By virtue of even posting, YOU are pushing datacenter demand. These comments don’t just live in the ether.
I write software for a living. There’s an enormous difference between a text-based service that functionally runs as a primitive CRUD and a compute-heavy service like LLM inferencing. The latter runs on servers that individually consume multiple kilowatts while the former can run on a raspberry pi but gains efficiency by sharing a server that consumes maybe 1kW with dozens of other people. Heck, Lemmy doesn’t even need to do video transcoding like all the toxic social media sites do.
The point being you are MORE upset about the lesser evil, not that one is justified evil and the other isn’t.
No, people are upset about all the evils. Being more or less upset about one or the other is irrelevant. You’re just choosing to cherry pick one thing that everyone who’s upset should ignore.
Mordikan@kbin.earth 3 hours ago
Can run on a raspberry pi and does run on a raspberry pi are two different things. I don't think raspberry pis are the solution here. I'm not cherry picking either as both are clear evils. That is the point. Current non-AI datacenters are using more power in total than AI datacenters. That may (and likely will) change, but the argument that AI datacenters are destroying the environment when the other is destroying it even more (but let's ignore that one) is moronic by nature.
That by definition isn't cherry picking. It's pointing out the logical fallacy in popular thought.
I work in devops. There is little difference in current overall consumption rates.