“Car does it all”?
What will it do when someone on a bike drives between you and the car in the lane next to you? What will it do when you’re rounding a corner and there’s a kid playing with a ball ahead of you on your right and the ball goes rolling in front of you? Will it be able to handle “Park next to my green SUV in front of the left garage door”?
How about a road construction area? Will it read the signs? Will it handle a detour? Will it refuse to take over when there is a heavy rain/snow?
If it can’t do all of these it shouldn’t try to do any of these. The human remains responsible for everything the automation does.
This is an old story, now 8 years old. On March 23, 2018, a Tesla crashed on California State Route 101 near Mountain View, which killed 38-year-old Apple engineer Walter Huang. The crash was caused by a combination of factors involving the Tesla Model X’s Autopilot system, driver behavior, road conditions, and infrastructure issues. Now, obviously the Tesla self-driving feature has improved a lot in 8 years, but this points out an unfortunate case where all the holes in the Swiss cheese lined up.
The Tesla, traveling at approximately 71 mph, was in the high-occupancy-vehicle lane using Autopilot (Traffic-Aware Cruise Control and Autosteer). It veered left into the paved gore area (divider) between Highway 101 and the Highway 85 exit ramp, striking a damaged crash attenuator barrier. The impact caused the Tesla to catch fire, and it was then hit by a Mazda and an Audi in adjacent lanes.
The NTSB and other investigations identified multiple contributing factors, with no single cause definitively pinpointed:
Tesla Autopilot System Limitations: The Tesla was operating on Autopilot, which was engaged with adaptive cruise control set to minimum follow-distance. The system failed to detect faded lane lines and steered the vehicle into the gore area, accelerating to 70.8 mph before striking the barrier.
Faded Lane Markings: The NTSB noted that the highway’s left lane lines were faded, causing Autopilot to follow a bolder, incorrect line into the barrier.
Camera-Based System: Tesla’s Autopilot relied on cameras, not lidar or radar, which struggled with worn lane markings or branching lanes.
Prior Complaints: Huang had reported to family and Tesla that Autopilot repeatedly swerved toward the same barrier on prior drives, but Tesla couldn’t replicate the issue at the dealership.
NTSB Critique: The NTSB criticized Autopilot as a “Beta system” not fully developed, noting it allowed drivers to disengage attention, and Tesla’s marketing overstated its capabilities, fostering over-reliance.
Tesla’s Defense: Tesla claimed Autopilot wasn’t a hands-free system, requiring driver supervision, and that Huang didn’t respond to warnings. They also noted 85,000 safe Autopilot trips on that stretch since 2015.
Driver Distraction: The NTSB reported Huang was playing a video game on his smartphone at the time of the crash, suggesting distraction contributed. Data showed his hands were off the wheel for six seconds before impact, despite earlier visual and audible warnings.
Context: Huang had five seconds and 150 meters of clear view of the barrier but took no action, per Tesla’s logs.