I mean... y'all knew I wouldn't leave this one alone. It was just a matter of time before this catfish ate the stinkbait.
I did take my time and watch the entire video and did multiple viewings of all of the actual tests.
You'll all be shocked to hear that I think the video is a bit disingenuous and manipulated to give pre-determined results. Here are my observations that inform my skepticism:
Rober doesn't use any FSD build and limits his testing to EAB and Autopilot. Yes, the sensor suite and hardware is the same, but the software is vastly different between autopilot and FSD 12+. Rober pits the latest version of Luminar's system against an older generation Tesla system.
In each test where the screen is visible (except for the EAB test), there is an out-of-focus warning notification present prior to impacting the dummy or the wall. Anyone who uses the Tesla TACC, autopilot, or FSD system will tell you that if you have the accelerator pressed and are going above the set speed, you get the following warning: Even EAB will keep going when the accelerator is pressed, which is what appears to happen in the video. It won't give the visual warning notification, but it will give an audible alarm which you hear about two seconds prior to impact in the EAB test in the video. To my eye, it looks like Rober is pressing the accelerator and not actually allowing the driver assist systems to maintain speed on their own.
In the raw footage that Rober released in response to criticism, rather than setting the autopilot early in the test which is easily doable, he waits until very late (<5 seconds) to attempt to engage FSD and fails on two attempts to engage before getting it to engage on a third attempt. There's no reason he couldn't have engaged autopilot much earlier. It's just really shady and looks like he is intentionally engaging late to confuse and cause errors in the system. https://x.com/MarkRober/status/1901449395327094898
The "fog" test was not fog, but smoke. LIDAR is significantly degraded in actual fog when water in the air scatters and absorbs the lasers. Credit to the LIDAR system for "seeing" through the smoke, but that is an extremely uncommon scenario for actual driving (no safe driver proceeds at full speed into an obscuring smoke cloud) that was engineered to give the LIDAR system an easy "win" in this comparison.
In the "rain" test, which did not simulate a realistic rain scenario that one would actually drive in, the LIDAR system stopped because it read the significant water spray as a solid object, not because it saw the dummy. Judge's review does not award the point to LIDAR in this comparison. LIDAR is degraded in rain just like in fog due to attenuation.
The CEO of Luminar (the LIDAR system featured in the video) is present in the video and is a previous collaborator and admitted friend of Rober.
For posterity: Motherfuck Elmo for being a nazi asshole.
You may now proceed with the knob-slobbing gifs.