Richard Morris - Chatbot Metabolism!

Richard Morris - Postdoctoral Fellow, Faculty of Agriculture and Life Sciences, Lincoln University, New Zealand 

INTRO: Most of us have used Chatbots. They are pretty much a staple of online life - whether we like it or not. And we are aware that AI programs like ChatGPT use a lot of resources to run. But does typing “please” and “thanks” really add to energy usage? RTRFM’s Allan Boyd spoke with an expert… 

Conversation Article: Does adding ‘please’ and ‘thank you’ to your ChatGPT prompts really waste energy? -  January 15, 2026 

Transcript

Copy Transcript Copied

SPEAKER_00 0:00

Does being polite use extra energy when chatting with your AI chatbot? AI is embedded into our digital existence, so it's becoming ubiquitous statistics. So more and more of us engage AI chatbots in our work and personal lives every day. ChatGPT has over 800 million weekly users who ask around 2.5 billion prompts or questions each day. That's 30,000 prompts being made per second. And that's just one chatbot service. So without any questions, it does adding please or thank you to your ChatGPT prompts. On the surface, this does seem to make sense as AI systems work harder than non-AI systems. With the bigger prompts, you require more computation, so it would maybe use more energy. But what's a few extra polite words compared to the vast amount of energy required to run data centres? To unpack this and to talk about AI energy metabolism, I'm joined by Dr. Richard Morris, postdoctoral fellow from the Lincoln University in New Zealand. Welcome, Richard. Oh G'day, how are you, Alan? I'm good, yeah. So do we even need to say thanks to our AI? I mean it's a computer programme, right?

SPEAKER_02 1:13

I do habitually. I think you get better results if you're polite. Um I I I had heard the rumors that you know just cutting up please and thank you was was adding billions of watts of energy from somewhere, and um so I became a bit conscious about it, and uh that's that's sort of what prompted me to to look into it and see if there's any truth behind it.

SPEAKER_00 1:32

Yeah, that's your article in the conversation goes deeply into it. I guess we don't need to say thanks, we do, it's kind of human habit, I I guess, because what mum told us to do. So we always say thanks and please, and it's and it's yeah, it's better better to do that, I suppose. But how much energy does chat GPT and other AI infrastructure use? I mean it must it's quite read all the time how heavy it is on usage. How much does do you think it uses?

SPEAKER_02 1:54

Well it's it's it's a massive amount, but the I think the interesting thing to me is how much it's growing. And I'm I'm reading really recent um evidence stuff that's coming out in in the media that you know by 2030, you know, the global usage of um of um AI is going to be doubling and and so so are the the data centres that are that are accompanying that. So, you know, that's the y the demand from those data centres globally is projected to to double by 2030.

SPEAKER_00 2:19

Wow, it would be hard to put a finger on it because it there's so many of these new data centres being built across the planet.

SPEAKER_02 2:25

Yeah, and they they've they're calling them by a different name. Even these these hyper hyper centres and things like that now that that that are going for like you know one gigawatt each um in terms of in terms of power consumption. So these are these are massive and you know they've they've come along quite suddenly and they've come along at a time when you know everything's quite stressed ecologically and and also with the with energy prices coming along and and these things seem to come along and and put extra pressure on and and the speed at which they're um they're coming along too, I think we just haven't really registered how significant they are. I think we still think of computing as something up in the cloud. Ruffines have got a very um what we call a a metabolic impact, you know, when it's when it's chewing up so much land, so much water, and so much so much electricity.

SPEAKER_00 3:11

Yeah, on that on that, the uh this idea of um urban metabolism, um could you s explain a little bit about what that is?

SPEAKER_02 3:17

Urban metabolism is a way of looking at urban systems where it's drawing in energy from from the outskirts of the city and and basically dissipating waste back into the into the outskirts like an like an ecosystem. And so we liken it to a human body, I guess, is is uh is the analogy. Yeah. Um and then when you've got data centers being thrown in, I've got a metabolism of their own, they're just they're consuming so much water, um, they're taking up a hell of a lot of land and impacting the land around it. And and obviously the the amount of energy they're drawn, where where our grids are straining generally at the moment anyway. So this this is a metabolic impact, and I I think people just overlook that uh when they still think about computing. May you know it's been something sort of uh intangible and up in the cloud somewhere.

SPEAKER_00 3:57

Yeah, well you don't see it, do you? You you just you just do your stuff on a computer or your or on your phone, and you don't think of the connection to a data center. We know they're there, but we don't think of that like that every time you make a request into your computer, it it needs energy. And so the more we do it, and the more apps and stuff we have on our phones and our computers, the more it uses up. And there's waste as well.

