After a self-driving Uber crashed and killed a woman walking her bike across a Tempe, Arizona, road last week, the technology behind autonomous vehicles has been questioned and scrutinized.
Intel, the company behind the driver-assistance software Mobileye — which is used in certain autonomous cars, but not in Uber — took the footage Tempe police released from the crash and ran their software through the fatal incident.
Mobileye CEO and CFO Amnon Shashua wrote in a editorial Monday that “despite the suboptimal conditions, where much of the high dynamic range data that would be present in the actual scene was likely lost, clear detection was achieved approximately one second before impact.”
How the system would have classified the scene with the pedestrian and bicycle in the road is shown above, based on the exterior video police released last week.
So Intel, which also announced this year it was working on autonomous cars with BMW, Nissan, and Volkswagen, claims it would have detected and classified the situation accurately. But even if the vehicle had a one-second warning about 49-year-old Elaine Herzberg in the road, would that have done anything to spare her life? Shashua, who has written about the safety of autonomous vehicles before, doesn’t go so far as to say his technology would have saved Herzberg, but he does point out that the sensors and software they used on the low-quality video is a “basic build block” for autonomous driving.
“It is the high-accuracy sensing systems inside [advanced driver assistance systems] that are saving lives today, proven over billions of miles driven. It is this same technology that is required, before tackling even tougher challenges, as a foundational element of fully autonomous vehicles of the future,” he wrote. “The video released by the police seems to demonstrate that even the most basic building block of an autonomous vehicle system, the ability to detect and classify objects, is a challenging task.”
An Intel spokesperson responded to further questions about the post asking, “If you don’t have the basics to understand the environment with high accuracy, how can you make driving decisions such as to stop, swerve, or speed up?”
This seems to question the technology used in Uber’s self-driving program and how it works, for example, in a situation with a pedestrian suddenly in the roadway. A New York Times report from last week revealed problems within Uber’s self-driving operation and its struggle to keep up with competitors.
Uber didn’t respond to request for comment about Intel’s post and its software analysis of the crash.
UPDATE: March 27, 2018, 1:05 p.m. PDT Uber said it wasn’t able to comment on Intel’s post.