columbo.photog/Shutterstock
Tesla is often under scrutiny for its self-driving mode, which has led to questionable, dangerous behaviors and even fatal incidents. This has led the National Highway Traffic Safety Administration (NHTSA) to investigate Tesla's Autopilot for safety multiple times over the past few years. YouTuber Mark Rober put Tesla's self-driving technology to the test, expressing concern that it only uses "simple cameras" to navigate the streets. It turns out that his concerns could be very legitimate.
Rober decided to run a series of tests to see if vision-altering effects would trick Tesla's cameras. He first had Tesla race at 40 miles per hour towards a dummy dressed as an unassuming (and very spaced out) child standing in the middle of the road. The car's automatic braking system didn't react in time, slamming right into the kid. However, turning on autopilot allowed the car to stop before hitting the dummy. Unfortunately, it got worse from here. When Rober added a wall of fog, Tesla's cameras couldn't detect the kid. It also couldn't detect the dummy with a bunch of water pouring down on it, meaning the Tesla hit the kid two more times. It stopped in time when Rober had bright lights blasting at the car to resemble high beams, so it was a pretty mixed bag.
Then came the test everyone was waiting for — would the Tesla slam right into a wall? Rober set up a wall meant to blend into the background environment ("Wile E. Coyote-style") in an attempt to trick the cameras once more. He then drove 40 mph right at the wall — and the Tesla did not touch the brakes at all. He crashed right into the wall at full speed.
Is Lidar safer than Tesla's cameras?
Lidar stands for "light detection and ranging," which essentially means it uses pulsed lasers to create a three-dimensional model of its surroundings faster than sound. It was initially used by NASA to track satellites, but has since found its way into a wide range of vehicles including BMW, Mercedes-Benz, and Waymo — one of Tesla's rivals in the autonomous taxi race. Lidar can detect almost anything in its path, which is why the vehicle using Lidar passed all of the same tests that Tesla's vehicle failed in Rober's viral video. But Tesla CEO Elon Musk keeps arguing against it.
Musk has argued that Lidar isn't necessary for his cars since it can't identify what it's looking at — while it can sense everything around it, it cannot specifically identify any objects. Tesla believes that using a vision-based system can have that same environmental detection but can actually make sense of what's around it, making Tesla vehicles more capable in unpredictable situations. This sounds great in theory, but Rober's test showed that the cameras weren't all that capable in even fog and rain.
Not even Tesla wants you to let its cars take over
RoClickMag/Shutterstock
Tesla's autopilot was not meant to be fully ignored, as Rober found out four dummies later. Tesla itself would tell you that — there's even a camera in your Tesla that is meant to detect if you're paying attention or not. Unfortunately, it has been found to be relatively unreliable in detecting drivers' behavior, leading Tesla users to run red lights and drive on the wrong side of the road while using Full Self-Driving mode. As a result, Tesla's FSD is under investigation by the NHTSA, adding to the long list of investigations.
Tesla's guide on FSD states that the car must be supervised, urging drivers to stay attentive and not assume the car is autonomous. If you see a wall that's painted to look like a hole in the side of a mountain, you'll definitely need to press on the brakes yourself.
