By John Divine
With a new decade now upon us, the 2020s promise much in the way of innovation: 5G technology, advancements in artificial intelligence (AI) and virtual reality (VR), and, last but not least, autonomous vehicles (AVs). Self-driving cars.
The race for this technology is on. Silicon Valley, Detroit, China—it’s not just the U.S., but the entire developed world that’s hard at work on making this future a reality. The holy grail, specifically, is Level 5 automation: a car that needs no human monitoring to get from point A to point B, regardless of the road or its condition. These kinds of vehicles won’t even require steering wheels.
Level 4 automation isn’t too far behind that: these driverless cars can perform all driving functions as well, but only under certain conditions. Unique roadways may require human intervention, for instance.
The timeline for when these sorts of changes will come—and when government regulators will give them the green light—is still very much up in the air. But there are cultural, societal, and economic signals that the days of Level 4 and Level 5 automation may be closer than people think.
After all, multibillion-dollar companies like Uber and Lyft are built almost exclusively around the premise that the eventual rise of self-driving cars will virtually eliminate their biggest variable costs (i.e., labor), allowing these companies to enjoy ever-growing operating leverage as their scale grows. They simply weren’t designed to be economical without cutting out the driver.
Another sign that this looming step-change may come sooner than expected came in the form of a unique political candidate vying for the presidency in 2020. This year, for the first time ever, the U.S. saw a major-party candidate whose platform was built entirely around the threat that automation will pose to millions of blue-collar U.S. workers. Businessman Andrew Yang sprang from virtual anonymity to fight his way far beyond where pundits thought he was capable of going. One industry Yang mentioned again and again as being on the verge of massive changes was the trucking industry, which is one of the largest employers in the U.S.
It’s a matter of when, not if.
So, where are we in this arc? Is Level 5 automation really that attractive? And what will the eventual emergence of Level 5 automation mean for the insurance industry?
Level 5—Improvement, Not Panacea
One question readers might have is simply: “Why?” Why is Level 5 automation something we should strive for? Is this technology actually miles better than the current sophistication cars have?
The commercial incentives for Level 5 AV development readily spring to mind—aside from applications in ride-hailing and trucking, there’s market-shifting potential in areas like local food delivery and e-commerce.
These applications might enjoy newfound appreciation from governments and citizens across the globe in the wake of the COVID-19 pandemic, the disease caused by the novel coronavirus. Few technologies are better suited for limiting person-to-person interaction than fully autonomous vehicles. Imagine ordering groceries from Walmart and getting them delivered to your residence by a driverless car. Sounds preferable to anxiously braving the produce aisle in order to adequately stock up.
Productivity gains should also be meaningful, given no one would have to waste time in their day overseeing the machine shuttling them around—they could be working remotely or reading or, well, wasting their time, instead.
But the real “X-factor” that makes Level 5 automation so desirable is safety.
According to the National Highway Traffic Safety Administration (NHTSA), 94% of all serious auto accidents are the result of human error. In 2018, over 36,560 people were killed as a result of auto-related crashes in the U.S. Clearly there’s quite a bit of margin for improvement there, with the complete elimination of human error by means of a perfectly safe automated substitute saving over 34,000 lives annually.
Most AV operators and experts aren’t panglossian with their promises and projections. A 100% error-free AV system won’t likely be seen in the 2020s.
Which, as long as self-driving vehicles are safer than average human drivers, is still fine. The ultimate goal, after all, is to reduce human pain, suffering, and death by incorporating self-driving technology. To that end, the phrase “perfect is the enemy of the good” applies beautifully.
Like every other technology, progress in autonomous vehicles will come in stages rather than all at once. Fighting the conception that AV tech must be perfect or near it in order to be adopted at all is an important step that consumers, regulators, and stakeholders at all levels need to combat—sooner rather than later—in order to accelerate adoption and reach Level 5 automation (and the concordant safety benefits).
