Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

How much of Tesla's FSD is learned from watching a lot of recordings of human drivers and how much of it is explicitly coded rules?

I'd expect school bus stop signs to be challenging to learn because of how context dependent they are compared to regular stop signs. Some examples, drawn from the rules in my state (Washington) are below. These may be different in other states.

With a regular stop sign you stop, then you go when it is safe to do so.

When you stop for a school bus sign you are supposed to stay stopped as long as the sign is there. You don't go until the school bus retracts the sign. It is essentially like a red light rather than a stop sign.

Whether or not a school bus stop sign applies to you depends on the direction you are traveling and on the structure of the roadway.

It applies if either of the following are true: (1) you are traveling in the same direction as the bus, (2) it is a two lane roadway and the lanes are not separated by a physical barrier or median. Three or more lanes or a barrier or median and cars traveling opposite the bus don't have to stop.



Just a data point, no argument:

My 2025 Tesla stops for road work flaggers spinning their signs from "Slow" to "Stop".

This issue here, correctly, is that it should come to a stop for 1) a bus flashing red, 2) with or without stop signs, 3) on an undivided road. Or, in our automated future, at least come to a crawl like FSD does now when entering a parking lot or pedestrian likely location.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: