You are viewing a single comment's thread. View all
6
MI7BZ3EW on scored.co
1 year ago6 points(+0/-0/+6Score on mirror)2 children
Having worked tangentially to self-driving research, I can definitively state that it will never happen. We were promised that self-driving cars would be more safe than human drivers. This will never be true.
What people want are trains, not self-driving cars. And even then, those are run by actual humans.
I remember back in the 2010s someone thought it would be cool to have your car drive you to work, then have the car endlessly loop around to avoid paying parking fees, and then pick you up for lunch or whatever. It was at that point that I realized we were looking at another 3DTV situation -- tech that looked cool but nobody wanted, and tech that investors were willing to flush their money down the toilet for.
1 year ago2 points(+0/-0/+2Score on mirror)2 children
you're a larping faggot who has never worked anywhere near self-driving research. the data wholly rejects your assertions. having worked in AI/ML for over a decade...
> We were promised that self-driving cars would be more safe than human drivers. This will never be true.
nah. this is already true. has been for years. the accident rate for fully human driven cars is over 10x the rate for self driving cars. and the accident rate for assisted driving (e.g. auto-brakes) is still many times higher than full self driving.
it's not even fucking close. self driving cars were already beating humans on safety a decade ago. this comes from the fact that self driving cars not only don't get sleepy or drive drunk or while on their phone, they also have hive intelligence. instead of x% of teenage boys getting into the same stupid accident under the same exact conditions, the self driving fleet did it once and learned from it, never repeats it again across the whole fleet.
if you would get in a taxi but you wouldn't get in a self driving car, then you're a victim of media misinformation.
>the accident rate for fully human driven cars is over 10x the rate for self driving cars
And who is primarily causing the accidents, I wonder? Especially on a per-capita basis?
>if you would get in a taxi but you wouldn't get in a self driving car, then you're a victim of media misinformation.
There are videos of NIGGERS lying on the front of self-driving taxis that have lone female occupants inside them to prevent them from moving in the hopes of raping them. A human driver would slam on the accelerator or make an attempt to defend its passenger. A robot does not know that some foreign-born rapists are trying to breach the passenger compartment via brute force by taking advantage of a pre-programmed weakness. It does not understand that the "obstruction" it has stopped for is a hostile one and it needs to get moving. It cannot rationalize something so complex that is otherwise second nature to a human of a mature mental age. It's like a house's automated sprinkler system activating itself as a wildfire surrounds it, not knowing how futile its efforts will be.
A plane may fly 95% of its flight on autopilot. I'm still not getting in one that doesn't have a few men in the cockpit. Truth be told, I wish we still had Flight Engineers, but airlines are too jewed to do that nowadays.
His safety record data point is BS. "Autonomous" vehicles are causing all sorts of accidents. They all require the driver to remain alert and attentive. People who spend any amount of time using those features know that it will go haywire from time to time and wander out of the lane or avoid things that aren't there, or completely miss things that are otherwise obvious.
Human drivers are WAY more safe than the current state of the art with autonomous vehicles. They have way more accidents for the amount of time spent driving than humans do.
Remember, even the best AI image recognition software can't hold a candle to a newborn babe in the most basic tasks, such as identifying faces and people.
If you're still in AI/ML as a true believer, I can't help you. Get all the money you can and don't take stock. That's my recommendation.
> accident rate
The fact of the matter is no one has fully autonomous vehicles yet and we're not even close. The best we have are cars / trucks that run on a closed circuit until it rains and some enhanced driving features on actual roads. We're nowhere close to what a human driver can do.
Of course, you can prove me wrong. Who has actually developed the self-driving technology? As far as I know, they all need drivers to remain alert or need remote drivers to get them out of a jam from time to time.
> the self driving fleet did it once and learned from it, never repeats it again across the whole fleet.
BS. I call BS. I call BS a thousand times. This has NEVER happened with ANY self-driving technology. Stating something like this shows you have no idea how this technology even works.