This isn’t a philosophical assertion; a late 2017 study conducted by the influential think tank RAND Corporation concluded that even AV technology that was 10% safer than the average human driver would save more lives if rolled out immediately than waiting until such systems were 75% or 90% safer before rolling them out.
In modeling the potential life-saving impact of various strategies, RAND termed the 10%, 75%, and 90% improvement rates Improve10, Improve75, and Improve90, respectively. Here’s an excerpt from the RAND study that discussed the policy implications. It refers to highly autonomous vehicles as “HAVs”:
“In the short term (within 15 years), more lives are cumulatively saved under a more permissive policy (Improve10) than stricter policies requiring greater safety advancements (Improve75 or Improve90) in nearly all conditions, and those savings can be significant—hundreds of thousands of lives. The savings are largest when HAVs under Improve10 are adopted quickly.”
The longer-term benefit of rolling out 10% safer autonomous vehicles was even more staggering:
“In the long term (within 30 years), more lives are cumulatively saved under an Improve10 policy than either Improve75 or Improve90 policies under all combinations of conditions we explored. Those savings can be even larger—in many cases, more than half a million lives.”
A couple of important factors contribute to this stark difference in expected fatalities under the various scenarios. The most intuitive one is the fact that it will take time to develop a system that can improve from a 10% reduction in fatalities to a 75% reduction.
Over that time, the number of HAVs on the roads will gradually increase, and the growth in adoption will help accelerate the total number of HAV miles driven. The degree to which rolling out autonomous vehicles commercially will reduce accidents and casualties is highly dependent on HAVs gaining “share” of total miles driven.
Waiting for a more perfect solution, RAND concludes, is counterproductive largely because of how delaying the rollout will delay HAV adoption, and thus HAV miles driven will experience a much slower ramp-up.
It’s also important to remember that after rolling out 10% safer AV technology, the safety profile will still continue to improve over time, and in fact should improve at a higher rate due to the miles being driven and the data being amassed.
In that same vein, over-the-air (OTA) software updates of the sort that Tesla began building into its vehicles offer extremely important life-saving potential as well. If you can improve the safety profile to an entire fleet of vehicles via an overnight software upgrade—instead of requiring consumers to buy entirely new cars with each improvement—these improvements, even if incremental, can help spur the age of Level 5 autonomy on an accelerated timeline.
Level 4 and Level 5—Insurance Implications
It’s clear that Level 4 and Level 5 autonomy will change car insurance markets; when technology reaches the stage where cars can drive themselves with negligible human intervention—or none at all—there should be a fundamental shift in how liability is dealt with.
But because that evolution will take place incrementally, it may also take some time to see major changes.
While Level 5 AV adoption—and even widespread Level 4 automation—may not come in the 2020s, the rise of earlier-level autonomy is already seeing corresponding changes in the field of insurance.
It’s not that these early changes are fundamentally reshaping industry economics. They’re not—and for the largest players, these shifts may not even seem material.
Still, cracks are starting to materialize. Specifically, major insurers have been reluctant to insure the most highly autonomous street-legal vehicles available to the public. Perhaps that’s unfair: They will insure them, it’ll just cost the policyholder an arm and a leg.
In some cases this has presented smaller players and upstarts with an opportunity to develop a niche business. One harrowing tale relayed by Bloomberg in 2019 shares the story of Dan Peate, who tried to insure his Tesla Model X, but was quoted $10,000 a year, so he promptly started his own insurance company to address that market inefficiency.
Arguably larger long-term threats to industry mainstays, however, are the autonomous vehicle manufacturers themselves. Some, most notably Tesla, have already started offering their own insurance policies to their customers.
In 2019, Tesla made good on one of CEO Elon Musk’s many promises when it began offering Tesla drivers in California discount auto insurance under its newly spun up Tesla Insurance arm.
“Tesla Insurance is a competitively priced insurance offering designed to provide Tesla vehicle owners with up to 20% lower rates, and in some cases, as much as 30%,” the automaker’s site reads. “Tesla Insurance offers comprehensive coverage and claims management to support Tesla owners in California and will expand to additional U.S. states in the future.”
