Tesla’s Alleged Self-Driving Tech Defeated By YouTuber With Cartoon-Style Foam Wall

Even if you pay extra for Tesla’s so-called “Full Self-Driving” software, no Tesla is actually capable of self-driving. The weird nerd you know who still wrongly believes Elon Musk personally invented the front trunk may disagree, but as ole Ben Shapiro loves to say, facts don’t care about your feelings. Plus, while other companies use LIDAR on their self-driving prototypes, Tesla relies entirely on cameras. 

Cameras are cheaper than LIDAR, but they also don’t do all the same things, either. That’s been known for a while, but in case you still need a demonstration of LIDAR’s superiority, former NASA engineer and current YouTuber Mark Rober recently put both systems to the test. 

While even a base Toyota Corolla offers adaptive cruise control and more advanced driver assistance than still comes standard on some luxury cars, passenger vehicles equipped with LIDAR are much less common. For that part of the test, Rober got some help from Luminar, the supplier Volvo uses for the LIDAR in the EX90. As for the Tesla, Rober used a Model Y that he personally owns. He then evaluated how well each vehicle handled tests meant to simulate a child standing in the road, a child running into the road, heavy fog, heavy rain, blinding lights and, in the ultimate test, a giant foam wall showing an image of the road behind it, Wile. E. Coyote-style.

Obviously, the Tesla lost. There’s simply no way cameras are going to outperform LIDAR, period. That doesn’t mean the Tesla’s cameras were useless, though. With the deceptively named “Autopilot” engaged, the Tesla was able to stop from 40 mph before hitting the stationary child-size mannequin, and also detected the same mannequin as it “ran” out from behind another car. The Luminar car equipped with LIDAR also stopped in time for both tests, leaving the two technologies tied going into round three. 

The third test, however, is where LIDAR’s superiority began to shine, and the limits of a camera-based system became apparent. While the lasers cut through the fog with no problem, the Tesla simply drove right through the mannequin. As for the torrential downpour test, while the wall of water initially hid the mannequin from the LIDAR, it still detected it in time to come to a stop safely. Considering how those tests went, it would have also made sense if the Tesla failed the blinding lights test, but it actually did manage to spot the mannequin just in time. 

The most interesting challenge, of course, was the foam wall made to look like the road behind it, just like in those old Road Runner cartoons. Would the Tesla’s cameras pick up on the obstacle in its path and stop just like the LIDAR car? Of course not. After passing the blinding light test, you could be forgiven for thinking there was a chance, but nope: It just plowed right through. It probably isn’t something you’ll ever encounter in real life, but it still does a great job of highlighting just how flawed Tesla’s camera-only approach is. 

As you can imagine, those convinced Tesla is on the verge of solving autonomy and becoming a $5 trillion company by selling a humanoid robot to every human on Earth — all very real things Musk has actually promised on earnings calls with investors — are not happy with a famous YouTuber broadcasting even more evidence to the world that a camera-only system is inferior. They’re saying it was unfair to use Autopilot instead of Full Self-Driving, as if more software could make up for the inherent limitations of a camera system. They’re arguing humans drive using a vision-only system known as our eyes, as if that’s proof LIDAR isn’t superior. They’re saying he must have faked the results and not used Autopilot at all. Basically, they’re throwing anything they can think of at the wall and hoping it sticks. 

Now, I wasn’t there. I didn’t witness Rober’s testing in person and can’t provide conclusive evidence that he didn’t cheat in any way. And even if I could, it’s not like the Tesla zealots would accept it from someone who’s been on the record for years saying Autopilot is bad — a position I still hold to this day. However, Rober isn’t some part-time YouTuber with a few thousand subscribers, a day job selling tires and no real assets of any kind. He has more than 65 million subscribers and likely makes substantial money from his channel. He has assets that can be used to satisfy a legal judgment. If Tesla believes this video was deceptively edited to defame the automaker, it has every right to, and absolutely should, sue Mark Rober. 

If Tesla sues and wins, then we’ll have our proof. Until then, the question stands: If this video really is deceptive and inaccurate, why hasn’t Tesla sued yet?