Comment on Things at Tesla are worse than they appear
AA5B@lemmy.world 3 days agoI tried fsd demo this spring and it’s getting pretty good. I wouldn’t use it but it was perfect on well marked roads. The thing is it made me realize just how poorly maintained our roads are and everything is an edge case. For example it didn’t stay in lane at one Intersection but the intersection was a weird offset plus the lines were all faded away. Although I also disnt give it any chance to recover so I suppose it could have been ok: Im not risking it not recovering
It might surprise everyone but mostly by staying in a well maintained well mapped area, like Waymo did. There’s no way it fulfills the claim of self-driving everywhere without more improvements
The robots is will have the next generation computer and higher resolution cameras which may help. However that also allows more overhead for the next ai update
tfm@europe.pub 3 days ago
I get what you mean but it’s still stuck at level 2 and it always will be. No matter how good it is, if you move your eyes from the road, it will eventually kill you. Cameras alone are not sufficient enough for autonomous driving.
AA5B@lemmy.world 3 days ago
I disagree with this assertion, because they’re correct that the only being that can currently drive is relying on vision. Vision alone is sufficient for driving.
But autonomous driving really hasn’t succeeded yet. We still have no idea what is required for autonomous driving or whether we can do it at all, regardless of sensors.
So you’re implying that we can definitely do autonomous driving but can’t do it the way humans do, whereas I say we won’t know the requirements until we find some that succeed, and we may never
tfm@europe.pub 3 days ago
Yeah sure. If you want the same bad results as humans deliver, in terms of crash rates, than it’s possible. I wouldn’t trust it. Also human vision and processing is completely different from computer vision and processing.
AA5B@lemmy.world 3 days ago
Presumably we have the intelligence to set requirements before something can be called self-driving - that’s usually what the fuss is about, whether the marketing is claiming it’s something it’s not.
If they fail with their approach, I’m fine with that, just like I’m fine if Waymo fails with their approach. Obviously there’s a problem if it runs over some old lady at a stop sign and drags them down the street but that’s clearly a failure for them