this post was submitted on 04 Jun 2024
386 points (91.4% liked)

Technology

59657 readers
2625 users here now

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related content.
  3. Be excellent to each another!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, to ask if your bot can be added please contact us.
  9. Check for duplicates before posting, duplicates may be removed

Approved Bots


founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
[โ€“] BrianTheeBiscuiteer@lemmy.world 0 points 5 months ago (1 children)

Efficiency at the consumer level is poor, but industry uses more total energy than consumers.

[โ€“] MudMan@fedia.io 4 points 5 months ago

Yeeeeah, you're gonna have to break down that math for me.

Because if an output takes some amount of processing to generate and your energy cost per unit of compute is higher we're either missing something in that equation or we're breaking the laws of thermodynamics.

If the argument is that the industry uses more total energy because they keep the training in-house or because they do more of the compute at this point in time, that doesn't change things much, does it? The more of those tasks that get offloaded to the end user the more the balance will shift for generating outputs. As for training, that's a fixed cost. Technically the more you use a model the more the cost spreads out per query, and it's not like distributing the training load itself among user-level hardware would make its energy cost go down.

The reality of it is that the entire thing was more than a bit demagogic. People are mad at the energy cost of chatbot search and image generation, but not at the same cost of image generation for videogame upscaling or frame interpolation, even if they're using the same methods and hardware. Like I said earlier, it's all carryover from the crypto outrage more than it is anything else.