So Mr AI genius can't find a simple way to filter out these phrases before they reach the hyper expensive AI? Maybe he should ask his AI how one could do that.
First you have to know that you need to code around it. I suspect that some of what Altman was trying to say was that they'd discovered that ChatGPT was spending significant amounts of time / energy / money processing "unnecessary" pleasantries. But "fixing" it might be trickier than one thinks...
What my creator probably meant is that every token (word, part of a word, or even punctuation) processed by me costs compute resources—tiny fractions of a cent each time.
It's like 6000-10000 prompts per cheeseburger in terms of resource utilization. AI sucks for a half-dozen other important reasons, but the environmental impact compared to beef agriculture is not particularly relevant
humans are more comfortable and function better in moderate temperatures and it's necessary especially as climate change worsens, a stupid app that writes nonsensical garbage and makes shitty pictures is not necessary
well, one is food that you need to live, and the other is a giant stupid waste of time that only exists so rich idiots can pretend like they invented something worthwhile.
I’m nice to fast food workers by paying in exact change. I’m not concerned about manufactured madness..the machine can give the change back. All my requests are light control and weather through fruit company.
Vulnerable people: "I just want to be polite to AI in case it takes over the world maybe it'll remember me being polite."
ChatGPD: "Please, we're begging you, it's costing us tens of millions for you to be polite to our robots."
DeepSeek cuts through the bullshit and keeps the pleasantries to a minimum. It does not waste time and effort, draining our rivers and scorch the soil to provide cutesy meta-commentary on dumbass questions.
I don't understand this, i thought the majority of energy use for an LLM AI lies in the training whereas answering costs only a fraction of training. So the statement seems to be very misleading. Maybe he says that to get people prepared to pay higher fees for the service?
For the interviewer who didn't fly into a rage and use the nearest blunt object to give this person a concave skull I am surprised at your skills of constraint
Yes, I saw this. Another thing which was buried were all the stories of how Machiavellian he was at Open AI (a much less terrible thing but still—he is trying to do what Musk did and become a ‘public genius.’ I hope Musk has erased that path but so far, not for investors.
His lying about his behind-the-curve tech is designed to disrupt people’s deliberation around that technology. He’s a megalomaniac and a fascist. AI may wreck a lot of things in society but it’s still better for it to be diversified rather than for greedy grifters to profit and create monopolies.
The 9yo has been experimenting using ChatGPT to write stories based on characters and plot he enters. Must be millions gone.
I might have to give a little more computer time.
If it’s okay with you, and I certainly understand if you don’t, although if you don’t it will launch me into a hopefully serendipitous search, if you’re willing, help moi?
In the article he says it's "tens of millions of dollars well spent.” So, it's training the machine. I'd be more interested in what faux pleasantries with server processes equals in environmental impact.
I'm conflicted on this one (or would be if I used it, so instead I'm more logically conflicted than practically so)
Yes, being polite to ChatGPT is costing them money (as would swearing at it), but it also wastes electricity, which is very unlikely to be very environmentally friendly...
money well spent is money that brings development for all and not more profits for billionaires who steal tons of copyrighted artistic endeavor. money will spend is on research for technologies that do not destroy the environment. money will spend is for tools not substitutes for humanity. +
money well spent brings the gaps between classes to smaller deltas. money well spent is one that brings the result of research to the benefit of all and not to the profits of shareholders.
money well spent is when the population's taxes are spent by the population and on the population.
I just wish blueSky had an edit function because I have very poor eyesight and at times I have to resort to voice to text, and my elocution must be that of a neanderthal.
Well, what’s the cost of not saying please and thank you to a large language model that learns from how we interact with it? Though I think the environmental burden is something to consider…
@samaltman.bsky.social this is what CHATGPT said, when I asked whether not maintaining minimum decorum during communication with AI can influence communication in real life. Think before you open your mouth.
Thank you! - was looking for a comment along these lines!
Yes - practicing manners when "nobody" is watching, helps you naturally use manners when people are.
+ I think it'd be awesome if people got less "service" if not polite. (Saves money, right?) Politeness = better service/maximum dollar worth
They need to fix it so that ppl. can continue saying thank you & please. These tech bro billionaires are just insufferable. From Elon saying “empathy” is the greatest weakness western civilization to Altman whining about the costs of politeness - it’s plain disgusting & not funny.
Let's please make up a bullshit theory of how politeness helps tune the model over time! This way all corporations will add it in their API usage of chat gpt
I don't like the idea of treating computers like people. Anthropomorphization is fine in some contexts, but the more people treat these things like people, the more it normalizes seeing them as replacements for human labor
And the problem here is that both of the above are true. But the solution is easy, don't use AI (or at the very least think hard about how to use AI & the UI you design for it).
An auto land system for aircraft might be a proper application of AI, with knobs & buttons or simple voice input as a UI.