The company explains that its firsthand knowledge of its vehicles, their safety profiles, technology, and repair costs give it a unique informational advantage allowing the company to price risk more effectively. Additionally, “Tesla Insurance pricing is reflective of Tesla’s active safety and advanced driver assistance features, which come standard on all new Tesla vehicles.”
The electric automaker currently uses aggregated, anonymous driver data to inform its pricing, and while it specifically says it doesn’t currently use vehicle-specific data that any given car’s GPS or camera footage could provide, Tesla plans “to expand the product offerings to incorporate more types of data over time.”
It makes sense. Why wouldn’t a company like Tesla get into this business? In Tesla’s case, a recent examination of company filings and disclosures conducted by MIT research scientist Lex Fridman revealed that over 737,000 Tesla cars equipped with Tesla Autopilot Hardware 2 or 3 are already on the road.
More illuminating are Fridman’s calculations on total Autopilot miles driven. As of early 2020, Fridman estimates over 2.2 billion Autopilot miles have been driven overall.
It’s hard to overstate the value of the data that was generated and collected over those 2.2 billion miles. It’s rich, proprietary, unique—and wildly expensive and time-consuming to replicate.
That’s all well and good, but if there’s not a clear way to monetize that data, then it’s just another statistic. That’s not an issue in this case: Tesla primarily uses the insights it gleans to improve its self-driving technology and make its vehicles safer, which dramatically increases the value proposition to consumers. Long-term plans for a Tesla self-driving taxi fleet are also potentially worth untold billions.
“When one vehicle learns something, they all learn it,” Musk has said.
Setting other applications aside, the value of this real-world dataset—which according to Fridman’s estimates also included roughly 17 billion miles of driving without Autopilot—in insurance markets is plain.
Aside from the clear pricing advantages gained from Tesla’s substantial informational edge, Tesla enjoys an elite brand and the ability to offer policies at the point of sale to its famously devoted customer base. The Tesla Insurance website already claims that California drivers can purchase policies on its site in as little as one minute. It may not be long before other ambitious AV manufacturers follow along.
This sort of competitive threat won’t materially ding the bottom line of companies like Geico, Allstate, Progressive, or Nationwide in the near future. But the first-mover advantage (and the growing treasure trove of first-party data) that a company like Tesla could opportunistically exploit does pose a possibility of making the manufacturer-insurer business model an intimidating force to be reckoned with long term.
The Liability Question and Rate Impact
As self-driving technology becomes more and more advanced, the central question surrounding higher levels of autonomous driving looks to be the liability question. If a human driver isn’t operating a self-driving car and the vehicle crashes or hurts a third party, who is responsible?
This question has yet to be satisfactorily resolved in the U.S., and will remain a major barrier to not just mainstream adoption, but rapid innovation, until a framework of divvying up liability among the manufacturer, operating system developer, and the individual is ironed out.
Canada, which is one of the world’s leaders in AV innovation, may serve as an interesting control for what works and what doesn’t as the U.S. seeks to navigate these waters. Like the U.S., Canada is a country composed of a number of smaller geographic regions—territories and provinces rather than states—which each have particular rules.
Daniel Gardner is a law professor at Université Laval in Québec.
Similar to the U.S., Gardner told me, “Every province and territory in Canada has its own rules. For instance, public insurance companies—working as a monopoly—are in place in four provinces (Quebec, Manitoba, Saskatchewan, and British Columbia) but the system in place varies from pure no-fault to a system where some civil suits are still authorized.
“Other provinces operate with private insurers, each of them with its own rules,” says Gardner. Quebec’s system, which utilizes a public insurer, covers all Quebecers around the world, “regardless of who is at fault,” Gardner notes, citing Section Five of the Automobile Insurance Act.
That makes things doubly simple for Gardner and other Quebecers when AV technology eventually rolls around in earnest. Because of its no-fault system, he says, “in Quebec, nothing will change with the introduction of AVs.”
