Bloc wanted to share this , so I carried the torch.

  • RamenJunkie@midwest.social
    link
    fedilink
    English
    arrow-up
    2
    ·
    edit-2
    6 days ago

    You can avoid the environmental impacts by using locally-hosted AI models.

    Does thst reply not realize that is just spreading the impact around and not really reducing it? If anything it makes it worse because your local NPU/GPU lacks the efficiency of scale that say, a ChatGPT data center would have.

    • Swedneck@discuss.tchncs.de
      link
      fedilink
      English
      arrow-up
      1
      ·
      3 days ago

      I think it’s worth considering that running locally places a limit on how resource intensive it can be (it can’t magically draw more power than your PSU can supply, and your components are only so capable) which means people will adjust their usage to fit within those limits.

      It’s like how making people drive longer distances is actually good for the environment, because it ends up with people only driving when they REALLY have to.

      • RamenJunkie@midwest.social
        link
        fedilink
        English
        arrow-up
        1
        ·
        2 days ago

        There is an argument to be made for peoppe getting impatient and using it less. But if it takes X Watts (or whatever) to run a prompt, and its bottle necked by your PSU, its just going to run longer to get there, burning more excess heat etc. Vs a controlled data center space designed to run these sort of quieries. Its not like they have millions of literal consumer desktops in an office space running these systems.