Comment by dsf2d
9 days ago
Its not ever going to get ready.
Getting this to a place where it is better than humans continuously is not equivalent to fixing bugs in the context of the production of software used on phones etc.
When you are dealing with a dynamic uncontained environment it is much more difficult.
Waymo is in a place where it's better than humans continuously. If Tesla is not, that's on them, either because their engineers are not as good or because they're forced to follow Elon's camera-only mandate.
It's the camera-only mandate, and it's not Elon's but Karpathy's.
Any engineering student can understand why LIDAR+Radar+RGB is better than just a single camera; and any person moderately aware of tech can realize that digital cameras are nowhere as good as the human eye.
But yeah, he's a genius or something.
Digital cameras are much worse than the human eye, especially when it comes to dynamic range, but I don't think that's all that widely known actually. There are also better and worse digital cameras, and the ones on a Waymo are very good, and the ones on a Tesla aren't that great, and that makes a huge difference.
Beyond even the cameras themselves, humans can move their head around, use sun visors, put on sunglasses, etc to deal with driving into the sun, but AVs don't have these capabilities yet.
4 replies →
I have enjoyed Karpathy's educational materials over the years, but somehow missed that he was involved with Tesla to this degree. This was a very insightful comment from 9 years ago on the topic:
> What this really reflects is that Tesla has painted itself into a corner. They've shipped vehicles with a weak sensor suite that's claimed to be sufficient to support self-driving, leaving the software for later. Tesla, unlike everybody else who's serious, doesn't have a LIDAR.
> Now, it's "later", their software demos are about where Google was in 2010, and Tesla has a big problem. This is a really hard problem to do with cameras alone. Deep learning is useful, but it's not magic, and it's not strong AI. No wonder their head of automatic driving quit. Karpathy may bail in a few months, once he realizes he's joined a death march.
> ...
https://news.ycombinator.com/item?id=14600924
Karpathy left in 2022. Turns out that the commenter, Animats, is John Nagle!
Using only cameras is a business decision, not tech decision: will camera + NN be good enough before LIDAR+Radar+RGB+NN can scale up.
For me it looks like they will reach parity at about the same time, so camera only is not totally stupid. What's stupid is forcing robotaxi on the road before the technology is ready.
3 replies →
If you ride enough Waymo, you will realize it is a far more cautious driver than a human but not a better driver. If you need to get somewhere even at an average speed, you still take uber/lyft.
Waymo still takes many wrong turns and can easily get stuck in situations where a human would not.
citation needed. Waymo says they are better, but it is really hard to find someone without a conflict of interest who we can believe has and understands the data.
I reject the premise of your comment. If Tesla wants to convince people that Robotaxi is safe, it's on them to publish an analysis with comparative data and stop redacting the crash details that Waymo freely provides. Until they do, it's reasonable to follow the source article's simple math and unreasonable to declare that there's no way to be sure because there might be some unknown factor it's not accounting for.
1 reply →
It's clear that camera-only driving is getting better as we have better image understanding models every year. So there will be a point when camera based systems without lidars will get better than human drivers.
Technology is just not there yet, and Elon is impatient.
Then stop deploying camera only systems until that time comes.
Waymo could be working on camera only. I don’t know. But it’s not controlling the car. And until such a time they can prove with their data that it is just as safe, that seems like a very smart decision.
Tesla is not taking such a cautious approach. And they’re doing it on public roads. That’s the problem.
> So there will be a point when camera based systems without lidars will get better than human drivers.
No reason to assume that. A toddler that is increasing in walk speed every month will never be able to outrun a cheetah.
3 replies →
Lidar and radar will also get better and having all possible sensors will always out perform camera only.