10 comments

[ 3.3 ms ] story [ 30.2 ms ] thread
What an absolute farce.

"Anyone relying on LIDAR is doomed" - daddy musk

Edge cases are hard.
Those shadows are confusing.
Wouldn't be to humans though. Foveated vision and excellent adaptive contrast, would be super obvious
I'm pretty sure humans mis-see and bump into stuff all the time.
Do we have good statistics over how well it does to a human without assistance facing the same conditions yet?
All of the cars in front of it avoided the barrier
And I presume there are many other places where a car has crashed that cars with advanced cruise control didn't crash.
I am no Autopilot fan, but that doesn't seem like the best road design/layout.
People are comparing this with humans. In this case, this type of narrowing of the bridge has existed for a long time I presume - it's not new. Thousands of humans have gone through it without an issue. I can see why without a LIDAR, Tesla would need to do a lot more with subtle vision cues that humans usually pick up due to high dynamic range of our eyes.

Data is available. Figure out when this narrowing of the bridge was constructed, average number of cars passing through the left lane in a day and you can estimate how well humans behave in this situation during mid-day when the shadows look similar.

I am willing to guess that humans do way better in this case but we don't have the data of how many Tesla's went through this situation without an issue.