HN Theater @HNTheaterMonth

There is a great argument for having ultrasonic sensors and radar in a recent video by Rayan from FortNine discussing two fatal accidents involving tesla autopilot https://www.youtube.com/watch?v=yRdzIs4FJJg

It may be a contributing factor to recent cases where Tesla Autopilot rear ended motorcycles at night. Thus video has a visualization: https://youtu.be/yRdzIs4FJJg

One of these crashes was right by my home, where I rode to work nearly every day prepandemic.

More recently, Tesla seems to struggle with motorcyclists sometimes: https://www.youtube.com/watch?v=yRdzIs4FJJg

⬐ salty_biscuits

Too be fair, so do a lot of human drivers. Something about motorbikes makes them difficult to perceive. Not sure if it is car pillar design, not expecting to see them there (a prior of just looking for cars) or just a small frontal area.

⬐ hulitu

Just today someone said in another thread that Tesla is good at SW. They might be good at software but they are bad at engineering. It is not enough to write the SW. SW must be also tested. In the real world.

⬐ isitmadeofglass

Tesla is testing in the real world. Unfortunately this has led to the death of many.

⬐ intothemild

This is a company that thinks stop signs are standardized so you can get a Tesla to stop early by using a slightly larger stop sign.

Relying on CV alone is stupid.

⬐ adamdusty

He glosses over the part where tesla autopilot shuts itself down just before impact. It doesnt matter that much but it's an interesting point I didn't know.

As a rider, I'm not really too worried about tesla autopilot, mostly because I don't do much freeway riding and I'm probably not the safest rider to begin with. Plus there are plenty of normal drivers that make questionable maneuvers solely because they don't see bikes. In fact, even as a rider myself, at night I often have trouble distinguishing a motorcycle from a car with one light out.

⬐ mwint

AP safety stats include crashes where the system was active within the previous x seconds - I think x is something like 10 or 30. Plenty of time for a human to become responsible.

⬐ adamdusty

I'm not arguing that it's to game the stats, but it's still an interesting point I didn't know. I wonder what the reasoning behind the shut off is, and why 1 second (or whatever it is). Why not 2 or 3, or a quarter second?

⬐ mwint

My guess is the path planner hits an impossible scenario (“we don’t have enough brake authority to stop in time”) and that triggers the disengagement. Doubtful there’s a DISENGAGE_SECONDS_BEFORE_CRASH constant somewhere.

⬐ ed

> Tesla‘s vehicles have been found to shut off the advanced driver-assistance system, Autopilot, around one second before impact, according to the regulators.

https://www.washingtonpost.com/technology/2022/06/15/tesla-a...

⬐ mwint

Sure, because it’s given up. Those crashes are included in the safety stats.

⬐ TheLoafOfBread

How do you know it when you can't independently verify it?