Quote:
Originally Posted by redpoint5
Which is the brilliance of Comma AI, and why it's the only autonomous system that's profitable.
|
Comma AI is not an autonomous system. You can't take it and put it on a car without an autonomous system and have it drive. It is like a piggy-back ECU in that all it does is trick the signals in the factory system to get a different result.
The Level 2 systems automakers are putting in their cars are profitable.
Quote:
Originally Posted by redpoint5
Musk's mistake was starting at level 4, which as you and I both agree, will not be achievable this decade.
|
Musk's mistake (or brilliance depending on perspective) was selling a system that will never be more than Level 3 as Level 5
Level 4 is technically achievable today and there are Level 4 systems on the road today driving. I just don't think the cost of Level 4 will get to the point were it makes sense for personal vehicles in this decade.
Quote:
Originally Posted by redpoint5
Comma AI gets around legal issues by only selling the hardware, which can accomodate open source software. It seems they are somewhere between level 2 and level 3 automation.
While people are killing themselves by not assuming responsibility for driving in their Teslas, Comma AI requires driver attention. No water bottle in the steering wheel trick.
|
Reading their website Comma AI is very clearly Level 2, only if the base vehicle is Level 2 and limited to highway use.
Quote:
Originally Posted by redpoint5
I'm surprised at the push for level 4, as that shifts liability. I'd work at perfecting level 3 as a goal since liability remains with the driver, and it's simpler to accomplish. Level 4 is a problem for the 2030s, and level 5 should never exist. There is always need to occasionally assume control.
|
There is a push for Level 4 because that is were the money is there to be made. The commercial space is where the money is at. An over-the-road truck with no driver to pay and no 11 hour per day limit could mint money for the fleet owner. Interstate driving is also the easiest use case for autonomous driver tech to meet.