I hadn’t done any calculation, but I guess hundreds of watts over a few seconds that datacenters need to generate an image is way less energy and water than what an artist consumes during several hours while he draws the same image. Plus the electricity for lights or computer consumes.
Thanks, you are being really helpful…
I at least tried to understand the problem and explain my reasoning. And yes, I do not know much about the topic, but everyone here is complaining how I am wrong without saying why so.
So, to have an idea, let’s do the calculation.
Generating 1000 images takes on average 2.907kWh (Power Hungry Processing: Watts Driving the Cost of AI Deployment?, A. S. LUCCIONI et. al., 2024), though with very large varience (standard deviation of 3.31). So generating a single image consumes on average 2.91Wh.
I have to make a few assumptions about the artist. First of all, I will ignore the energy their body would consume, since it is pretty safe to assume, they would need the energy anyway.
Let’s assume it would take the artist one hour to produce the same image (based on nothing, just the ease of calculation; feel free to correct me).
If the artist was drawing using a PC monitor, they would consume tens of watt-hours based on the monitor (Internet article: What is PC Monitor Power Consumption? A Complete Guide, Akash, 2026). Computer with all peripherals would consume even more.
If the artist would choose iPad, using official parameters (Apple Inc.), the iPad should last up to 10 hours with its 28.93Wh battery, so the drawing would consume at least 2.893Wh. This is slightly less then AI, but charging the iPad isn’t 100% efficient. Also they would probably use a stylus for drawing, which also uses some electricity, so I would say the total power needed would be comparable (please don’t force me to calculate these efficiencies).
If the artist would draw on a paper, it would get so much complicated and probably lost in all of the assumptions about materials used, their production complexities, etc. But just for a comparison, a efficient LED light consumes from 4W (Internet article, How Much Electricity Does a LED Light Bulb Use?, 2025), so using a bulb for 44 minutes consumes more energy than generating an image.
So overall under my assumptions, generating a image using AI is at least comparable, probably more efficient then hiring an artist to do the same.
I ignored training the AI, because the more it is used, the less effect it has on the generation, and goes to 0 over time. In the same way I ignored the monitor / iPad / light bulb energy footprint during its production and transfer to the artist, since with more paintings this effect goes to 0 too.
Please do not force me to do any more calculation. I think, this was enough.
That’s a lot of fucking words that all just boil down to “I have no artistic or moral integrity and no respect or regard for the value of art or the human endeavor.”
This shows me how hateful and stubborn you are. I never said (and nobody in this comment chain) that gen-AI is better then artists. Of course I agree with how unfairly AI companies treat artists and that they should be supperted.
The only thing I said was that I think gen-AI uses less power and so should be more environmentally friendly. Then you asked for proof, so I gave you a proof. The proof was based on a lot so assumptions, so it could be wrong, but it is still better then saying “I don’t like AI, so it must be worse in every way”.
It seems insane to me that most people in here refuse to admit that there could be one single perspective in which AI is better, there is one single example where it is useful…
If you have to bend over that far backwards to try to get one tiny element of something to not sound abhorrent then maybe question what the fuck exactly it is that you are defending.
Just tell me, what you think I am defending… I said that I support human artists in this fight. But everything has its pros and cons. I am defending the fact that one of the pros of gen-AI is that it is (according to my calculations) greener. It is also faster. That’s all I am saying. I am not saying these pros outweight the cons, I’m just saying, this is advantagious about gen-AI. Maybe the question is why you refuse to admit there are any pros of gen-AI and why you assume everyone who disagrees a little bit with you has to be wrong and completely on the other side of the argument.
There is nothing wrong with admitting there is something good in something I hate. For example I hate cars in cities, mainly the big ones, but I use them often, because in the current state they are the most efficient mean of transfer. I hate Trump, but have to admit, he is sometimes right…
What would you base that supposition on?
