Elon Musk wants to turn Tesla's fleet into AWS for AI – would it work?
theverge.com
theverge.com
This is basically the same thing as websites running crypto miners in the background, just for cars.
I wonder how he comes up with this stuff.
That's all of it, and as long as you trust he's worked out the details, the trick works.
If you sit down and start thinking about it though, you quickly realize it's utter nonsense.
In this case, the cost per operation will be much higher than if you do it at a datacenter for a long list of reasons I'm too bored to repeat (but if someone wants to know, ask me).
So talking breathlessly about how much compute sits idle and unused in all the Tesla cars is like this joke:
"Sure, we lose money on every T-shirt we sell, but we'll make it up in volume"
My hardware 3 computer @144 TOPS draws 72 Watts.
At low speed drag and friction losses are much lower and range increases beyond the EPA number. Also cities have many charging locations so you never worry about running out.
In my personal case I can charge in my garage and almost never get below 50% charge (only charging to 80% overnight to maintain a 20000 charge cycle life of the battery).
Ridiculous ...
I don't know about what chips Tesla uses but if they're using the right chips and have enough storage on board, this might actually be the first time this idea pays off (other than the plethora of crypto miners).
I mean who hasn't come up with "what if we use peoples' phones, computers, etc. to do compute while they're not using it?!?" as a startup idea?
or battery life from someone else's vehicle they might need to get home?
or additional power from someone else's line?
You didn't even spend one minute to skim the article and yet you are spreading false claims
The idle compute power of millions of Tesla idle vehicles _does not belong to him_. The electricity in their batteries _does not belong to him_. The line power charging those batteries _does not belong to him_.
Of course.