silicon Valley is unable to: Google is not the strongest ally in the robocar wars because superior technology will not be decisive. War, in fact, is not even an apt metaphor. Cooperation in a better way to look at this market: If the right parties cooperate, they may be able to balance (1) the desire to introduce new technology against (2) the negative economic impact of doing so on car manufacturers as well as the wider economy. Uber has done something extremely clever: It has decided to cooperate with the gatekeepers that matter most in an effort to protect and expand the strength of its core business.