I want a Cruise-equipped Camaro with different drive modes: Comfort/Eco to commute; Sport would be like riding a long with a good driver: quick launches, higher Gs in the corners. Would be a total blast!!
200k per day is less than 80 million per year. meanwhile waymo has driven several billion virtual miles already. Tesla has a billion real world miles (mostly highways)
Yeah, and this is just a rendering of the paths the AI agents took based on very clean data, and not likely the actual simulators output. Every AV can drive perfectly in perfect noise-free simulations. Real world data is the key to really doing it though. And if they are limited only to areas that are pre-scanned, their a.i. isn't that good as pre-scans mean the pipeline isnt actually interpreting new situations. It compares its internal mapping to the pre-map, then drives as instructed, taking into account obstacles. The second it goes into a new scene, its a dud because its not actually able to interpret it and look up any pre-defined paths - because they dont exist yet. The real self driving tech doesn't need to know anything about what the environment looks like until it encounters it for the first time. This is why Waymo and Zoox etc. are stuck in certain cities. They have not seemingly created a generalized solution, and rely on humans to pre-map first, define the route, then the bots navigate through.
Ask yourself. "Would I fly in an airplane with nobody in the cockpit" no you would not. Same here. Imagine getting in of of these and it runs a red light. Jeez, what a waste of money.
I really thought this was real for a minute. That's incredibly photo-realistic.
I want a Cruise-equipped Camaro with different drive modes: Comfort/Eco to commute; Sport would be like riding a long with a good driver: quick launches, higher Gs in the corners. Would be a total blast!!
200k per day is less than 80 million per year.
meanwhile waymo has driven several billion virtual miles already. Tesla has a billion real world miles (mostly highways)
This is a great simulation! Keep advancing guys!
The outdoors are messy. Simulate more outside.
Launch?
No turn signal at 1m14s ?
The virtual environment looks way too clean
Yeah, and this is just a rendering of the paths the AI agents took based on very clean data, and not likely the actual simulators output. Every AV can drive perfectly in perfect noise-free simulations. Real world data is the key to really doing it though. And if they are limited only to areas that are pre-scanned, their a.i. isn't that good as pre-scans mean the pipeline isnt actually interpreting new situations.
It compares its internal mapping to the pre-map, then drives as instructed, taking into account obstacles. The second it goes into a new scene, its a dud because its not actually able to interpret it and look up any pre-defined paths - because they dont exist yet. The real self driving tech doesn't need to know anything about what the environment looks like until it encounters it for the first time. This is why Waymo and Zoox etc. are stuck in certain cities. They have not seemingly created a generalized solution, and rely on humans to pre-map first, define the route, then the bots navigate through.
That sawed-off car isn't big enough to be a cab ...
Using LIDAR is nonsense plus you lose much precious aerodynamic thus lowering the range on EV's
Ask yourself. "Would I fly in an airplane with nobody in the cockpit" no you would not. Same here. Imagine getting in of of these and it runs a red light. Jeez, what a waste of money.
4 fucking lidars in a single car? That's unnecessary and price ineffective.
Sorry it was 5
Simulation is not as good as real world miles from Tesla production fleet in customers hands.