We’ve known for some time that Elon Musk has a problem with using radar sensors on Teslas. We’ve also known that the automaker is under federal investigation for crashes involving Autopilot, and that U.S. senators and others have decried the use of the terms Autopilot and Full Self-Driving as deceptive and dangerous. But the New York Times today posted an article, “Inside Tesla as Elon Musk Pushed an Unflinching Vision for Self-Driving Cars,” that does just what the headline says: The Times interviewed 19 people involved with development of Autopilot over the years, who described internal edicts by Musk, disagreements from engineers on the project, and statements by the CEO that they say misled the public.
Central to the inner story of Tesla’s quest for a self-driving car is this insight into Musk’s reasoning. The Times wrote:
“For several years, Autopilot incorporated radar, and for a time Tesla worked on developing its own radar technology. But three people who worked on the project said Mr. Musk had repeatedly told members of the Autopilot team that humans could drive with only two eyes and that this meant cars should be able to drive with cameras alone.”
Musk insisted a Tesla should be capable of driving door-to-door unassisted, using only cameras. Eight of the Tesla engineers interviewed by NYT disagreed with this technological direction, as did outside experts; some left the company, while others followed Musk’s marching orders. But for this reason, along with cost, Musk’s view of a radar emitter’s aesthetics on the vehicle, and his preference not to use third-party suppliers, he pushed for cameras-only technology and in May announced that Tesla was dropping radar.
Yet NYT quotes computer-vision specialist Schuyler Cullen, who oversaw autonomous-driving research at Samsung, as saying a cameras-only approach was fundamentally flawed: “Cameras are not eyes! Pixels are not retinal ganglia! The FSD computer is nothing like the visual cortex!”
Engineers on the project also objected to the marketing team’s plan to dub the driver-assist technology Autopilot, for the same reasons that Tesla’s critics and regulators have since cited — it implies capabilities the technology does not have. Their preferred alternative name was Copilot.
The article also describes the making of this video that is still on Tesla’s website, which purports to show how well Autopilot 2.0 can self-drive. In reality the route was mapped out ahead of time using three-dimensional software not available in the consumer version of Autopilot — and the car hit a roadside barrier during the filming.
NYT rounds up a series of Musk pronouncements over the years that autonomous driving was just around the corner, when it clearly was, and is, not, and says engineers believed Musk was promising capabilities that are not possible. They were “surprised and concerned” when Musk as early as 2016 was promising that all Teslas then being built had the components for “full self-driving.” Less than a month after a fatal crash involving Autopilot in May 2016, Musk stated that autonomous driving was “basically a solved problem” and that Teslas could already drive more safely than humans.
The article concludes with this quote from Amnon Shashua, chief executive of Mobileye, a former Tesla supplier, who advises, “One should not be hung up on what Tesla says. Truth is not necessarily their end goal. The end goal is to build a business.”
Head over to the New York Times for the full article.