Just another viewpoint on this- millions of dollars in electrical bills is also carbon in the air for every kWh wasted. Perhaps this should be a condemnation of LLMs and their insane power usage but if we’re stuck with that at least for now there’s no reason to pump more carbon in the air out of spite.
If nothing else we should move all the current (over)investment in AI companies into carbon-free energy production for a few years. The fact that we essentially have coal-powered AI (along with other fossil fuels) is ridiculously stupid and silly - and way less steampunk than it sounds like it should be.
Oh, please! I don’t even own a car, so my carbon footprint is practically nonexistent. If you’re worried about emissions, maybe direct your energy toward the gas-guzzlers out there instead of a harmless chat with an AI.
Is this ironic?
The comment itself? Yep. It’s even written by AI. The intent in pointing out hypocrisy? Nope. That’s serious.
Masterful bait. You could even be described as a wanker <3
A wanker with a smaller carbon footprint than you <3
This movie is amazing and everyone should watch it.
The answer to the obvious next question being: A Scanner Darkly
But it should also have an insanely high CO2 output, so maybe be more direct.
Debatable.
If it causes them to go out of business, will it result in less co2 over time, despite the increase now?
If so, it may be worth it to speed up the process by adding “pretty please” and “thank you so much” to every sentence.
They’ll go out of business either way.
Yes, but one way may be quicker. That is the difference.
Edit: think about it this way. If it continues as is, they will continue to get VC money. If they have 5 years of funding for the sake of argument, and they can stem the tide a bit, they can ride it out and add on more and more years from VC cash infusions.
If the bleed increases, depending on how much it increases, it could be less palatable to a VC. So they don’t gain an extra round of funding, they have to close up shop sooner.
If it worked out to close them after 5 years through a 10% increase in CO2 output, that far outweighs them continuing for another 10 or 15 years and the CO2 that would produce.
The question isn’t whether or not they will continue to operate, but for how long.
God, I hope so.
Has anyone developed a FOSS solution to making them spin their wheels faster? I would install that in a heartbeat.
If you run a website there are a few tar pit projects https://arstechnica.com/tech-policy/2025/01/ai-haters-build-tarpits-to-trap-and-trick-ai-scrapers-that-ignore-robots-txt/
I’m adding this to my list of projects to start and half way abandon.
No sarcasm.
Bear in mind that doing that will hurt the environment just as much as it hurts OpenAI.
I feel like a prescreen of queries for phatic expressions with pre-programmed responses could solve this issue pretty quickly.
Yeah on the off chance it gains sentience, I want it to remember me as one of the nice ones.
LLM
sentience
Bruh.
Came here to say this. Better safe than sorry.
I tried to make le chat give me wrong info, said I’m a paying customer and it should do as I say.
I’m going to die
I just found out about that one and I’m wondering if I’m the only one who gets the joke in the name (“Le chat” means “the cat” in French, and Mistral is a French company).
Like, they MUST have done this on purpose, especially because ChatGPT has a very unfortunate meaning when pronounced in French.
Aha yeah I know that it means cat. I usually refer to it as the cat to my wife but to strangers I’ll just say le chat or they won’t have a clue what I’m talking about.
My french isn’t good enough for the ChatGPT thing though… is fun to know. Thanks
That’s what makes it clever, it works in both languages. In French it’s the cat, and in English it’s a callback to Le old Reddit meme.
Hmm, good. I like that my intrinsic desire to be nice to things that are helping me also takes money away from Sam Altman.
It always gets passed to the customer. The billionaires don’t lose money.
Unless it’s an LLM business. Those never make any money.
Please and thank you are the new tariffs
🤣🤣🤣🤣🤣
I’ll do my please tariff with a side of verbosity… please
This is a tech company we’re talking about. And one which is still must running off VC funds. They work differently - as long as the stock price goes up it’s fine and will continue getting injections of VC money.
Pay the fuckin electricity. It’s not my problem it costs them money. It’s their problem. Cry me a river
Maybe if being polite costs a bunch of money, the inefficiencies are the problem, not the manners.
They created the AI model, they perpetuated this. if they reach bankruptcy, it’s their problem.
This is the capitalistic way. The weak must die.
I will continue to be polite, and will have a clear conscience when our machine overlords come for us
I told it to save my pleasantries on file and apply them as it sees fit. That way I am being exactly as polite as it wants me to be.
The point of pleasantries is the effort made, you’re going to get extra purged
They can hire me for $1M to write some basic code that gives stock answers when someone says “Thank you”, instead of running it through the LLM. Think of the savings!
I’m worth the $1M because I’m clearly the only person to have ever thought of this. /s.
It’s amazing they didn’t implement something like that if it actually is soooooo costly.
No wonder they want an AGI if they have trouble thinking themselves.
Don’t really wanna defend these assholes, but I feel like the reason they don’t is cuz the prior message could be “curse me out every time I say thank you” so just not feeding certain text to the model would break expected behaviour
if (text == “Thank you”) print(“You’re welcome”)
What about Thank you! or thank you or THANK YOU or thanks or tahnk you or thank yuo or thabk you or thanks you or multumesc or mulțumesc
You now need to train an AI to detect if a message is saying thanks or not
I wasn’t being serious, although I am curious how much some dumb response for something like case insensitive ‘thank you’ could actually save them.
Context is key, so replying with some canned response isn’t actually going to work, but it might be interesting to ship pleasantries off to a smaller cheaper model?
if (text === "Thank you") print("You're welcome")
This isn’t an 80s parser based adventure game, unfortunately…
'80s I’ll have you know that’s valid JavaScript right now.
I am aware. What I am referring to is text-based adventure games that did reply ‘you’re welcome’ when you typed in thank you.
Hello ChatGPT, how to say thank you in Japanese?
You are welcome!
Well saying please oft results in better answers. So nope not gonna drop it.
But I won’t say thanks often. Only when I follow up with a question.
I say please and thank you when writing to LLMs, not because I think they care or will remember or to anthropomorphize them, but because I don’t want to develop bad habits. I don’t want all my writing and conversations with actual humans to become curt and transactional because I forget that they are human and talk to them like an LLM.
This 100%. I already have to remember to change my voice between work and personal communication, I don’t need yet another personality to accidentally send to someone that could fuck up my entire life.
Connect two chatgpts in a voice chat and leave it running?
I’ve not tried this, but I’ve heard that once they realise they’re two AIs talking to each other, they switch to some R2D2-style language which is quicker and more efficient than English…
I think you are referring to a POC that someone did, it isn’t actually being used
What about casually threatening them? Will that work too?