Huge backlash over a video claiming to show Tesla’s LiDAR-free autonomous systems as inferior!

In the video, supported by LiDAR manufacturer Luminar, Tesla is depicted crashing into a wall painted with a fake road image—reminiscent of the Road Runner cartoon—and failing to detect obstacles in heavy fog.

The main issue with the video is that, instead of Tesla’s Full Self-Driving (FSD) system, an older software infrastructure with limited capabilities—essentially a basic autopilot—was used. In other words, rather than testing Tesla’s true autonomous driving mode, the video evaluated the simplest ADAS features, such as adaptive cruise control and lane-keeping.

The second issue is that the autopilot wasn’t even active during the scenes involving the crash into the wall with the fake road image or the foggy road scenarios. Raw footage released later revealed that the wall scene was repeated multiple times to capture the “best” crash moment. Already failing to conduct a fair comparison, Mark Rober is now accused of editing the video to portray the autopilot in the worst possible light. Luminar, which had initially started using the content for promotional purposes, removed the video from its website following the backlash. While the video may be flawed, it does not change the fact that LiDAR is superior to camera-based systems in obstacle detection. However, it is evident that an expensive obstacle-detecting sensor is not essential for autonomous driving.

We recently saw in China that there is currently no LiDAR-based autonomous driving system that can rival Tesla’s FSD software on “real roads.” For autonomous driving, it is crucial to develop a neural network capable of performing well even in extreme edge cases.

 

 

 

 

Bu gönderiyi Instagram'da gör

 

 

 

 

 

 

 

 

 

 

 

licarco (@licarcommunity)'in paylaştığı bir gönderi