SPEAKER_02 4:18

That's right, Alan. You had a kind of a really interesting point about the difference between AI and your normal computer, right? Because your normal computer's got all of its memory and data that it needs right there, so it just goes in and grabs it and comes back. Difference with AI, every time you do a prompt, it goes off into the into the ether and carries out a big, big long chain of algorithms and then comes back with an answer. So you're you're igniting a long chain of of connected events, much different to you know just getting something off your local hard drive.

SPEAKER_00 4:46

Yeah, so I suppose um the difference there would be fetching a file and displaying it in a browser, which is lots a lot less energy heavy than an endless round of questions to your chatbot.

SPEAKER_02 4:55

Yeah. Yeah, it's more it's more that exponential use of energy and then it goes out and learns stuff, you know. So it's it's constantly been activated.

unknown 5:03

Yeah.

SPEAKER_02 5:04

And and this is where I think obviously that with where the growth and the amount of memory required and the size of these data centers is required just because it's going out and learning every single time you make an inquiry.

SPEAKER_00 5:14

Yeah, so every time I I use Chat GPT, it does ask me more questions as I ask a question to it, it wants more information. And so to have to refine the prompts and yeah, it's a constant play. Energy must be being churned up.

SPEAKER_02 5:27

It's and Chat BT GPT is quite polite in itself. It'll often say please and thank you to you, but it can't be that damaging if it's doing it itself, you think.

SPEAKER_00 5:34

Yeah, right.

SPEAKER_02 5:35

That's right.

SPEAKER_00 5:36

How dare it be so polite. So yeah, you say that uh it uses more new metabolic load into areas already under strain from climate change, population growth, and competing resource deniance. And I suppose that's along the lines of urban metabolism there. Um this is a entirely new thing that wasn't there a few years ago.

SPEAKER_02 5:53

That's right. And I we've had in the past, you know, I liken it to you know, we've had ports and and things like power stations come along. These are big infrastructures, which as our civilization has developed, these things have come along. And and over time we've we've gotten really good at building them. But these these have come along really quick. And I just wonder if, you know, how we will adapt to it, we will work out how to use them to the best. But um I think it's just something we need to reflect on and step back and say, wait, this is much more different to a normal factory.

SPEAKER_00 6:19

But as you say in your article there, the AI infrastructure is planned and assessed separately, as if it were a digital service rather than an actual thing, like the persistent physical presence. That's right. Yeah.

SPEAKER_02 6:29

We need to start considering it. Like that. It's part, you know, it's not going away. And it's something we just need to build and constructively into our into our the way we put our cities together.

SPEAKER_00 6:37

Yeah, it's not like you're um advocating against uh artificial intelligence or chatbots, yeah, are you? Not at all.

SPEAKER_02 6:43

No. No, it's just it's just something I think we just need we we just need to sort of um realize the bigger picture.

SPEAKER_00 6:48

As as AI systems develop, they are becoming a bit more uh, I don't know, granular. And so they've they they have special tasks with certain things which which reduces their load. So will that those kind of development improvements uh help with efficiency, do you think?

SPEAKER_02 7:01

Could do. I think I think we're at such a stage now where efficiency's secondary. Uh we really have to understand how how this thing impacts us you know as a system. Uh and the efficiency savings I think are going to be um minor compared to the importance of understanding where this whole thing slots in.

SPEAKER_00 7:17

So well what can we do? If if saying please or thank you doesn't really make much difference, what should should we not even ask questions to these things?

SPEAKER_02 7:24

Well, I think that it's it's how we it's it's about asking the right questions, isn't it? You know, um I don't think there's anyone to blame as things come along. The problem with our society is we we do think in silos. Like we'll have one one department dealing with the environment, another one dealing with energy, another one dealing with digital matters. And and these sort of complex problems need all those people to be um pulled in the same room and and working together for solutions. Well, we we're so specialized and and technical in our society these days that we're very we're very siloed and pigeonholed. So it that's why these things kind of creep up on us. We're not really equipped to spot them coming.

SPEAKER_00 7:56

So you're suggesting we need a much more holistic approach to planning.

SPEAKER_02 8:00

Yeah, holistic's a word that gets thrown around, but it's it's purely a systems view of things. It's looking at all the um upstream and downstream impacts of these things. We definitely got the ability to do it, but we just we just tend to just put ourselves into into boxes. My my sort of work I'm involved in is more of you know looking at things from a regional nature-based solutions sort of way and how we're going to deal with climate adaptation, sea level rise and all those problems. And this just slots into another challenge as part of that, but it's not something we can turn away from. We have to look at how we can incorporate uh the energy requirements and also the energy that's dissipated from these things and and try and fit them in the best we can. But as I said, there's there's no point turning our back and and and hoping that these things go away. We've got to um put them in the list of all the other things we've got to deal with.