4 ms·

> where it successfully navigates the most dangerous road in America Twistiest, maybe. Most dangerous, not even close unless you or someone driving towards yo

by wolrah 5y ago

> where it successfully navigates the most dangerous road in America Twistiest, maybe. Most dangerous, not even close unless you or someone driving towards you do something dumb. Historically Autopilot's weak spots have been crappy/ambiguous road markings, low speed corners on high speed roads, and large vehicles stopped on/across the road. This video hit none of those points. The lane markings are clear, the posted limit is low, and large vehicles aren't even supposed to be there.

reaperducer 5y ago
Twistiest, maybe. Most dangerous, not even close It's just CGP Grey, not NHTSA. Hyperbole is to be expected. I know he's a darling among the chattering tech class, but it's been several years since he jumped the shark from "insightful, carefully researched, and well thought out" to "just another vlogger doing and saying what is needed to get views."
maxdo 5y ago
your historical observations are irrelevant for FSD beta. It's a new system, it works amazing without markings on the road at all.
- wolrah 5y ago
  > your historical observations are irrelevant for FSD beta. It's a new system 1. I find it very unlikely that Tesla completely scrapped the previous system to start from scratch for the FSD beta. 2. If they had, that would make me even more suspicious of the new system because that would mean no one would understand its weaknesses yet. Either way, being most interested in where the previous generation system and other semiautonomous systems have historically failed makes sense. > it works amazing without markings on the road at all. I have yet to see any videos of the new system handling poorly marked high speed roads one way or another, but the crash where it tried to veer in to oncoming traffic doesn't provide confidence. All the unmarked/poorly marked roads I've seen so far have been low speed residential/urban, and even there performance has been unpredictable at best. Let's see it on a highway where night time rain has made the markings hard to see. Let's see it when snow covers the road and your only hing or lane The state of the art is such that it can simultaneously be amazing and terrifying to think that normal people are allowed to "test" it wherever they feel like trying on public roads. What it does with the limited inputs it has is incredible, but at the same time it has dangerous failure modes that do not seem to be predictable even to those who use the system regularly. --- Don't get me wrong, there is nothing I'd love more than never having to manually handle a traffic jam, or "go home car, I'm drunk", but the way this is being released will kill people.
seoulmetro 5y ago
Not to mention it's dangerous because of the amount of people acting irrationally and speeding there. There were barely any cars on the road with him (and he did that on purpose, possibly to save people the hassle or actually not kill others).

by wolrah 5y ago

reaperducer 5y ago

Twistiest, maybe. Most dangerous, not even close It's just CGP Grey, not NHTSA. Hyperbole is to be expected. I know he's a darling among the chattering tech class, but it's been several years since he jumped the shark from "insightful, carefully researched, and well thought out" to "just another vlogger doing and saying what is needed to get views."

maxdo 5y ago

your historical observations are irrelevant for FSD beta. It's a new system, it works amazing without markings on the road at all.

wolrah 5y ago
> your historical observations are irrelevant for FSD beta. It's a new system 1. I find it very unlikely that Tesla completely scrapped the previous system to start from scratch for the FSD beta. 2. If they had, that would make me even more suspicious of the new system because that would mean no one would understand its weaknesses yet. Either way, being most interested in where the previous generation system and other semiautonomous systems have historically failed makes sense. > it works amazing without markings on the road at all. I have yet to see any videos of the new system handling poorly marked high speed roads one way or another, but the crash where it tried to veer in to oncoming traffic doesn't provide confidence. All the unmarked/poorly marked roads I've seen so far have been low speed residential/urban, and even there performance has been unpredictable at best. Let's see it on a highway where night time rain has made the markings hard to see. Let's see it when snow covers the road and your only hing or lane The state of the art is such that it can simultaneously be amazing and terrifying to think that normal people are allowed to "test" it wherever they feel like trying on public roads. What it does with the limited inputs it has is incredible, but at the same time it has dangerous failure modes that do not seem to be predictable even to those who use the system regularly. --- Don't get me wrong, there is nothing I'd love more than never having to manually handle a traffic jam, or "go home car, I'm drunk", but the way this is being released will kill people.

seoulmetro 5y ago

Not to mention it's dangerous because of the amount of people acting irrationally and speeding there. There were barely any cars on the road with him (and he did that on purpose, possibly to save people the hassle or actually not kill others).