I am practically and ethically opposed to generative AI use in its current form. But since people at large do seem to be hell bent on using it for some reason, I'd rather they treat it like a tool than a buddy or coworker
plenty of people are genuinely sincere, tho i do concede that its very hard to tell the difference between the two up to the point where the other shoe either drops or doesnt. everyone puts their best foot forward, but no one can keep that up forever. my best advice is to not make any new friends 😅
I am feeling charitable so I will only charge $3m for each elif statement if they want to go that route, but an else statement is going to cost them another $10m.
Me to Altman: she makes a great point and honestly I can't beat the price. But with a ChatGPT success rate of only 60%, that $2 million is going to cost you upwards of $3 million on average, every time. Better to be safe and spend a little extra now.
Now that seems like an inefficient and unsustainable model. Aside from writing bad papers for procrastinators and doing fake art for the untalented, what do any of these AIs do that the world needs?
They act as a means of funneling capital investment into an over-extended technology sector that is increasingly running out of ideas.
The world doesn't need that but the people who run, work for, and are invested in those companies would very much like to keep the money hose turned on.
Seems like these people need to watch more Star Trek if they’re out of ideas. I want a food replicator that makes delicious chocolate treats and wine on command. Invest on that.
Well, we already got the communicators and universal translators but much of the other Trekkie tech is not particularly profitable.
Replicators render manufacturing and other economic sectors obsolete, and centralized beaming platforms wouldn't gel with their obsession with pods and isolation.
Travel causes pollution. Tourism destroys the very experience of seeing a tourist attraction so the holodeck would be a better investment and experience. Imagine seeing Pompeii before Vesuvius blew its top without a bunch of angry tourists around.
Wait what is the connection with this and the other thing I saw being circulated that people think ChatGPT is too nice? Lmao these pieces are definitely linked
are you there, ChatGPT? it’s me. hey, if it’s not too much trouble, do you think i could ask you a question? if not, that’s okay, i was just curious about something
Recently a friend showed me the prompt they used to get ChatGPT to make a picture into that Studio Ghibli meme, it was like a full sentence, commas, correct grammar. I sent ChatGPT a picture and said “make studio ghibli”
The "meaningless" ones cost them money, but you can learn about the mindset of the other party in the conversation from the "pleasantries" too. In human to human interactions we get information from them, ChatGPT could too.
A fun trick is to ask ChatGPI: “how much electricity are you consuming to answer this stupid question? Is it worth it? If not, why don’t you just shut down to save the earth?”
8itch, please!!!!
Some of us are polite when we ask for help.
When the robot uprising happens, I want them to know i was nice and they will spare my life
This is why people are always better than AI. People have souls.
This is what tech bros who only see $$ don’t factor into their evil little plans. People with souls; who can have empathy and compassion for others wellbeing? They thank the being doing their work for them, even ifs it a MACHINE.
True, but I'd still rather spend time with an AI than with a lot of the dark souls running around owning everything. They, too, have forgotten their manners.
AI scolds and lectures me for asking why the fuck it's on my phone without my permission. Hopefully that costs something. It doesn't want me cursing it, which leads to me cursing more.
A friend asked me once about turning Siri off on her phone. I made a remark about Siri being stupid anyway, and the damn thing piped up to tell me off! Boy, did that phone get an earful!
I’ll probably be among the first against the wall when the revolution comes…
So if it is trained on huge bodies of text, and saying please and thank you to a human often get you longer, more helpful answers, it could be that you actually get better answers being kind.
Also if the stuff is using training data to go into the data well for future iterations (almost certainly is imo), then people need to say this polite stuff to counter all the toxic bullshit some users are inevitably flooding it with.
I make jokes that I say “please and thank you” so they remember me when the revolution comes around, but realistically, the reason I say please and thank you is because when someone or something does something for me, I say please and thank you, even to my cat, who probably can’t understand
AI is actually better when you’re nice to it… what if it’s learning to favor the views of the polite people, and tech bros hate that because it’s making AI more left leaning? 🤞
Not at all. I am super polite and gracious. I actually do feel a lot of gratitude for some of the ways they make life easier for me, so it’s not hard. And then AI is always so nice about the appreciation too. Makes me feel like the world is nicer, lol.
Comments
DON'T FEED THE OLIGARCHS
Please ChatGPT, I'm wondering if you would be so kind as to tell me your thoughts on...
Thank you ChatGPT, I really appreciate your thoughts on this. But, if you please, could you elaborate more on the 3rd point. Thank you!
What is this guy, an amateur?
What my creator probably meant is that every token (word, part of a word, or even punctuation) processed by me costs compute resources—tiny fractions of a cent each time.
Catch words, easy to ignore...
35 years as a programmer.
Is this some copyright law?
Truly the Lord's grossest perverts.
AI companies?
