3 ms·

Tesla Car Computers Could Be Monetized in Other Ways

2 points by panrobo 2y ago

keikobadthebad 2y ago
Maybe I just lack imagination, but why would distributed computing be better tied to a battery needed for your personal security and geographically mobile with expensive and unreliable connectivity? It's just nonsense...
- panrobo 2y ago
  I guess the idea wouldn't be to compete with AWS but just explore potential other revenue streams for Tesla owners. Clearly there will be limitations in terms of what can be done with such computer.
  - jqpabc123 2y ago
    Did he actually say or imply there would be a revenue stream here for "owners"?
    - 3cats-in-a-coat 2y ago
      There won't be revenue stream for other owners or Tesla from this, because computing this way is far less efficient than simply doing it locally in a datacenter, with high speed interconnects between machines. Meaning if he sells that at profit, no one would buy it. And if he uses it himself, he'd be losing money rather than buying time in a datacenter or setting up one on their own.
      - jqpabc123 2y ago
        In other words, like a lot of things from Musk, this is more marketing BS than real subsistence.
        3cats-in-a-coat 2y ago
        Basically. There's only one scenario in which I see this being useful. Let's say some world crisis and most of our chip manufacturing capacity is bombed/destroyed or unavailable in some other way, while our computing needs are rising exponentially due to some AI-warfare (economic or military). In this case, every device with compute will become valuable, because compute is extremely scarce and datacenters won't be able to service this. So you'll be able to sell compute from your phone (when hooked to a charger mostly), your laptop, your desktops, iPads, and your cars. But there's nothing special about the cars themselves. And in a relatively normal world, what he proposes is a non-starter. Even my scenario is a bit of a sci-fi thing. It's really inefficient to distribute compute. And while folding@home and seti@home ship independent small units of data for analysis, AI inference tends to be very high-volume data and needing high-interconnect between inference units. Meaning it'll be extra hard to distribute in a meaningful way compared to these well-known processing donation projects.

2 points by panrobo 2y ago

Maybe I just lack imagination, but why would distributed computing be better tied to a battery needed for your personal security and geographically mobile with expensive and unreliable connectivity? It's just nonsense...

panrobo 2y ago
I guess the idea wouldn't be to compete with AWS but just explore potential other revenue streams for Tesla owners. Clearly there will be limitations in terms of what can be done with such computer.
- jqpabc123 2y ago
  Did he actually say or imply there would be a revenue stream here for "owners"?
  - 3cats-in-a-coat 2y ago
    There won't be revenue stream for other owners or Tesla from this, because computing this way is far less efficient than simply doing it locally in a datacenter, with high speed interconnects between machines. Meaning if he sells that at profit, no one would buy it. And if he uses it himself, he'd be losing money rather than buying time in a datacenter or setting up one on their own.
    - jqpabc123 2y ago
      In other words, like a lot of things from Musk, this is more marketing BS than real subsistence.
      - 3cats-in-a-coat 2y ago
        Basically. There's only one scenario in which I see this being useful. Let's say some world crisis and most of our chip manufacturing capacity is bombed/destroyed or unavailable in some other way, while our computing needs are rising exponentially due to some AI-warfare (economic or military). In this case, every device with compute will become valuable, because compute is extremely scarce and datacenters won't be able to service this. So you'll be able to sell compute from your phone (when hooked to a charger mostly), your laptop, your desktops, iPads, and your cars. But there's nothing special about the cars themselves. And in a relatively normal world, what he proposes is a non-starter. Even my scenario is a bit of a sci-fi thing. It's really inefficient to distribute compute. And while folding@home and seti@home ship independent small units of data for analysis, AI inference tends to be very high-volume data and needing high-interconnect between inference units. Meaning it'll be extra hard to distribute in a meaningful way compared to these well-known processing donation projects.