this post was submitted on 16 Jun 2024
156 points (94.3% liked)
Technology
59422 readers
2855 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- 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
view the rest of the comments
That's cool and all, but we're talking about the AI companies that are trying to get valuated at trillions of dollars and want taxpayers to pay for the upgrades to the grid. The sad part is it's likely going to work
I'm all for having companies pay for their electricity use and their impact on the grid but that has nothing to do with AI.
Llama took 2 600 mWh to train over 6 months and can run on much less than what's needed for gaming. ActivisionBlizzard used 86 000 mWh of energy in 2022 for both the datacenters for their games and the development of them. Yet no one in their right mind would suggest to curb stomp gaming to save on energy.
Openai has bigger costs but they run inference, and having them run it actually makes it more efficient, even though I rather open source models you can run on your own machine.
The clear solution is upgrading to a more robust green energy grid, not blocking innovation.
And if we are going to ban things because of their energy use, there are much better candidates than software. A transatlantic flight takes up 500 mWh, so essentially 1000 people flying to Europe and back use up as much energy as the llama model took to train, a model that has been downloaded 3.5 million times in the past month alone on hugging face (only with the official 8b included, and not counting the other sizes or the thousands of finetunes).
Have you got something to read up on regarding comparisons of energy consumption? Sounds really interesting, but I know close to jack shit about this.
Most big companies publish their energy usage like the two examples above. For the plane bit, I just found multiple people calculating it and coming up with the same number online, so that one might be hot air.
That's completely besides the point.
Blizzard isn't asking taxpayers to subsidize them billions "to advance humanity".
As you say yourself, there are way better models than what is being funded right now, and what is likely to get the monopoly on energy, at our expense.
I'm just stating that "AI" is a broad field. These lightweight and useful transformer models are a direct product of other AI research.
I know what you mean, but simply stating "Don't use AI" isn't really valid anymore as soon these ML models will be a common component. There are even libraries and hardware acceleration support for tensor operations on the ESP32-S3.
I didn't say don't use AI.