This has been so obvious for so long. Just look at open weights offline AI capabilities. A consumer cutting edge 24 GB GPU is still nowhere near powerful enough to make executive decisions at this scale, speed, and scope. Maybe it is possible with one or a few A100’s, but those are $10k monsters that drink enough power to justify data centers investing in nuclear power plants. That’s not going into some ultra light EV with lithium cells.
This has been so obvious for so long. Just look at open weights offline AI capabilities. A consumer cutting edge 24 GB GPU is still nowhere near powerful enough to make executive decisions at this scale, speed, and scope. Maybe it is possible with one or a few A100’s, but those are $10k monsters that drink enough power to justify data centers investing in nuclear power plants. That’s not going into some ultra light EV with lithium cells.
If you don’t use ai, then self driving autonomy is likely feasible in the compute power available in most of these cars.
The problem is sensors (like lidars) and software, that’s highly tested and validated. The testing and validation is the real issue.
Didn’t Tesla decide to go completely with cameras? No other sensors, just some rudimentary version of machine vision?