I bought FSD on Sept 19th, 2025. I have not driven my Tesla on a public road since then. It is all done by FSD. I just completed a trip from Pittsburgh to Louisianna to Florida and back to Pittsburgh without touching the wheel. Perhaps the author ought to try it.
I went from San Francisco to Vegas and back last week, and it's still not quite there, but it almost there. It had problems navigating parking lots, making turns too early or too late, but the problem with it is that there's no tiny windshield wipers on the cameras. When dirt from driving accumulates on them, it can't self drive, and then you're totally lost and confused. The real problem with it though, is that because it's Tesla, it's under extreme scrutiny, so the driver monitoring is non-trivial to hack, and if you don't, you get strikes for not watching the road and then it doesn't self drive any more. Comma.ai gets to be a bit more forgiving because of how they're positioned in the market.
"Still not quite there"?
Imagine common operation of vehicles without windshield wipers!
"you get strikes for not watching the road"
What does Doctorow call this risk: where you confuse a requirement for your vigilance to a machine with liberation?
Another risk looms as vehicle occupants are officially required to be back-seat drivers for a system that allows no driving experience...
Everywhere you look at AI applications you see the hazard of mechanical mad-cow disease.
"Kenyans don't talk like AI!"
> Or, to put it another way, if self driving cars truly exist, you’ll have a lot more types of robotics completely autonomously operating.
Which is why you solve the general humanoid-robot problem because then you’ve solved many autonomy problems including that of driving because then the humanoid-robot does the driving… as well as perform many other general world tasks.
Mr. Bullwinkle acknowledges Comma.ai, but doesn't dig into them deeply. It's really really good cruise control on freeways. That task is well defined. Every new car that is sold today should be that good. It's not self driving, it's not sentient, it doesn't change lanes for you. But I can do the city driving. It's the long boring endless stretches of freeway that was the problem and they've solved that. Yes there's more to go, but cameras + radar is good enough. The insightful thing is to get the insurance companies on board. Money talks.
I saw this headline a while back
Full Self-Driving (Supervised) | Tesla United Kingdom.
Supervised!
we are all going to live on mars one day, (Supervised)
we will be able to have day trips to the moon and back, (Supervised)
Everyone will have AGI within the next three months, (Supervised)
Everyone will have personal quantum computers in three months, (Supervised)
I bought FSD on Sept 19th, 2025. I have not driven my Tesla on a public road since then. It is all done by FSD. I just completed a trip from Pittsburgh to Louisianna to Florida and back to Pittsburgh without touching the wheel. Perhaps the author ought to try it.
I went from San Francisco to Vegas and back last week, and it's still not quite there, but it almost there. It had problems navigating parking lots, making turns too early or too late, but the problem with it is that there's no tiny windshield wipers on the cameras. When dirt from driving accumulates on them, it can't self drive, and then you're totally lost and confused. The real problem with it though, is that because it's Tesla, it's under extreme scrutiny, so the driver monitoring is non-trivial to hack, and if you don't, you get strikes for not watching the road and then it doesn't self drive any more. Comma.ai gets to be a bit more forgiving because of how they're positioned in the market.
"Still not quite there"?
Imagine common operation of vehicles without windshield wipers!
"you get strikes for not watching the road"
What does Doctorow call this risk: where you confuse a requirement for your vigilance to a machine with liberation?
Another risk looms as vehicle occupants are officially required to be back-seat drivers for a system that allows no driving experience...
Everywhere you look at AI applications you see the hazard of mechanical mad-cow disease.
"Kenyans don't talk like AI!"
> Or, to put it another way, if self driving cars truly exist, you’ll have a lot more types of robotics completely autonomously operating.
Which is why you solve the general humanoid-robot problem because then you’ve solved many autonomy problems including that of driving because then the humanoid-robot does the driving… as well as perform many other general world tasks.
Mr. Bullwinkle acknowledges Comma.ai, but doesn't dig into them deeply. It's really really good cruise control on freeways. That task is well defined. Every new car that is sold today should be that good. It's not self driving, it's not sentient, it doesn't change lanes for you. But I can do the city driving. It's the long boring endless stretches of freeway that was the problem and they've solved that. Yes there's more to go, but cameras + radar is good enough. The insightful thing is to get the insurance companies on board. Money talks.
I saw this headline a while back
Full Self-Driving (Supervised) | Tesla United Kingdom.
Supervised!
we are all going to live on mars one day, (Supervised)
we will be able to have day trips to the moon and back, (Supervised)
Everyone will have AGI within the next three months, (Supervised)
Everyone will have personal quantum computers in three months, (Supervised)