4 ms·
Proof that this system lacks night vision. No different to the video of the deer getting run over by another Tesla on autopilot [0] and the near miss that happe
by rvz 5y ago
Proof that this system lacks night vision. No different to the video of the deer getting run over by another Tesla on autopilot [0] and the near miss that happened when the system malfunctioned and tried to hit another car [1] - both cases at night.
Sounds like the risks in using both either autopilot or FSD systems at night is even more dangerous but very surprising that these systems still have no night-vision system.
Shocking.
[0] https://news.ycombinator.com/item?id=28732566 https://news.ycombinator.com/item?id=28732566
[1] https://news.ycombinator.com/item?id=29515784 https://news.ycombinator.com/item?id=29515784
- jeffreygoesto 5y agoActive sensing like Lidar is even better than night vision. The side view of a damaged truck is really hard to train as a relevant class reliably, even for Karpathy.
- Workaccount2 5y agoAll the data should be there for the computer to decide that yes there is a large object, yes it is in the road, and yes we are on track to hit it.
- jeffreygoesto 5y agoYou'd be surprised how much modeling is needed and how little the raw sensor data has to do with 'objects'.
- KaiserPro 5y agonot really, its a generic obstruction. You don't need to classify it as a car, just a generic object blocking the way.
- jeffreygoesto 5y agoWhat class are "generic" objects and what would be a typical representative? ;)
- KaiserPro 5y agoroad obstacles. rock fall, tree, building. Given that the road is occluded, this needs to be in the training data, if only for autonomous lane following.
- tyingq 5y agoWikipedia seems to indicate the car should have had front-facing radar. I'm curious why that appears to have not helped.