- YouTuber Mark Rober tested Tesla’s Autopilot against a LiDAR-equipped car to assess its obstacle detection.
- Tesla’s Autopilot struggled in extreme fog, rain, and painted wall tests, unlike the lidar systems.
- Elon Musk has criticized lidar in the past and argued that camera-based tech is sufficient for self-driving cars.
A YouTuber’s homemade Tesla test looks like something straight out of a children’s cartoon.
CrunchLabs founder and YouTuber Mark Rober put Tesla’s Autopilot through its paces to see if it could detect the same road obstacles as a vehicle that uses LiDAR — a system that Elon Musk has previously called a “fool’s errand.”
The verdict? Kind of.
In the video, the Tesla Model Y on Autopilot successfully detected a child mannequin in front of the car, one that appeared with less than a second of braking time, and another standing in front of the vehicle amid bright lights to simulate sunset conditions.
However, it failed to avoid hitting the mannequin when obscured by intense fog, heavy rain, and a Wile E. Coyote-style Roadrunner-painted wall.
The alternative car, a Lexus with LiDAR in its self-driving system, passed every test in Rober’s video.
Most cars with self-driving technology use LiDAR to map roadways, other cars, and possible obstructions around the vehicle. LiDAR stands for Light Detection and Ranging and is similar to radio detection and ranging, but uses laser lights invisible to the human eye instead of radio waves.
Musk said in the company’s last earnings call in January that “obviously, humans drive without shooting lasers out of their eyes.” The Tesla CEO said that humans rely on their eyes and biological neural networks to navigate while driving, and the digital counterpart to that is cameras and digital neural nets, or AI.
The Tesla CEO said that “the entire road system was designed for passive optical neural nets.”
“Look, we even have a radar in the car and we turned it off,” Musk said.
Rober said the last test, in which he drove the Tesla through the fake road wall at 40 miles per hour, was the “only one that matters.”
“While that looks sort of convincing, the image processing in our brain is advanced enough that we pick up on minor visual inconsistencies, and we wouldn’t hit it,” Rober said about the wall.
Tesla did not immediately respond to a request for comment from Business Insider.
While the fake-wall test highlighted that Tesla’s Autopilot cameras and AI system still have a ways to go in some scenarios, it’s also highly unlikely that a human would ever face such an illusion in real life.
The weather tests, which were closer to reality, also created smoke and rain conditions so dense that a human wouldn’t be able to see through it. However, the goal of self-driving systems is to create technology so advanced that it detects potential incidents faster than a human driver, increasing road safety.
In response to the YouTuber’s test, some Tesla fans and others on social media criticized the video and Rober’s decision to use Autopilot instead of the more advanced Full-Self-Driving beta software, which is a paid upgrade.
Rober responded to some of the criticism by sharing what he described as the raw footage of the test.
Musk has previously said that a Tesla on Autopilot is almost 10 times less likely to crash than the average car, a claim that some experts have said is misleading.