Idk, maybe I’m a big idiot, but I have issues with the points made in this article. I will concede the point that individual Chatgpt use is not that big of a deal, environmentally speaking. If I can trust the numbers in this article, then it has successfully convinced me that I dont need to worry about the energy cost or emissions or water use of individual prompts.

The case I take issue with is the author’s point that LLMs are inherently useful. I don’t care if chatgpt is kind of just a better Google. I still hate every other thing about it. Using Chatgpt is clearly developing psychoses in some people, and even for people like this author that can use it responsibly, I think it’s just intellectually lazy. It encourages the user to abandon critical thinking and let the robot do it for you. What’s more, as a search tool, it’s destroying the internet. If no one ever goes to websites to read the info, why would people keep making websites with reliable information? Why should I even read this article? Why don’t I just have chat summarize it for me, and I never give this author any traffic or money? Then of course there is the plagiarism problem…

So idk. Maybe now I’ll stop harping on the environmental point. But I’m still going to avoid LLMs like the plague, because at their core I think they rob us of some of the finer points of a life well-lived. I’d rather spend my time poring through articles to understand the why and how of a question, rather than have a robot just spit the “what” out at me.

I think in terms of energy usage on this issue, as per usual it’s not so much about individual consumption habits as it is corporate spending and ecologically destructive habits.

• Collectively, big players in AI are ramping up the construction of AI datacenter megaprojects that will consume massive amounts of electricity and water. If managed well this will be in areas of the world where there is electricity and water to spare, or added to grids that will be expanded in responsible ways. Utility companies in the US often seem all too happy to put the cost burden from expanding infrastructure onto individual users, instead of the massive corporation who wants to build the massive thing. This is bad when most people are barely scraping by.
• Training AI uses a lot more resources than using it when it’s done. Developers of commercial AI models such as OpenAI spend a ton of resources on models that will never be used, and these practices are fairly opaque. Additionally, it’s very hard to measure the full cost of developing a model against that of any individual query because a lot of usage data isn’t public.
• The point of AI is to replace workers. So so many people are struggling as a result of broader economic issues in addition to the new tech putting them out of a job, in a deregulated economic system where it’s increasingly work or die. This technology is not unique in the regard that it replaces human jobs, but it threatens increasingly more people, and increasingly centralizes resources and control into fewer and fewer hands. I don’t think that’s particularly good or healthy for society.
• The AI industry now is trying to do what Uber did over the last decade: beat out the competition, monopolize, and enshittify to squeeze its users for better returns every quarter until the end of time. Using ChatGPT may be fine now, but if it’s the only game in town I guarantee it will start costing a lot more and get a lot worse too. OpenAI in particular has not turned a profit since it went public and doesn’t expect to turn one until 2029.

Most importantly, using these tools signals to the companies making them that there is demand for them to tap into. Your use is not significant on its own, but in so using it you are making yourself into a part of their market. That gives them more justification and green light to keep going, and to waste ever more resources on ever marginal improvements.

I don’t think the issue is primarily where things are at now in terms of resource consumption by AI, but that the industry shows little to no signs of slowing down let alone stopping without popping the catastrophically large investment bubble it’s developed into. No matter how this ends, it likely doesn’t end well.

this is such a dumb take.

“if you are talking to a chatbot you arent doing other things like driving!”

Watching the anti-AI crowd get played and manipulated over the past year or so has been really sad. It reminds me a lot of the early days of the anti-GMO movement when legitimate criticisms and efforts at regulation were supplanted by fear mongering and the whole thing got transformed into more of cult or a kind of subcultural identity. You can see the same pattern playing out where they increasingly are sprouting disproven nonsense, actual substantive criticism is left behind for whatever sounds scariest and/or makes them feel righteous. They are on the path to discredit themselves and lose the fight for AI regulations just as badly as the anti-GMO movement lost.

Thought this was an interesting look at an oft-repeated claim, that AI is bad for the environment. I think it’s assumed to be true, especially on Lemmy, but that might not actually be the case.