"We cannot do it."
That's all an engineer from one major Japanese automaker could say after seeing the central control unit for Tesla's (TSLA 4.96%) Autopilot technology, extracted during Nikkei Business Publications' recent teardown (an investigative disassembling of tech hardware) of a Model 3.
Japanese automakers might not be able to match Tesla's self-driving technological achievements, but that doesn't make Tesla the leader by default.
Other companies, most notably Alphabet's (GOOG -1.96%) (GOOGL -1.97%) Waymo unit, have been hard at work on self-driving vehicle technology for years.
And while Elon Musk may have the world's attention, Tesla simply doesn't possess as many resources to throw at the self-driving problem as the techies in Mountain View, California. Instead, for what Tesla's lacked in resources, it's made up for in aggression: Autopilot was the first technology advertising "self-driving" available to the public.
This early push has given Tesla more visibility, and access to more real-world data, than its competitors, but it's also exposed Tesla drivers to the risks of using a work-in-progress system. Will that be enough to keep it ahead, particularly as Autopilot-involved accidents continue to crop up in the news?
Over the long run, the real question won't be "Who can't do this?" but "Who will do this successfully and safely enough to lead the market?"
Let's see how Tesla's track record stacks up against two of the most notable self-driving developers: Waymo and Uber (UBER -0.07%).
Tesla Autopilot sensors visualized. Source: Tesla.
Which contender has the most real-world self-driving experience?
Tesla last announced its owners' cumulative Autopilot usage in November 2018 with a tweet celebrating one billion Autopilot miles. Extrapolating from available data, technology researcher Lex Fridman estimated Tesla owners had driven 2.2 billion total miles with Autopilot by the end of 2019.
But Autopilot isn't truly autonomous yet. Tesla's Autopilot site notes only that the technology is "designed to assist you with the most burdensome parts of driving," such as steering, braking, and accelerating within traffic lanes. "Current Autopilot features," the site says, "do not make the vehicle autonomous."
By making incomplete automation available to the buying public, Tesla seems to be treating its customers as beta testers, particularly since Musk has for some time implied a fully hands-free Autopilot is right around the corner.
This matters a great deal, because the perceived safety of self-driving vehicles is a critical obstacle to widespread acceptance of the technology on public roads.
In January, Waymo announced its autonomous cars had accumulated 20 million miles of total real-world experience, up from 10 million total miles reported six months earlier.
Some residents of Phoenix, as well as a few Silicon Valley-based Google employees, have been using Waymo's autonomous ride-hailing service since late 2018.
In 2019, The Information evaluated 10,500 Waymo rides, finding that 30% received negative reviews because of awkward drop-off points, erratic driving, or inefficient routes.
Meanwhile, Uber reported 2 million miles driven at the end of 2017. Its progress slowed significantly in 2018 after an autonomous Uber hit and killed a pedestrian. The company only gained self-driving access to California's roads earlier this month, making it the 66th autonomous-vehicle developer to win the state's approval.
Which contender is best protecting the public?
Self-driving cars are designed to be safer than human drivers. When something goes wrong, it's usually a big story.
Uber's fatal strike in 2018 was devastating in many ways, but it wasn't the company's first accident. A National Transportation Safety Board review found Uber's autonomous fleet to be involved in 37(!) non-fatal crashes during the 18 months prior to the deadly accident, 33 of which involved another vehicle hitting Uber's testing cars.
Whatever Uber poached from Waymo doesn't seem to include safety precautions. The unit reported a total of 30 minor accidents between 2009 and 2018, only one in which its autonomous tech was undoubtedly at fault.
Waymo-equipped Chrysler Pacifica. Source: Waymo.
Rather than report on total incidents, Tesla tracks and reports Autopilot-related accidents on a per-miles-driven basis each quarter.
In the fourth quarter of 2019, Tesla owners had one accident per 3.1 million miles with Autopilot engaged; one per 2.1 million miles while using Tesla's standard active safety features; and one per 1.6 million miles when not using safety features.
There have also been at least five reported fatalities blamed on Autopilot-engaged Teslas since 2016. TeslaDeaths, which tracks all deadly incidents involving Teslas, has recorded 119 total deaths to date. The last deadly accident in which Autopilot was known to have been engaged was a Florida crash last March. As authorities complete lengthy investigations into other crashes, we may see more accidents and deaths connected to faulty Autopilots.
Five deaths over 2.2 billion miles works out to a fatality rate of about 0.2 per 100 million miles. The overall auto accident fatality rate in the U.S. is 1.13 per 100 million miles.
On the surface, this seems like a great improvement over the baseline, and it's markedly better than Uber's track record of one fatality in less than 5 million miles. However, as the most visible and most widely used "self-driving" system on the market, Tesla's Autopilot is under more scrutiny than others. Any flaws will be magnified and picked apart. Waymo's total avoidance of accidents resulting in injury or death to date also sets a high bar for safety that Tesla will only ever be able to match if Waymo itself is ever discovered to be at fault for a fatal crash in the future.
Which contender will come out ahead?
Tesla could become the dominant force in autonomous technology, but it seems just as likely it could play the foil to the eventual true market leader -- most likely Waymo -- in a similar fashion to the way Apple's iOS has battled Google's Android over the past decade. However, cars aren't smartphones: the profit margins for the most well-run automakers, in their best years, still come up far short of the margins on Apple's or Google's earnings reports.
It's only by being less cautious, and thus putting his customers at risk, that Musk has pushed Tesla's Autopilot so far so quickly.
This might work in the short run, before Waymo and other self-driving developers make functional systems available to automakers and end-users. Once that happens, the relative safety of various self-driving systems, compared to human drivers and to competing systems, will become clearer.
If Tesla's Autopilot is better than human drivers but worse than Waymo, it won't have an Apple-like toehold on the autonomous vehicle market. It'll wind up looking more like Blackberry, which had a pretty good run too, before Apple and Google knocked it down from top-notch to "just good enough," which eventually gave way to "just hanging on."