Even before then, however, Quebecers enjoy a simplified and low-cost auto insurance ecosystem. “By far, we have the lowest auto premiums in North America because we operate a pure no-fault system,” Gardner says.
The numbers back that story up. A 2018 report from the Insurance Bureau of Canada found that Quebec had easily the lowest car insurance premiums of any province or territory in the country, with annual premiums clocking in at $717. The next highest? Prince Edward Island at $816. In British Columbia, the rates average more than $1,800, so the numbers vary widely depending on the region.
And while car insurance premiums in the U.S. also vary widely—annual ranges clock in between $850 on the low end (Maine) to over $2,600 (Michigan)—America’s litigiousness plays a large role in making the average annual premium in the U.S. about $1,650 annually, a higher average than nine of Canada’s 10 provinces.
This litigiousness is simply a fact of life in today’s system. But if the U.S. wants to continue to be on the cutting edge in the development of such a paradigm-shifting technology, it will need to confront the liability problem head-on and develop specific federal legislation to deal with this issue.
As it stands now, AV manufacturers face the risk of costly lawsuit after costly lawsuit every time someone gets hurt while using higher-level autonomous driving features. Tesla has already faced several lawsuits stemming from fatal accidents that occurred while Autopilot was enabled.
The motive to sue the companies behind the AV hardware and software also causes problems. You can go after an individual, but the amounts recoverable from doing so pale in comparison to the deep pockets of a Tesla, Alphabet’s Waymo, or GM’s Cruise Automation.
As it currently stands, it won’t matter if, for instance, Tesla’s Autopilot proves to be much safer than human-driven vehicles: They’ll be hearing from lawyers every time they’re not perfect. For what it’s worth, Tesla’s self-reported numbers certainly seem to paint the picture of an impressive safety profile: In Q4 2019, there was one accident for every 3.07 million miles driven using Autopilot. With Autopilot turned off but other active safety features on, there was still just one crash per 2.1 million miles, and accidents occurred once every 1.64 million miles when both were turned off. For comparison, Tesla cites NHTSA data of the U.S. average, which is one car crash per 479,000 miles.
It should be noted that critics have taken issue with Tesla’s juxtaposition of these numbers, which compare newer Teslas with a much older average vehicle fleet for the entire U.S., and which list Autopilot stats that largely consist of highway driving.
Regardless of whether it’s true today that commercially available AV-enabled vehicles are safer than human drivers, the problem really lies in the hypothetical: Even if AVs were unequivocally safer than humans, the threat of litigation for being anything less than perfect may justifiably disincentivize the commercial rollout of those vehicles.
RAND’s research has already emphasized the enormous potential human cost to rejecting incrementalism and waiting for something dramatically safer than human drivers before going to market. In missing the forest for the trees, hundreds of thousands of American lives could be needlessly lost over the long term. And RAND’s paper doesn’t even cover the economic opportunity cost faced by the U.S. if such reticence allows other eager innovators like Singapore, Germany, China, or Canada to take the lead.
A forthcoming paper from Anthony Paolino III in the University of Arizona Law Journal of Emerging Technologies proposes a solution that directly addresses the problems posed by mainstreaming new AV technology.
From the abstract:
“During the period of trial and error, I propose a federally mandated insurance policy to indemnify injured passengers and third parties on a no fault system, so long as they can prove involvement in a self-driving car accident. Self-driving car companies will fund this policy. Structured as a compromise, like workers’ compensation, those companies will have to pay a lower amount of damages per injury, only the plaintiff’s economic damages. The policy will help avoid the litigation process, guarantee that every injured person gets something, and facilitate the long term benefits of a likely usefully technology.”
As for how autonomous vehicles will impact car insurance premiums, the overall consensus seems to be: negatively. On a high level, increased visibility into granular driving data should produce lower margins as companies with access to such data drive down prices. Perhaps more to the point, the biggest ethical argument for AVs—their improved safety profile—should substantially reduce the number of accidents and fatalities and meaningfully drive down the volume of claims.
