Silicon Valley has thrown its technological and deal-making weight behind getting humans away from the steering wheel, but will have to do what tech companies are typically least equipped for before real progress in driverless cars: wait.
That’s because it is stubbornly hard to get artificial intelligence, a key component of driverless-car technology, mighty enough to parse out every situation that may arise on the road, to say nothing of technical difficulties that still prevent much more than equipping our priciest vehicles with advanced driver-assistance systems. Moreover, the road to driverless cars will likely be littered with failed business partnerships and the need to compromise with Washington, D.C., as well as rife with consolidation and deals in the coming years.
In recent weeks, the association between car maker Tesla Inc. TSLA, -1.78% and hardware company Nvidia Corp. NVDA, -0.14% one of the most powerful in the Valley, has given signs it may have run its course, a key development to watch in 2018.
Artificial intelligence’s algorithms and technologies are changing our everyday lives, and AI likely won’t alter anything as soon and as thoroughly as it will alter how we move people and goods from point A to point B.
While machine learning is already weaving its way into many of the technologies humans use regularly, autonomous automobiles will be the biggest turning point for the technology. For most of us, driverless cars will be gateway AI, one of the first domains in which humans will be asked to trust the reliability and safety of an AI system for a very complex task.
“Autonomous transportation will soon be commonplace and, as most people’s first experience with physically embodied AI systems, will strongly influence the public’s perception of AI,” according to a 2016 report on long-term impact of AI by academics and industry observers.
For investors, the question is not if, but when.
For now, those looking to bet on the space can put money on household names such as auto parts maker Delphi Technologies PLC. DLPH, +1.71% as well as Nvidia and Tesla, to name three companies that Wall Street widely holds as the front-runners at the intersection of AI and driverless cars.
For the future, dozens of startups globally are vying to become the next publicly traded giant of autonomous-car technologies, which means waves of consolidation will occur until one or two dominant companies emerge.
And that’s to say nothing about players yet to be launched to serve markets yet to surface, and that today may sound as far-fetched as the Jetsons’ Rosie the Robot Maid did decades ago.
Deals, deals, deals
Take Tesla, for example: it ditched advanced driver assistance systems maker Mobileye NV in 2016 as the two sparred over Autopilot, the car maker’s suite of ADAS. Now, Tesla has signaled it is considering severing its relationship with Nvidia to take on chip-making in house.
Other companies may take a page out of Intel Corp.’s INTC, +0.07% playbook and look at leapfrogging the competition through an acquisition. Intel bought Mobileye in March for more than $15 billion, a deal hailed as Silicon Valley’s largest pure bet on driverless cars and one that vaulted Intel’s position in vision-based ADAS.
In 2016, General Motors Co. GM, -1.17% bought Cruise Automation for $1 billion, and Cruise, now part of GM, recently demonstrated its driverless-car technology on San Francisco streets. GM has promised driverless taxis by 2019.
Later that same year, Google parent Alphabet Inc. GOOG, -0.70% spun its driverless-car efforts into its own business unit, which it called Waymo.
Ride-hailing companies such as Uber Technologies Inc. and Lyft Inc. are racing to offer “robo taxis,” and Uber and Waymo have been involved in a bitter litigation, in which Waymo seeks billion-dollar damages from Uber for alleged cloak-and-dagger stealing of trade secrets related to Waymo’s self-driving program.
Moreover, virtually all auto makers are pursuing driverless capabilities, often pairing such goals with producing more electric cars. Ford Motor Co. F, -0.79% has promised a fully self-driving car by 2021, and has teamed up with Lyft. Uber has paired off with Volvo Cars, owned by China’s Zhejiang Geely Holding Group Co. GELYF, +0.91%
Toyota Motor Co. 7203, -0.19% expects to test driverless cars, equipped with its virtual assistant named Yui, on the roads in 2020, and its Toyota Research Institute has teamed up with several companies to explore the use of blockchain technology, of bitcoin fame, on driverless cars. Tesla has promised a Los Angeles-New York City driverless ride at some point in 2018.