I hadn’t done any calculation, but I guess hundreds of watts over a few seconds that datacenters need to generate an image is way less energy and water than what an artist consumes during several hours while he draws the same image. Plus the electricity for lights or computer consumes.
Oh, ok, so then you just have absolutely no fucking idea what the fuck you’re talking about.
That tracks.
Thanks, you are being really helpful… I at least tried to understand the problem and explain my reasoning. And yes, I do not know much about the topic, but everyone here is complaining how I am wrong without saying why so.
So, to have an idea, let’s do the calculation. Generating 1000 images takes on average 2.907kWh (Power Hungry Processing: Watts Driving the Cost of AI Deployment?, A. S. LUCCIONI et. al., 2024), though with very large varience (standard deviation of 3.31). So generating a single image consumes on average 2.91Wh. I have to make a few assumptions about the artist. First of all, I will ignore the energy their body would consume, since it is pretty safe to assume, they would need the energy anyway. Let’s assume it would take the artist one hour to produce the same image (based on nothing, just the ease of calculation; feel free to correct me). If the artist was drawing using a PC monitor, they would consume tens of watt-hours based on the monitor (Internet article: What is PC Monitor Power Consumption? A Complete Guide, Akash, 2026). Computer with all peripherals would consume even more. If the artist would choose iPad, using official parameters (Apple Inc.), the iPad should last up to 10 hours with its 28.93Wh battery, so the drawing would consume at least 2.893Wh. This is slightly less then AI, but charging the iPad isn’t 100% efficient. Also they would probably use a stylus for drawing, which also uses some electricity, so I would say the total power needed would be comparable (please don’t force me to calculate these efficiencies). If the artist would draw on a paper, it would get so much complicated and probably lost in all of the assumptions about materials used, their production complexities, etc. But just for a comparison, a efficient LED light consumes from 4W (Internet article, How Much Electricity Does a LED Light Bulb Use?, 2025), so using a bulb for 44 minutes consumes more energy than generating an image.
So overall under my assumptions, generating a image using AI is at least comparable, probably more efficient then hiring an artist to do the same.
I ignored training the AI, because the more it is used, the less effect it has on the generation, and goes to 0 over time. In the same way I ignored the monitor / iPad / light bulb energy footprint during its production and transfer to the artist, since with more paintings this effect goes to 0 too.
Please do not force me to do any more calculation. I think, this was enough.
That’s a lot of fucking words that all just boil down to “I have no artistic or moral integrity and no respect or regard for the value of art or the human endeavor.”
Go fuck a robot.
This shows me how hateful and stubborn you are. I never said (and nobody in this comment chain) that gen-AI is better then artists. Of course I agree with how unfairly AI companies treat artists and that they should be supperted. The only thing I said was that I think gen-AI uses less power and so should be more environmentally friendly. Then you asked for proof, so I gave you a proof. The proof was based on a lot so assumptions, so it could be wrong, but it is still better then saying “I don’t like AI, so it must be worse in every way”.
It seems insane to me that most people in here refuse to admit that there could be one single perspective in which AI is better, there is one single example where it is useful…
If you have to bend over that far backwards to try to get one tiny element of something to not sound abhorrent then maybe question what the fuck exactly it is that you are defending.
Again, go fuck a robot.
Just tell me, what you think I am defending… I said that I support human artists in this fight. But everything has its pros and cons. I am defending the fact that one of the pros of gen-AI is that it is (according to my calculations) greener. It is also faster. That’s all I am saying. I am not saying these pros outweight the cons, I’m just saying, this is advantagious about gen-AI. Maybe the question is why you refuse to admit there are any pros of gen-AI and why you assume everyone who disagrees a little bit with you has to be wrong and completely on the other side of the argument.
There is nothing wrong with admitting there is something good in something I hate. For example I hate cars in cities, mainly the big ones, but I use them often, because in the current state they are the most efficient mean of transfer. I hate Trump, but have to admit, he is sometimes right…