The benefit they point out is 200fps vs. 2000fps, but is that actually that useful? In the post of OpenAO's robot, they said that improving the response times gave no significant gains. Only thing I can imagine would be allowances for wider and deeper neural nets.
Tesla is building its own AI chips for self-driving cars
11–20 of 238 posts
Re: Tesla is building its own AI chips for self-driving cars
#12What are the real-world benefits going from 200 to 2000 frames a second? 200 seems quite fast.
Re: Tesla is building its own AI chips for self-driving cars
#13This is pretty bad from a hardware/software verification perspective. People who care about correctness can just go home, I guess. It's amazing that we're gonna let these things on the streets. When you weigh almost 2 tons and can go >100mph, it doesn't matter if you have a gun mounted on you or not, you are a weapon .
Re: Tesla is building its own AI chips for self-driving cars
#14This is pretty bad from a hardware/software verification perspective. People who care about correctness can just go home, I guess. It's amazing that we're gonna let these things on the streets. When you weigh almost 2 tons and can go >100mph, it doesn't matter if you have a gun mounted on you or not, you are a weapon .
Re: Tesla is building its own AI chips for self-driving cars
#15So much for no new hardware required for L5
Just wait until they admit that they need LIDAR too... Cha Ching! Various Autopilot accidents strongly imply that whatever the computer “brain” is doing doesn’t matter if it doesn’t have sufficient “eyes” to actually see things like a fire truck, cop car, or bollard.
Re: Tesla is building its own AI chips for self-driving cars
#16This is pretty bad from a hardware/software verification perspective. People who care about correctness can just go home, I guess. It's amazing that we're gonna let these things on the streets. When you weigh almost 2 tons and can go >100mph, it doesn't matter if you have a gun mounted on you or not, you are a weapon .
does correctness even work with deep neural networks to begin with? especially with the existence of adversarial examples, out of domain situations, etc
Re: Tesla is building its own AI chips for self-driving cars
#17So much for no new hardware required for L5
Re: Tesla is building its own AI chips for self-driving cars
#18Earlier quoted context omitted.
Does it matter if Tesla does the hardware swap on existing vehicles for free? Costs should not be high to do so if it’s their own silicon and the swap is done as part of regular maintenance at Tesla service centers. The benefit of high margins and sophisticated engineering is you can get away with a lot of unknowns and promises.
Yes. It demonstrates that their claims have been complete BS.
Re: Tesla is building its own AI chips for self-driving cars
#19Earlier quoted context omitted.
Just wait until they admit that they need LIDAR too... Cha Ching! Various Autopilot accidents strongly imply that whatever the computer “brain” is doing doesn’t matter if it doesn’t have sufficient “eyes” to actually see things like a fire truck, cop car, or bollard.
And yet squishy meatbags manage to sensor fuse a stereoscopic image (plus nascient sound based imaging) into a reliably rendered picture of the world...
Re: Tesla is building its own AI chips for self-driving cars
#20So much for no new hardware required for L5
Just wait until they admit that they need LIDAR too... Cha Ching! Various Autopilot accidents strongly imply that whatever the computer “brain” is doing doesn’t matter if it doesn’t have sufficient “eyes” to actually see things like a fire truck, cop car, or bollard.
The current model makes assumptions that sometimes turn out to be false. So yea it's not going to get to perfection, but level five only needs better than human not perfection.
* Granted I am likely a bad driver, but I don't think I could do this well.