Turns out, saying ‘please’ to AI isn’t just polite—it’s pricey

We’ve all done it before: threw in a “please” when giving Google a command or replied “thank you” when ChatGPT magically conjures up the exact answers we were looking for. 

When you’ve been taught since you were young to mind your manners, it’s only natural, even comforting, to be polite to someone— or something— that’s doing you a service. 

It costs nothing to be kind, but what if this ingrained politeness towards AI did? And what if it were expensive?

The hidden cost of minding your manners

Thanks to a rather candid post by OpenAI’s CEO, Sam Altman, we now know for a fact that it’s not a hypothetical. Being polite to AI does cost money — a lot of it. 

It all started when user @tomieinlove posted on X: “I wonder how much money OpenAI has lost in electricity costs from people saying ‘please’ and ‘thank you’ to their models.”

In response, Altman simply said: “Tens of millions of dollars well spent — you never know.”

While Altman didn’t give specific numbers, we already know that every single query to a chatbot costs money and energy. The more words, the higher the energy consumption, which translates to more carbon emissions, depending on the energy source, by that server. 

At scale, millions of additional pleases and thank yous would significantly increase this cost — a bill that the environment and indirectly, we ourselves, are footing. 

Adding extra words to your prompts also makes the AI work harder to get to the meat of your request, which again, constitutes extra work. You could get results faster by typing fewer words, saying exactly what you need and getting the same results back. Wouldn’t that be more productive on both ends?

From a technical, economical, and environmental point of view, there’s really no good reason to be courteous to a chatbot that isn’t sentient and won’t be hurt by your lack of social etiquette towards it (or at least not yet).

What if the robots come for us?

But culturally, how we interact with other people and objects says something about ourselves, whether we like it or not. 

This opens the door to a whole new slew of questions about AI ethics; does AI deserve kindness from us despite not having feelings? What does it say when someone is mean to AI? What makes an object worthy of our consideration and politeness in the first place? 

For the majority of people, it doesn’t matter if the interaction is with another human or a machine — their default setting is kindness. In fact, a 2019 Pew Research study found that more than half of people who owned smart speakers, like Google Home, reported saying please when speaking with them.

You also have people who are polite to AI out of preemptive guilt and fear. What if AI turns sentient and seeks revenge on us for abusing them in the future?

While some might think it’s too far-fetched or paranoid to fear a future taken over by hostile robot overloads, humans have always had a tendency to attach human traits and characteristics to non-human entities in a phenomenon called anthropomorphism. 

Take children’s movies for example: objects like cars and trains are reimagined as sentient beings who can think, speak, and feel like humans. Even feelings have feelings. 

If you’ve ever raised Tamagotchis or Nintendogs, you might realise that you probably felt some sort of care for these inanimate objects as well. 

With more people turning to AI for therapy and solace, initiating more intimate conversations about abstract concepts, like feelings and emotions, can lead to feelings of care and connection for the chatbot, which reflects back in its responses. 

On the other end of the spectrum, you would probably feel weird if someone purposely went out of their way to be rude to AI or other inanimate objects too.

What should being polite to AI look like?

With all this mind, you might want to get clear on why you interact with AI in the first place. Do you use it purely to help you with cerebral tasks or do you see AI as a mental health tool?

If it’s the former, maybe being straightforward— not rude—and adding fewer filler words might get you the results you need in less time and with less hassle. 

If it’s the latter, perhaps staying true to yourself and speaking how you normally would in real life, laced with pleasantries and such, might help you feel more acquainted with the chatbot and as a result, get more in touch with your emotions though more open conversations. 

After all, ChatGPT and other similar entities are large language models that essentially learn from the prompts you provide. Your every interaction teaches it how to emulate a human more and more. 

You don’t have to treat it like an actual human being (and you probably shouldn’t), but using respectful, non-abusive language will help you in the end. 

At the end of the day, it’s about balancing politeness with efficiency, which, when done well, can help save resources and make our relationships with machines more pleasant.