Generative AI

119 readers

1 users here now

A community about news, questions, development and all related subjects of Generative AI

founded 1 year ago

MODERATORS

[email protected]

Generative AI’s environmental costs are soaring — and mostly secret (www.nature.com)

submitted 9 months ago by [email protected] to c/[email protected]

4 comments fedilink hide all child comments

you are viewing a single comment's thread
view the rest of the comments

[–] [email protected] 3 points 9 months ago (1 children)

I don't see articles complaining about the impact of data centers powering YouTube, Instagram, etc.

Which is, at least currently, orders of magnitude bigger.

Also, can we stop with the "datacenter used X amount of water"? Where do you think the water's going? It sure isn't despawning never to be seen again.

The water is used for cooling, meaning it either ends up evaporating or it ends up being released back a bit warmer.

[–] [email protected] 1 points 9 months ago* (last edited 9 months ago) (1 children)

There are many many articles out there talking about the problem of energy and water expenditure by data centers! It is one of the many issues underpinning the energy problem.

I found a nice one from the same source, but unfortunately it is not open access. But you can at least see the title/abstract, and I can help you get access if you would like: https://www.nature.com/articles/d41586-018-06610-y

Or, here is an open access article from a different source: https://energyinnovation.org/2020/03/17/how-much-energy-do-data-centers-really-use/

The point I want to make is that the use of energy by YouTube / Instagram is a known issue that is seriously considered when thinking about the future of energy. The article I posted does not ignore this, but rather is a continuation of this line of thought.

AI is more energy intensive in part because of its reliance on heavy GPU calculations, and so a world that expands its reliance on AI will inevitably use more energy. The question of "how much energy?" is a very important one, but, as the article points out, it is a very difficult question without the collaboration of the companies running these services. And they have decided to keep much of the information secret.

As for the water... I am not very knowledgeable about water management myself, but I can imagine that moving water, storing water, cleaning water, having access to water, and many other factors play into this - and all of this information is not going to be captured by a simple number like a volume or a percentage.

This article focuses on water usage in data centers. I have not yet read it myself, but I will try to because I am also curious. Hopefully it is helpful for understanding the details of the problem: https://www.nature.com/articles/s41545-021-00101-w

[–] [email protected] 2 points 9 months ago

My point mostly was, it's a issue of datacenters, not of AI itself, and it feels like the only reason they wrote the article is to get clicks on a buzzword.

But yeah, AI will gobble up insane amounts of energy, though I think it'd be interesting to see how much it ends up saving overall.

It's not like the point of AI is to throw the output into a black hole, it's to optimize workflows.

If I spend 30 minutes writing an article or 15s writing a prompt, hasn't AI actually helped me use less energy to power my computer, my brain and body etc?

Nothing is as black and white as it might seem, and AI is no exception