With less risk to insure, the car insurance market should contract.
One 2017 paper from the Bank of England forecast a 21% reduction in the U.K. motor insurance market by 2040 with the steady rollout of AV technology. It also highlighted the increasing need for insurers to forge partnerships with manufacturers and tech companies.
The good news for actuaries in this space is that, new decade aside, the car insurance industry seems destined to change more gradually than the underlying technology it’s insuring. And the growing need for new products and partnerships should create new opportunities in itself.
Another difficulty is establishing a satisfactory confidence interval around outcomes. Tesla’s estimated 2.2 billion Autopilot miles sounds impressive at first glance, but it doesn’t hold a candle to the 17 billion non-Autopilot miles Tesla vehicles are estimated to have racked up, nor the untold billions of miles that Big Three automakers and other majors have under their belt. These data are based on driving distances orders of magnitude larger than what Tesla has to work with.
Mobile apps that the biggest insurers now have, like Progressive’s Snapshot, allow companies to collect driver data without the expensive hardware and technology investments—and still reward customers for safe driving behavior.
Still, even if we envision a rosy future with far fewer accidents over time as safer self-driving technology filters out to the masses, insurers will need to build a higher level of certainty around outcomes before premiums can fall as swiftly as claim amounts.
Another challenge: the new and entirely unprecedented risk of low-frequency, high-severity events that may present themselves with the threat of a mass hack or operating system failure. Actuaries really have no historical data to work with in this regard, and if the early months of 2020 have taught us anything, it’s that so-called Black Swan events simply can’t be responsibly ignored. Insurers that want to stick around for the long run will be forced to build in a material margin of safety into their rates, or pawn these large, remote risks off on reinsurance companies and pass those costs on to the customer instead.
In short, there are some amazing things happening with AVs right now, and the 2020s are bound to be a decade of enormous innovation and progress. It’s not often that a technology with such profound potential ramifications on public health and productivity comes around. Once policymakers settle on what a tolerable safety profile is and how to prove it, the runway will be clear of debris and the commercialization of highly autonomous vehicles can begin in earnest.
Insurers are unlikely to see real near-term pain at the hands of this innovation. It’ll take time before the market is saturated with self-driving cars, and the industry should have plenty of time to learn how to price the risk of these vehicles more accurately, competitively, and responsibly in the meantime. The biggest risk for insurers, perhaps, is not a long-term decline in premiums, but the loss of market share at the hands of nontraditional insurers, upstarts and the do-it-yourself manufacturer-insurer.
JOHN DIVINE is a Washington, D.C.-based freelance writer. He previously appeared in Contingencies with “The Game Behind the Game,” about the National Football League and advanced sabermetrics.
 “Automated Vehicles for Safety”; National Highway Traffic Safety Administration.
 “The Enemy of Good: Estimating the Cost of Waiting for Nearly Perfect Automated Vehicles”; Rand Corporation; 2017.
 “Self-Driving Cars Might Kill Auto Insurance as We Know It”; Bloomberg; February 19, 2019.
 “Support–Tesla Insurance”; Tesla.
 “Over 730,000 Tesla Vehicles with Autopilot 2 & 3 on the Road”; Clean Technica; January 5, 2020.
 “Tesla Vehicle Deliveries and Autopilot Mileage Statistics–Sources”; Lex Fridman.
 “Tesla, the data company”; CIO; August 28, 2019.
 “Support–Tesla Insurance”; Tesla.
 “Car Insurance Rates Across Canada: Who’s Paying the Most and Why?”; Canada Drives; February 12, 2020.
 “What Is The Average Cost of Car Insurance?”; The Simple Dollar; March 23, 2020.
 “Average Cost of Car Insurance in 2020”; The Zebra; March 11, 2020. “What Is The Average Cost of Car Insurance?”; The Simple Dollar; March 23, 2020.
 Op. cit.; The Simple Dollar.
 “Tesla hit with another lawsuit over a fatal Autopilot crash”; The Verge; August 1, 2019.
