you are viewing a single comment's thread
view the rest of the comments
[–] 3 points 6 months ago (14 children)

You already shouldn't be using Chrome, and if this is what moves the needle, great.

At a billion-device scale the climate costs are insane.

No.

Just flatly, no.

Local models spin your GPU like a video game. Unless you think Overwatch is a climate disaster, please learn to separate datacenter condemnation from people running their own computers a little harder.

  • source
  • hideshow 14 child comments
  • [–] 6 points 6 months ago

    Agreed. Hate the data centers and corporate obsession with extreme waste for total value capture. Not the tech, especially on a local scale.

    A 128GB amd strix halo runs at 120W and can host large LLMs at home. If you ran it full time for inference it would still be under $20/mo in electricity costs.

  • source
  • parent
  • [–] 14 points 6 months ago (12 children)

    I'd say you should read the article a little more closely, but it's not written very well. But it brings up interesting things that have nothing to do with your local GPU usage. For example, it names an interesting point about simply delivering 4 gigabytes of data to that many people. If pushed out to ~15% of Chrome users without consent:

    • That'd be 500 million people
    • It would be 2 exabytes of data
    • 120 GWh of energy, equivalent to the annual electricity consumption of about 36,000 average UK households
    • 30,000 tonnes CO2 emitted, roughly the annual emissions of 6,500 cars

    And that's just for the initial data push. Models need ✨updates!✨

  • source
  • parent
  • hideshow 12 child comments
  • [+] -10 points 6 months ago* (10 children)

    We must kill the environmental disaster that is Steam.

    Youtube is history's greatest monster.

  • source
  • parent
  • hideshow 10 child comments
  • [–] 12 points 6 months ago* (9 children)

    Installing a game you want is different than hundreds of millions of people having something they didn't ask for getting pushed on them.

  • source
  • parent
  • hideshow 9 child comments
  • [+] -7 points 6 months ago (8 children)
  • [–] 1 point 6 months ago* (5 children)

    What happened to separating personal use from condemning data centers for expending unnecessary and unwanted energy?

  • source
  • parent
  • hideshow 5 child comments
  • [–] -4 points 6 months ago (4 children)

    Does consent change how much power a server uses?

  • source
  • parent
  • hideshow 4 child comments
  • [–] 2 points 6 months ago (3 children)

    Here, yes. Two exabytes of data transfer could have been one or zero.

    I don't get the point you're trying to make here.

  • source
  • parent
  • hideshow 3 child comments
  • [–] -1 points 6 months ago (2 children)

    The point is that servers don't belch black smoke when they send you one file. This model is the size of a four-hour Youtube video. How many people watch how many hours of video, every single day? We only see this hand-wringing minutia over internet use when talking about neural networks, and it's getting weird.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 4 points 6 months ago (1 child)

    Much weirder when people try to shift blame off corporations pushing stuff on people without their consent, and on people minding their own business.

    Weren't you just telling me that data centers would use energy regardless anyway? I can't keep track of these talking points, except it seems like they're all pro AI.

  • source
  • parent
  • hideshow 1 child comment
  • [–] -1 points 6 months ago

    The root comment opens with 'don't use Chrome.' This tangent is about specific overblown fixation on power use... for downloads.

    I am telling you to apply your own criticism of bandwidth to anything else Google does. Four gigs to every desktop Chrome user is still a drop in the bucket compared to a streaming service. If the average Chrome user has watched two movies online, they've done just as much environmental damage. Which is to say: not much.

  • source
  • parent