HawaiianBrian:

Ya but Tesla's new computer is 21 times more powerful than Nvidia's !!!-(see video in other post) 144 TOPS!!!!!!!! I want tesla to start making GPU,s. This new computer is THE BOMB!

Pegasus is 320 TOPS but it's also 500w - regardless Tesla definitely isn't the highest performing and is far less flexible than Nvidia's solution

fantaskarsef:

Bad Nvidia management, I'd say. Either too expensive or lacking in some other field, which usually is a maangement positioning thing.

They are expensive but Elon is all about vertical integration. The rest of Tesla and Space X are the same way, they design/build basically everything in house. Even if Nvidia's stuff was cheaper I think they'd still go this route.

cowie:

NVidia ha responded Nvidia fires back at Tesla's claim that it created the world's best chip for self-driving cars https://www.businessinsider.com/nvidia-fires-back-at-tesla-claiming-worlds-best-self-driving-car-chip-2019-4

“It’s not useful to compare the performance of Tesla’s two-chip Full Self Driving computer against Nvidia’s single-chip driver assistance system. Tesla’s two-chip FSD computer at 144 TOPs would compare against the Nvidia DRIVE AGX Pegasus computer which runs at 320 TOPS for AI perception, localisation, and path planning. Additionally, while Xavier delivers 30 TOPS of processing, Tesla erroneously stated that it delivers 21 TOPS. Moreover, a system with a single Xavier processor is designed for assisted driving AutoPilot features, not full self-driving. Self-driving, as Tesla asserts, requires a good deal more compute.”

Fox2232:

IIRC, autopilot issue is due to EU and requirement to approve each car model separately... which EU did not. I am not sure about details.

As anything from EU, it is simple to do complicated law 🙂 It is due to the "who is responsable in accident" wich differ from country AND the approval wich differ too. But EU cmpany do it for export (like the toxic product that are forbiden to use in food but are used for export 🙁 ).

0blivious:

All these cars can do is drive in easy situations. Throw them a real curve ball and they typically fail.

Yes for now but they make great advancement, most company do real life test near where i work (the cliff-the road-the gap 🙂 ) and we start to see autopiloted car that performing quite good now.

We are currently at "human drives, car intervenes to prevent accidents". We need to get too "car drives with no human intervention". That's a huge step. The problem is the in-between steps (in most self driving scales) are basically "car drives, human intervenes to prevent accidents". That just doesn't work as if the car's in charge the human is simply not capable of paying enough attention to make a split second intervention to prevent a crash. Yet that's essentially what Tesla want to turn on. Hence it's not surprising that Tesla are having trouble clearing it with EU health and safety.

Dribble:

We are currently at "human drives, car intervenes to prevent accidents". We need to get to "car drives with no human intervention". That's a huge step... The problem is the in-between steps (in most self driving scales) are basically "car drives, human intervenes to prevent accidents". That just doesn't work as if the car's in charge the human is simply not capable of paying enough attention to make a split second intervention to prevent a crash. Yet that's essentially what Tesla want to turn on. Hence it's not surprising that Tesla are having trouble clearing it with EU health and safety.

I thought I read several months back that it was discovered that some Tesla autos were found to be mining crypto using their on-board systems. Can't recall if that was with, or without, the knowledge of the owner. In any case, while I don't disagree with your assessment of where we're at, and where we're unfortunately trying to transition to, I want absolutely no part of that "car drives with no human intervention" concept. Personally, I think it's just a matter of time before someone hacks the living daylights out of these autonomous driving systems and gets someone killed. You know, just because they can. Seems like one downside of Tesla's decision to keep this tech "in house" would be that they're assuming all the responsibility if something unexpected happens; no opportunity to place at least some of that burden on the shoulders of the designer/supplier.