What I have seen is someone identifies a fault in the system, a human programmer intervenes and either manually codes in conditions and decisions or parameters, or they just add more training data and hope the new model does slightly better.
I have never heard of anyone even getting close to 90% accuracy with their models. Even beating 80% requires herculean efforts.
The issue is if you turn up the safety factor, the vehicles won't even move. If you turn down the safety factor, you are running over old ladies on bicycles because the car can't tell if it's a bicycle or a pedestrian.
What people want are trains, not self-driving cars. And even then, those are run by actual humans.
I remember back in the 2010s someone thought it would be cool to have your car drive you to work, then have the car endlessly loop around to avoid paying parking fees, and then pick you up for lunch or whatever. It was at that point that I realized we were looking at another 3DTV situation -- tech that looked cool but nobody wanted, and tech that investors were willing to flush their money down the toilet for.
> We were promised that self-driving cars would be more safe than human drivers. This will never be true.
nah. this is already true. has been for years. the accident rate for fully human driven cars is over 10x the rate for self driving cars. and the accident rate for assisted driving (e.g. auto-brakes) is still many times higher than full self driving.
it's not even fucking close. self driving cars were already beating humans on safety a decade ago. this comes from the fact that self driving cars not only don't get sleepy or drive drunk or while on their phone, they also have hive intelligence. instead of x% of teenage boys getting into the same stupid accident under the same exact conditions, the self driving fleet did it once and learned from it, never repeats it again across the whole fleet.
if you would get in a taxi but you wouldn't get in a self driving car, then you're a victim of media misinformation.
And who is primarily causing the accidents, I wonder? Especially on a per-capita basis?
>if you would get in a taxi but you wouldn't get in a self driving car, then you're a victim of media misinformation.
There are videos of NIGGERS lying on the front of self-driving taxis that have lone female occupants inside them to prevent them from moving in the hopes of raping them. A human driver would slam on the accelerator or make an attempt to defend its passenger. A robot does not know that some foreign-born rapists are trying to breach the passenger compartment via brute force by taking advantage of a pre-programmed weakness. It does not understand that the "obstruction" it has stopped for is a hostile one and it needs to get moving. It cannot rationalize something so complex that is otherwise second nature to a human of a mature mental age. It's like a house's automated sprinkler system activating itself as a wildfire surrounds it, not knowing how futile its efforts will be.
A plane may fly 95% of its flight on autopilot. I'm still not getting in one that doesn't have a few men in the cockpit. Truth be told, I wish we still had Flight Engineers, but airlines are too jewed to do that nowadays.
Human drivers are WAY more safe than the current state of the art with autonomous vehicles. They have way more accidents for the amount of time spent driving than humans do.
Remember, even the best AI image recognition software can't hold a candle to a newborn babe in the most basic tasks, such as identifying faces and people.
> accident rate
The fact of the matter is no one has fully autonomous vehicles yet and we're not even close. The best we have are cars / trucks that run on a closed circuit until it rains and some enhanced driving features on actual roads. We're nowhere close to what a human driver can do.
Of course, you can prove me wrong. Who has actually developed the self-driving technology? As far as I know, they all need drivers to remain alert or need remote drivers to get them out of a jam from time to time.
> the self driving fleet did it once and learned from it, never repeats it again across the whole fleet.
BS. I call BS. I call BS a thousand times. This has NEVER happened with ANY self-driving technology. Stating something like this shows you have no idea how this technology even works.
What I have seen is someone identifies a fault in the system, a human programmer intervenes and either manually codes in conditions and decisions or parameters, or they just add more training data and hope the new model does slightly better.
I have never heard of anyone even getting close to 90% accuracy with their models. Even beating 80% requires herculean efforts.
The issue is if you turn up the safety factor, the vehicles won't even move. If you turn down the safety factor, you are running over old ladies on bicycles because the car can't tell if it's a bicycle or a pedestrian.
Have fun with that. Cash out while you still can.