4 ms·

Lol what. It's like you'd be trying to convince everyone Bing has been the leading search engine for the last 10 years. You don't even mention who would be ahea

by maxlin 3y ago

Lol what. It's like you'd be trying to convince everyone Bing has been the leading search engine for the last 10 years. You don't even mention who would be ahead. Cruise, having to intervene every 5 minutes (literally, look it up) and basing on unscalable HD maps? Get real man. Tesla has even custom HARDWARE for their NN's even, designed by the chip designer that has created some of the greatest designs BOTH at Intel and AMD. And Tesla's battery lasts better than others. I remember seeing incredible endurance even for the first Model S batteries after like 10 years of driving. You buy a Tesla now or 5 years ago, it will likely last longer than any other car, all while it will receive SW updates to function as a Robotaxi when the time comes. You're funnily arguing the exact polar opposite of where Tesla is strong.

IshKebab 3y ago
> Tesla has even custom HARDWARE for their NN's Their chips are actually nice, but you're betraying your lack of knowledge here - the compute hardware isn't remotely the hard part about self driving cars.
- maxlin 3y ago
  Using words like "remotely not the hard part" that is just not true man. There doesn't exist hardware that can do inference with even close to as low as the 100 watts Tesla uses. Also it is not just to drive FSD, it is to provide data, to detect situations of from where to send data as they can't just stream everything to base (check the Autonomy day talk, it's quite interesting), which is needed for the development of the software which we agree is an extremely hard problem. To further demonstrate this, consider the fact that in the current AI landscape, hardware is actually the bottleneck. Everyone is trying to buy more GPU's NVIDIA simply can't supply fast enough. If Tesla still relied on NVIDIA for the in-car processing, the project would have to be shelved as they'd have to sell Teslas without the hardware, instead of just having it all cars even if they will never use it for running FSD directly on that car.

by maxlin 3y ago

IshKebab 3y ago

> Tesla has even custom HARDWARE for their NN's Their chips are actually nice, but you're betraying your lack of knowledge here - the compute hardware isn't remotely the hard part about self driving cars.

maxlin 3y ago
Using words like "remotely not the hard part" that is just not true man. There doesn't exist hardware that can do inference with even close to as low as the 100 watts Tesla uses. Also it is not just to drive FSD, it is to provide data, to detect situations of from where to send data as they can't just stream everything to base (check the Autonomy day talk, it's quite interesting), which is needed for the development of the software which we agree is an extremely hard problem. To further demonstrate this, consider the fact that in the current AI landscape, hardware is actually the bottleneck. Everyone is trying to buy more GPU's NVIDIA simply can't supply fast enough. If Tesla still relied on NVIDIA for the in-car processing, the project would have to be shelved as they'd have to sell Teslas without the hardware, instead of just having it all cars even if they will never use it for running FSD directly on that car.