Given the OP, maybe it’s not flawed vision, but close to human vision. Elon’s philosophy of trying to max out simulating people is not a great tech solution when there were better choices.
It is fundamentally flawed. Ask experts in the area. The state of the art is nowhere near what better classes of sensors will do. This is exactly the sort of stupidity that caused iRobot to be overtaken in the home space and go down to penny stock status.
And, yes, it comes down to Elon’s philosophy, which is often remarkably dumb.
I believe vision-only is driven by legal issues. There will be times when vision and other sensors provide conflicting info. If a car crashes, it will be deemed at fault if the vision recording shows it made a mistake (even though other sensors induce it made the right call) - this is because any judge or jury can really only interpret the vision data. Thus there is no point is having the car make any choices other than what vision indicates.
Automated systems have existed for a long time that can react much faster than visual stimulation.
Even our own body does this to prevent us from getting burned. If you touch something hot your pain receptors fire off much fast than it would take for you to see your hand getting burned. You pull away your hand quickly, which is another body reaction, preventing further injury.
We should be using multiple system just like the human body does to prevent injury. Any system we design should also be redundant just like the human body as well. We should not be relying solely on vision for autonomous vehicles.
Given the OP, maybe it’s not flawed vision, but close to human vision. Elon’s philosophy of trying to max out simulating people is not a great tech solution when there were better choices.
It is fundamentally flawed. Ask experts in the area. The state of the art is nowhere near what better classes of sensors will do. This is exactly the sort of stupidity that caused iRobot to be overtaken in the home space and go down to penny stock status.
And, yes, it comes down to Elon’s philosophy, which is often remarkably dumb.
I believe vision-only is driven by legal issues. There will be times when vision and other sensors provide conflicting info. If a car crashes, it will be deemed at fault if the vision recording shows it made a mistake (even though other sensors induce it made the right call) - this is because any judge or jury can really only interpret the vision data. Thus there is no point is having the car make any choices other than what vision indicates.
Wow. No. Just no. We don’t build tech based on a juror can appreciate. We build tech to work.
From an engineering point of view, of course we want to do that. From a business point of view it’s really not the goal, sadly.
the dumbest shit i heard today
Well it doesn’t fit with your preconceived notion. And I understand that can be tough for some people to deal with.
You know LIDAR data can be visualized as well, yes?
Sure, but people who can’t get out of jury duty, aren’t usually very smart.
Automated systems have existed for a long time that can react much faster than visual stimulation.
Even our own body does this to prevent us from getting burned. If you touch something hot your pain receptors fire off much fast than it would take for you to see your hand getting burned. You pull away your hand quickly, which is another body reaction, preventing further injury.
We should be using multiple system just like the human body does to prevent injury. Any system we design should also be redundant just like the human body as well. We should not be relying solely on vision for autonomous vehicles.
I’m not saying vision-only is a good engineering solution. Just that businesses have other considerations.