You're viewing a single thread.
It's really idiotic how many resources we're throwing on bad texts, hallucinations and images that look just off. I've got a coding license for my work and I really tried to use it in a meaningful way, but now it's just another tool that's unused because there's only so much mindless scaffolding you need to generate.
It's on average maybe 5x more than a google search. A 300W GPU running for a minute won't be a blip on your energy bill. What is costly is training, which is usually not done often.
That said, data centers as a whole are something like 1% of the global electricity use. So while AI may use a lot of energy relative to what you're getting, it's really not a ton in the grand scheme of things.
If the goal is to reduce electricity usage, there are bigger fish to fry. If the goal is to stop the small fish from getting too large w/o providing value, then go ahead, guilt away.
Yeah, exactly my point. It's more useful than harmful and it is a miniscule impact compared to other things. Enviro people have just latched onto AI as if it's the greatest evil ever, with lots of embelishments and straight up lies about its power usage.
Eh, I'm not so sure about the useful vs harmful tradeoff, but it's certainly not harming anything AFAICT. In fact, they'll be super motivated to get cheap, reliable electricity as it grows, so expect more investments like reopening 3-mile island, solar energy storage, etc. It could end up being a net win climate-wise (probably not), but in any case, it's not what's causing climate change.
If we want to address climate change, we need to address the 500lb gorillas in the room: transportation. Anything else, IMO, is a distraction.
Again, 5x the electricity of a single google search is nothing and it produces incredible results. Overall it is a tiny blip compared to anything else.
mindless scaffolding
So perfect, yet succinct - bravo!
Itâs always at best a starting off point, needing a real human to review and heavily edit. Even in the more refined spaces like surveillance whereâs thereâs a ton of research money thrown at it, itâs still just a pattern recognition probability gambler that needs human oversight.
It's better than spending the money on McKinsey. But not by a lot.