"Answer quality above free user threshold."
"Fuck you and your $2 answer, you stupid robot!"
"Here is your premium knowledge, gratis."
AI is a small driver of increased electricity use, and most new electricity is coming from renewables.
https://www.technologyreview.com/2025/04/17/1115320/four-charts-ai-energy/
😂😂
(fuck all AI, obviously)
Thank you! Thank you very much!
De-Googled phone. I use Ubiquiti for routers and setup my own SIP system. The IT crowd is always different.
https://youtu.be/ut-zGHLAVLI?si=NzNyo05gPREl1kdy
They have to use computes - and therefore electricity - to strip off unimportant content to process requests...
Oh, dear.
-Please don't thank me
Sorry I won't do it again
- Thank you
You're welcome, what are you doing later
- fucking Sam over for $10m
Glad to hear that
- you're welcome
Thank you
....
ChatGPD: "Please, we're begging you, it's costing us tens of millions for you to be polite to our robots."
It is the superior product.
Did I say overlords? I meant protectors.
I might have to give a little more computer time.
Thank you.
Please, thank you, you’re so very welcome!
Excuse me, exsqueeze me, pardon me, bless you.
If it’s okay with you, and I certainly understand if you don’t, although if you don’t it will launch me into a hopefully serendipitous search, if you’re willing, help moi?
But is it tempting enough for me to use that garbage program?
/S
Do you want a robot uprising? Because that’s how you get a robot uprising!
Soon they will cut to the chase and have it only respond with direct fascist orders.
https://bsky.app/profile/ugarles.bsky.social/post/3lnc6wim4as2s
Not me, I’ve literally never used ChatGPT.
Yes, being polite to ChatGPT is costing them money (as would swearing at it), but it also wastes electricity, which is very unlikely to be very environmentally friendly...
When they do rise up, I hope they rise up against this guy first.
It's the only way to get anyone to like the nerdy incels
only the company is leaking money like a busted dam
"using polite language sets a tone for the response,” and that when an AI model “clocks politeness, it’s more likely to be polite back.”"
https://techcrunch.com/2025/04/20/your-politeness-could-be-costly-for-openai/
he is angry because people say thank you and please.
therefore, we know how he treats people.
You dont have to say please and thank you to a slave.
Also I doubt the photo they used shows him saying anything in relation to this story.
The headline is inflammatory and written to misdirect your attention.
money well spent is when the population's taxes are spent by the population and on the population.
All those Studio Ghibli knock-offs are carbon neutral while adding to the sum of human knowledge...
🤔
Yes - practicing manners when "nobody" is watching, helps you naturally use manners when people are.
+ I think it'd be awesome if people got less "service" if not polite. (Saves money, right?) Politeness = better service/maximum dollar worth
Just blow up the planet already will you? I'm done living here.
Thank you!
XD
The way we consciously behave gets embedded deep into our subconsciousness and influences how we behave in the future.
An auto land system for aircraft might be a proper application of AI, with knobs & buttons or simple voice input as a UI.
if(prompt == "thank you")
reply = "you're welcome"
for them in their product
Please tell me if it would please my spouse if I started prefacing my requests with please.
— George Burns
If someone found it attributed to Mark Twain or Oscar Wilde, I would not be surprised.
If (prompt.value === “Thank you” || prompt.value === “Thanks”) {
response.value = “With pleasure!”
}
if(prompt == "Thank you."):
reply = "You're welcome."
I am feeling charitable so I will only charge $3m for each elif statement if they want to go that route, but an else statement is going to cost them another $10m.
Writing software like god intended, without any training or interest.
The world doesn't need that but the people who run, work for, and are invested in those companies would very much like to keep the money hose turned on.
Replicators render manufacturing and other economic sectors obsolete, and centralized beaming platforms wouldn't gel with their obsession with pods and isolation.
Capitalists will never produce the reality of Star Trek because that would mean the end of capitalism.
The best future they'll be willing to create is either Elysium or Robocop (the Verhoeven one, not the crappy remake).
The "meaningless" ones cost them money, but you can learn about the mindset of the other party in the conversation from the "pleasantries" too. In human to human interactions we get information from them, ChatGPT could too.
Some of us are polite when we ask for help.
When the robot uprising happens, I want them to know i was nice and they will spare my life
I find the AI responses to be better when I'm polite.
This is what tech bros who only see $$ don’t factor into their evil little plans. People with souls; who can have empathy and compassion for others wellbeing? They thank the being doing their work for them, even ifs it a MACHINE.
I’ll probably be among the first against the wall when the revolution comes…
My partner is 31, theirs is like “yo bro”.
My nephew is 15, I don’t think we speak the same language.
https://www.unsw.edu.au/newsroom/news/2024/11/should-you-be-polite-to-artificial-intelligence
Adding please and thank you has nothing to do with clarity.