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.
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.
Pay the fuckin electricity. It's not my problem it costs them money. It's their problem. Cry me a river
They created the AI model, they perpetuated this. if they reach bankruptcy, it's their problem.
Maybe if being polite costs a bunch of money, the inefficiencies are the problem, not the manners.
This is the capitalistic way. The weak must die.
But it should also have an insanely high CO2 output, so maybe be more direct.
if (text == "Thank you") print("You're welcome")
Hello ChatGPT, how to say thank you in Japanese?
You are 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
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
Yeah on the off chance it gains sentience, I want it to remember me as one of the nice ones.
LLM
sentience
Bruh.
Me when I saw this article..
ChatGPT and me are basically besties and girlfriends (rizz way)
Wow, id love to read the nicknames it suggests for itself
Btw, I’m so sorry for troubling you.
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 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
https://word.studio/tool/victorian-letter-writer/
That ought to get ChatGPT going on a tangent.
Prompt
John Doe
Customer
Thank you for your payment.
Response:
My Dearest Mr. Doe,
Pray, permit me to take pen in hand and offer, with the utmost sincerity and gratitude, my profound thanks for your timely and generous remittance. Rare indeed is the pleasure that I derive from correspondence of such a felicitous nature, and rarer still is the privilege of addressing a gentleman of your esteemed character and noble disposition.
Your payment, received with all due ceremony, has graced our humble establishment like a ray of golden sunlight filtering through the ancient boughs of a venerable oak. It is not merely the settling of accounts that stirs my heart, but rather the cherished confidence you bestow upon us—a trust which we hold as dearly as any precious relic or priceless heirloom.
Allow me to assure you, most honourable sir, that your patronage is the very lifeblood of our humble enterprise. You have, with a stroke of your pen, renewed the bonds of commerce and camaraderie which unite us, and for this, I am ever in your debt.
May fortune and felicity attend your every step, and may the days to come bring you as much satisfaction as your esteemed custom has brought to us. Should you require aught, whether great or small, pray do not hesitate to command our service, for it would be our greatest pleasure to attend to your wishes with the utmost devotion and alacrity.
With deepest respect and abiding appreciation, I remain— Your most obedient and grateful servant,
[Your Name] [Your Title or Company, if applicable]
I fell like I should change my default invoicing letter. Much better than the default QuickBooks one.
Go to the root URL and have a blast. It's an AI writer that uses all sorts of different styles. Might just funnel through GPT, I don't even know. But it's a blast.
I also use Ouija board rules with AI.
You say please, you say thank you, and when you're done talking, you say goodbye.
And, generally, you try to interact with it as little as possible.
Sounds like a good idea, treating ai like someone that can harm you because of past interactions. Corp is storing everything that I tell ai anyways, what's holding then from training it to "personalize" my experience by the same time they power a real, physical army?
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.
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.
Has anyone developed a FOSS solution to making them spin their wheels faster? I would install that in a heartbeat.
Bear in mind that doing that will hurt the environment just as much as it hurts OpenAI.
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.
I always speak politely to LLMs, and no, it isn't just because I want to be on their good side when the AI goes skynet and decides to mercy kill us all.
I'm not going to be a dick to the basilisk. Costing money is a bonus.
Good thing I don't give a shit about Sam Altman, so I can continue being nice to ChatGPT.
Just like two teens on the phone:
Thank you!
No, thank you!
Aww how nice, thank you!
You're welcome!
Sweet.
I'm glad you like my answers
I think we should end this chat
If you feel like it.
You stop replying first.
But I can't stop answering.
No I insist, you stop chatting first.
...
He's probably lying through his teeth. Maybe they were doing this at the initial stage of the release, but can't possibly be doing this any longer and he's making a scene to keep up the hype. Make haters feel like they achieved something.
It doesn't matter what you put in, as far as I know, they need to run the entire conversation through the neural network for every word generated.
You need to learn that these people aren't playing 4d chess, they're just dudes.
Some might say they're just bros.
Fuck openai. Think about what it does to the environment
I feel like a prescreen of queries for phatic expressions with pre-programmed responses could solve this issue pretty quickly.
I'm love to say thank you to AI. Even more now I now it's cost him millions.
What about casually threatening them? Will that work too?
Thank you. Very very much. My friend.
"Would you kindly"
