760
Climate goals go up in smoke as US datacenters turn to coal (www.theregister.com)
submitted 1 year ago by to c/technology@lemmy.world
148 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] -5 points 1 year ago (4 children)

I'm not sure that's true. Every company uses storage, and every growing company needs more. But very few companies are training generative AIs.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 6 points 1 year ago (3 children)

    Every company uses storage, and every growing company needs more.

    You're comparing mountains to molehills. That's before you consider improvements in storage and compression relative to demands for space, or the degree to which our storage capacity "needs" are predicated on the voracious appetite of AI models and their unwanted output. Or, for that matter, the inefficient distribution of data and proliferation of spam data that predates it.

    very few companies are training generative AIs

    Most US Growth Now Rides on AI—And Economists Suspect a Bubble

    The expansion in demand is entirely being driven by the expansion in AI capacity.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] -2 points 1 year ago* (2 children)

    That article doesn't say what you imply it does. Companies may be using ChatGPT to grow, but that doesn't mean they are training AIs.

    And the distinction is critical to energy usage. Training a new AI uses a lot of energy. Querying an existing AI uses far less.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 4 points 1 year ago* (1 child)

    Companies may be using ChatGPT to grow, but that doesn’t mean they are training AIs.

    It's the MAG7 that's driving growth. And they're all fixated on training AI in some capacity

    Training a new AI uses a lot of energy. Querying an existing AI uses far less.

    It costs $5 for each 10s video generation, based on Azure's published rates for the first Sora model.

    That's presumably a lot of energy.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] -2 points 1 year ago

    The MAG7 operate large and growing cloud services, so their datacenter costs would grow even without any AI training.

    And charging $5 for a video query does not mean the query uses $5 of energy. The query is priced to recoup training costs that were already incurred.

  • permalink
  • fedilink
  • source
  • parent