A growing body of research attempts to put a number on energy use and AI—even as the companies behind the most popular models keep their carbon emissions a secret.
  • @AlteredEgo@lemmy.ml
    -1•24 hours ago

    It’s not that complicated. If you buy an old server and put ~1500 GB of ram into it, you can run full deepseek on it, comparable to ChatGPT. It’s slow though, like a few words per second. That would be less than 1000 watts of power, only for when you actually use it. That is about 5 square meters of solar panels. Or if we finally make wind power through kites mature it would be about a 2 square meter kite and a winch. Plus some batteries that can be infinitely recycled. With hardware advances hopefully we’ll see cheaper and more sustainable ways of running AI.

    We can generate a near infinite amount of power in sustainable ways, the problem at what times, and how inefficient capitalism is in managing resources.

    The question seems to be a drive to regulate access to AI, similar to the “muh copyright” talking points. AI is a resources that ideally they want to monopolize, and regulations that limit who can use it are the standard way of doing that. You can’t electrify your own car because of regulations. Thank of the children! So I suspect there is a kind of struggle going on who gets to reap the rewards of using and commercializing AI. I bet it will end up somehow impossible to run your own AI models at home because if IP or energy laws.

    How much energy do gaming PC use? How much energy do playstation and consoles use? How much energy does online gaming use? Social media? Asking this question specifically about AI is suspect.