 “Tesla Vehicle Safety Report”; Tesla.
 “The Ultimate Insurance Policy: Autonomous Vehicles and Artificial Intelligence, a Statutory Proposal for a Complicated Product”; University of Arizona Law Journal of Emerging Technologies; November 6, 2018.
 “Quarterly Bulletin 2017 Q1–Potential impacts of autonomous vehicles on the UK insurance sector”; Bank of England; 2017.
Levels of Autonomous Driving
SAE International, once known as the Society of Automotive Engineers but now a trade group serving various technical professionals, has developed a set of international standards for autonomous vehicle systems, and the U.S. Department of Transportation adopted these standards in its “Federal Automated Vehicles Policy” in September 2016.
This common framework allows consumers, automakers, and regulators to have a shared group of assumptions about how much a vehicle can do on its own and how much responsibility the driver has behind the wheel.
Here are the official levels that define how autonomous a vehicle is:
Level 0—No Automation
The federal policy defines this simply as “the full-time performance by the human driver of all aspects of the dynamic driving task, even when enhanced by warning or intervention systems.” Drivers do it all on their own.
Level 1—Driver Assistance
The driver is still almost entirely in control, but specific tasks using specific information can assist them. The functions—steering assistance or acceleration controls, for example—are limited in scope, and a human at the wheel is still necessary to deploy them.
Level 2—Partial Automation
Much of “dynamic driving” based on changeable elements is still in the hands of the person holding the steering wheel. However, the vehicle itself can finally start to take action on its own, such as steering back to the center of the lane or maintaining and adjusting speed based on traffic. In some circumstances, the driver is disengaged from both the wheel and the foot pedals at the same time. Certain circumstances will allow the driver to quickly take back control, of course, but automated driving is possible for a short period.
Level 3—Conditional Automation
This is the step change between what is commonly viewed as driver assistance and the world of truly autonomous driving. In Level 3, driving at speed is fully automated and the system begins to monitor the external environment and act accordingly. The vehicle itself still has limits, however, but now is aware of those limits and will ask for human assistance when situations dictate. The rest of the time, it takes care of the driving on its own.
Level 4—High Automation
The difference between Level 3 and Level 4 is that the autonomous vehicle may still encounter dynamic driving situations that do not have clear or obvious solutions but will be able to respond reasonably appropriately even without human intervention. Whereas a vehicle with conditional automation has a human fallback, theoretically a car with Level 4 automation can perform all aspects of driving, even when the unexpected occurs—but only “in certain environments and under certain conditions.” In other words, even if highway driving is fully automated, there may still be some surface streets with unconventional traffic patterns that require a human hand at the wheel.
Level 5—Full Automation
This is the dream: a vehicle that has full-time automation, responding to all aspects of the road environment in real time in a way that is perhaps even more effective than a person (who is prone to human error or distraction).
Moving at the Speed of Innovation
A conversation with AV advocate Greg Rogers on the state of play
and where America is on the AV timeline
Greg Rogers is director of government affairs and mobility innovation at Securing America’s Future Energy (SAFE), a nonpartisan, nonprofit advocacy group focused on reducing America’s oil dependence. He focuses on electric and autonomous vehicles.
Uncertainty over regulation, and liability in particular, is an obstacle to innovation in AVs. What needs to be done, and how close is the U.S. to passing legislation reducing this uncertainty? Is there anything in the works?
AVs present a number of questions that our existing regulatory structure can’t answer yet.
All the rules were written for human-driven vehicles and how to assign human fault. Now we’re talking about a vehicle perceiving the world around it and how the vehicle makes decisions.
AVs may be able to make mathematical and accurate decisions that can save lives and do so faster than humans can, and there’s clearly a great benefit there.
The challenge is that regulation doesn’t move at the speed of innovation. We’ve been working on a bipartisan bill being written by Republicans and Democrats in the House and Senate that would create a flexible framework for how AVs are regulated in the U.S.