The deal-making and consolidation wave in driverless-car technology could mirror how financial technology whittled itself down a few years ago, with fintech-focused startups folded into major financial companies and a couple of sole players emerging, said Anand Rao, a consultant with PwC.
Valley optimism and on-the-ground realities
To ask Silicon Valley, with its deeply ingrained optimism about technology, the bulk of the tech needed for driverless cars is available today, and driverless cars will be turning a corner tomorrow.
While it may be the case that the technology itself will arrive first, we’d still need acceptance from individuals, from society and from regulators, and currently we are building that acceptance, Rao said.
Full, disruptive autonomy is likely further away than most may think, analysts at Evercore ISI said in a note earlier this year. It is a “science project,” currently, albeit one that “the greatest minds globally are tackling at a feverish pace,” the analysts said.
The analysts have singled out Delphi and Nvidia as the top companies to target for investment in the space.
The first truly driverless cars will likely be robo taxis, geofenced or physically confined to limited areas such as a campus, a business complex, or a downtown loop, until their use is more widespread and people start to get comfortable.
Driverless ride-hailing services could be offered first because they offer “a gentler slope toward autonomy,” and flexibility, said Jim Adler, managing director of Toyota AI Ventures and a vice president at Toyota Research Institute. If the weather is good, if the route is clear, a driverless car would be sent; if certain conditions are not met, then a car with a driver would be sent. In the early stages, only relatively few rides would be performed by driverless cars, Adler said.
And for years to come driverless cars will have to coexist with “traditional” cars, a transition period estimated to last at least a decade, PwC’s Rao said.
It takes about 11 years in the U.S. to turn over the fleet, so even if regulators were to mandate all cars to be autonomous from this day forward, it would still take at least that long to start moving toward predominantly driverless cars on U.S. roads, he said.
In addition, people who might be fine with summoning a “robo taxi” on occasion might still want to drive their private car on the weekends, for example, and that’s not to mention die-hard stragglers who will keep craving the driver’s seat. We’d still be around 15 to 20 years away from a time where most cars on the road would be autonomous, and transportation viewed as a basic service, Rao said. That time frame would vary from city to city, he said.
This summer, German regulators came out with guidelines from autonomous vehicles, trying to establish ethics standards for AI and driverless cars. Among 20 basic principles is the precept that the protection of human life enjoys top priority, and that systems must be programmed to accept damage to animals or property if personal injury can be prevented.
Dilemmas between one human life over another cannot be standardized, or programmed, and any distinction based on “personal features,” such as age and gender, is prohibited.
The Moral Machine, an online platform, asks people from all around the world to make judgment calls mostly involving hypothetical driverless cars, in order to gather human perspectives on “lesser of two evils” scenarios.
New business models, job roles emerging
New job roles could also emerge, such as people tasked with managing autonomous traffic in a way that would mirror air-traffic controllers. Beyond software companies, service and logistics companies could get into the mix as well, Rao said.
Another business model could render driverless cars as rental properties, with owners putting up cash for the cars and having companies or individuals manage the asset for them, Toyota Research Institute’s Adler said.
Level 5 automation—in which autos will take care of all the tasks of driving in every situation—is still a long way off. For an example of why, look at two often-talked up sensors at the core of driverless technology: lidar and cameras.
Lidar doesn’t see through snow, or through steam coming off a manhole; glare and certain long-distance conditions can fool cameras, Adler said.
“The sensors need to get better,” he said. Even if one assumes a perfect perception of the environment, cars will have to predict and understand a complex web of interactions and variables.
Examples abound: is the human with one arm up a police officer, in which case the gesture would quash a green-means-go rule, or someone pretending to be a traffic cop? Will the pedestrian at the corner obey walk/don’t walk signs? Is the person on an unicycle doing a stunt, or ready to move forward?
Society will have to grapple with how safe is safe enough for driverless cars. Many in the industry speak of “10 9s” — 99.9999999999% accuracy needed to move toward commercially viable autonomy, the Evercore ISI analysts said.
Throngs of data and simulation, not just driving around, are going to play a big role in exposing such situations and teaching cars to act like a social being, Adler said.
“We don’t know how it will play out,” he said. “We will benefit from the successes and learn from the mistakes.”