Yes it is a bubble and will pop. Yes it can be done sustainably. I literally have a local LLM made by Alibaba (Qwen) that runs on my laptop NPU and is still pretty useful for Linux debugging. Now it's made to be the like normal chatbot shit which is dumb, but it works. If you took a similar sized model and trained it specifically to be a helper for a specific distro (like train it on Arch Wiki and not on like all the other random shit they train it on) then it would be quite useful. But the thing is it isn't really profitable the way they think it will be. I'd compare it to like software in general.

I think we should push back on the AI term anyway. That's marketing BS. They are machine learning data processors. Their outputs are determined by their inputs. The viable economic strategy for them is to treat each LLM model more like a piece of software. You purchase X model and it does a specific thing. Or it comes bundled with something else. I think Apple will likely make a local model that runs on their Macs as a tech support helper at some point. And it'll just be trained on like all the issues that come up with the Mac and can download updates where if 1 person has an issue it'll update and suddenly all the machines will know about that issue. Or like how Photoshop has the "AI" tools that auto fill images and crop things for you.

Those are genuinely useful and don't take a data center to run.

What has basically happened is tech companies put spellcheck on crack out there as AI and are now trying to convince the world it can be the next stage of human civilization. When in reality it's best used as what it is. Spellcheck but for other things. You remember how back 10 years ago you had that thing where your phone would reccomend the next word for you to type and you'd click one of 3 or 4 options? That was an LLM. The tech isn't new. They just decided to pump so much data and energy and compute into it that it became sentient and it didn't work. So now they're trying to convince everyone it DID work so they don't go broke.

the ML field is sustainable and productive... LLMs and LLM derivatives are not. they have an inherent use case which is why they'll stay around (unlike NFTs), but there are two factors that lead me to not be concerned about their climate impact for now:

1. all these big numbers you see about the AI buildup and its impact will probably never happen because i think the enthusiasm runs out and kills the bubble before oversupply takes over and pops it.

2. chinese models are far more efficient and cheaper to run while being open weight, i think they will absolutely dominate unless US companies also make cheaper models. regardless in both scenarios the compute needed falls drastically. granted they are still not sustainable, but they are better in that regard.

i think the available infrastructure right now can perfectly handle the natural demand for this tech at this point. the issue is that they keep selling this idea of "AGI" and total AI domination, which is not going to happen, at least with transformers and LLMs. there's also the structural issues with ML approaches, like past is not a reliable indicator of the future and not everything can be optimised and represented by an objective function.

Yeah, the Chinese ones are well optimized and use way less power. The power itself doesn't need to come from fossil fuels. They could run them not as hard and use something other than ground water to cool them.

So I fact it could easily be done better

If by AI you mean LLMs and chatbot stacks? Not without a major breakthrough in either hardware or software architecture. In order to remain dubiously useful they need constant retraining as they contain no information between training start and the current date (there are ways to fiddle around this but it's basically all just appending new things to a prompt and re-running).

Diffusion algorithms? They're not the most inefficient thing. They're just pointless. If you want to see structured colour devoid of meaning press on your eyes.

I think nobody has demonstrated economic viability or particular usefulness. The technology exists, it'll probably cling on in some ways, but eventually they're gonna run out of investor money and someone is gonna be left holding the bag. The financials are all incestuous

Yeah you could make it sustainable for sure. I think in a rational economy the tech would only find its way to places where people actually want it, and the data center requirements would be a fraction of what they are now, and people wouldn't be nearly as annoyed by it because it wouldn't be getting actively propagandized to the tune of hundreds of millions of dollars. Even major corporations would be more careful about it because there wouldn't be an infinite money tap subsidizing its use - the economic incentive would be for fast, specific local models rather than massive, slow, and expensive cloud ones.

That said art communities would be still mad about image generation, schools would still need to figure out how to prevent it from breaking essay writing forever, and the whole question of massive copyright violation that occurred to make the tech would still be unresolved.

I think it's somewhat sustainable if it's locally hosted on your own computer with your own data and not just a constant energy drain/pollution factory

If it was heavily regulated and only accessible by scientific bodies

In response to the thread title.

Now, this isnt meant to excuse AI and it is certainly making things worse but to be brutally honest, climate change is already irrevocably and utterly fucked. Nothing short of some form of global, hardcore Eco-Communism is going to turn that wagon around or slow it down in any large scale way. So yeah, Im pretty much at peace with it and I don't really have an answer to it. If im seriously missing something anyone feel free to correct me.

Now, to your second question. I assume the AI bubble is going to pop or at least deflate at some point and various AI companies will probably go under. With that said.... I dont think that means the end of AI. It probably just means the industry consolidates into fewer companies. So AI slop and machine learning are probably just going to be part of the future we live in. Pure speculation on my part.

I hope I dont come of as a doomer but thats my 2 cents.

In terms of the environment I don’t believe there’s anything about AI that directly necessitates environmental destruction. All that electricity could be produced sustainably and afaik cooling could be sustainable too if better methods and infrastructure were in place.

I don’t think the tech will completely disappear but I think it has a long way to go to a fraction as useful as the AI evangelists claim.

Bubble but a ine that will be pumped way bigger than NFTs. I am not a nerd, I stomp nerds, but from my understanding the stuff people put under the umbrella of AI is stuff thst has existed and was useful subjected to such a large dataset to become fucking weird. The idea behind it has it's uses when fed specific data and programmed to do a specific thing but obviously the current consumer applications are absolutely bad and awful as well as more or less useless and unwanted by a majority of people.

Tldr people shouldnt be computers and computers shouldnt be people. Let them do their jobs to make ours easier. It's being used as mechanical perversion of the human spirit and drinks lakes to do it. I think a lot of people just wanna build a god of destruction before the planet shows us who's really in charge because they are terrified of forces beyond their control to the point they will kill us with forces under their control because power is and end unto itself even in the face of oblivion to these unthinkable monsters.

In growth markets, efficiency gains are redirected towards production.

China has mandated that new data centers source at least 80% of their energy from renewables and is actively building its massive computing infrastructure in regions rich in wind and solar power.

Return to Clippy seems to be the common sentiment.