Does it have a name?
No, it hasn’t been introduced yet. House and Senate staff have been talking with stakeholders: automotive companies, insurance companies, safety advocates, people with disabilities and everyone else across the field on how to write it.
It’s pretty impressive how bipartisan it’s been; if the coronavirus hadn’t happened we might’ve seen the full bill come out [in April].
Where is the U.S. in terms of the global autonomous vehicle race?
It’s closer to the Olympics than a race. There’s a global effort to reduce fatalities. There are just under 40,000 auto-related fatalities in the U.S. annually. About a million die each year across the world. It’s a senseless loss of life. So there’s a global effort, of course.
Nationally speaking, we’ve been the leaders in AV development—and we should keep leading in AV.
What are the major issues that must be ironed out for AVs on the regulatory side, and what has draft legislation on this issue attempted to address?
No. 1: We need a consistent regulatory framework. You can drive a car from New York to California without stopping at every single state line to get your car inspected because we have one national framework for regulating vehicles. The federal government needs to be sole regulator of autonomous vehicles. States have written this patchwork regulation; we risk each state basically acting as its own country when it comes to AV laws.
No. 2: We need to have timelines for the federal government to write standards for AVs. We need clear goalposts for when we’ll see regulations in place. We need to work with stakeholders on what those standards should be and make sure they’re written with the evolution of that technology in mind.
Lastly, we need to allow for vehicles with different designs. Today’s standards were written around human-driven vehicles. So it’s been difficult for companies trying to release vehicles that don’t actually need steering wheels and brake pedals because they’re driving themselves. But it’s illegal not to have them. Our system needs to allow for the development of these smaller, more efficient driverless delivery vehicles. Imagine how useful contactless, autonomous delivery would be today, during this pandemic.
Take this company called Nuro, which has autonomous delivery robots. They just got an exemption from NHTSA that would allow them to do this. But they applied to put these little driverless delivery vehicles on the road two years ago. It took the Department of Transportation two years to review their application.
Technology doesn’t move at that speed. We should be able to put these vehicles on the road far sooner. Imagine what could be put on the roads today if these are from two years ago. Even then, NHTSA was only allowed to give exemptions for 5,000 vehicles. That’s not the scale the auto industry works at.
Then take a look at Neolix in China: they’re basically the Chinese version of Nuro. When COVID-19 struck China, the Chinese government already knew that there was this benefit to contactless delivery for medical supplies and food.
So China expedited regulatory approval for Neolix and provided government incentives to get as many of them on the road as possible. They were completing deliveries and helping to reduce people’s exposure to coronavirus.
Level 5 is considered the ultimate goal in autonomy. Where are we in that process?
A Level 5 vehicle can drive on any road, in any part of the world, at any time, without any human guidance.
The real question isn’t so much Level 5 as much as a wider range of Level 4. Level 4 means you can operate in designated environments safely without a human driver. It’s more important to focus not on when, but where AVs will happen.
We’ll see autonomy appear in more places over the next three to four years. But it will be some time before an AV can drive coast-to-coast on whatever route it wants.
Well, so what level are we at now? And when do we move to the next one?
GM has its Super Cruise automated driving system. That’s Level 2, and that’s in some new vehicles on the market today.
Tesla Autopilot is categorized as Level 2. In these versions, humans must have eyes on the road and hands on the wheel at all times.
We’ll see more Level 3 systems coming out in next few years. Drivers will cede more control to the automated system.
But Level 4—where you have a robot driving you—you can call a fully driverless vehicle through Waymo in Arizona right now. Waymo is operating fully driverless vehicles that people can order for ride-hailing.
So it’s just a matter of expanding their capabilities on different roads and under different environmental stimuli?
Exactly. That’s why Waymo is testing in Arizona, because it has all the features that make it easier to drive in. Wide roads, lane markings are good, signage is clear, and there are predictable roadways. Having a more expansive driverless service, though—that’s the next step in this progression, and it’s going to take more time to see more and more fully driverless services across the country.