Please, stop with the AI generated images
Introduction
I have a strong stance on AI-generated content. NO! It’s as artificial and soulless as corporate style. I may have discriminated against a lot of "good content” just because of one image, but as soon as I see a generated image or video, I instantly leave. I cannot stand it personally. Despite that, I don’t hate artificial intelligence and LLMs. I use ChatGPT daily as a tool to help me learn, research, and automate mundane tasks. However I am still mostly AI-averse.
Elaboration
We are in the age of AI slop. In 2022, with the release of ChatGPT and DALL-E 2, the first AI slop content started to show up. The one “genre" I remember fondly was generated images titled “How will the end of the world look like?” showing dilapidated buildings, meteorites falling, or everything set alight like in some hell depiction. In that time, I was excited about the new possibilities. However, as you can see, I grew to absolutely despise it. Now I would much rather see a pixelated, badly cropped image than one more generated image. I hate AI-generated images the most, from all of the content that could be generated. AI-generated text at least provides some informational and rhetorical value, videos from as far as I’ve seen have comedic value, but images? There are millions of stock images, free or paid, floating around the web. You want a drawing of something? Learn how to draw. Sure, it will consume some time, money, and goodwill, but I promise you, it will be a billion times more valuable than something fabricated. Don’t get me started on searching for an image of something via a search engine. It’s filled to the brim with AI images if you don’t turn the filter on, and even if you do, it’s imperfect and still shows some.
Sustainability
Disclaimer: amounts vary by source!
Servers use significant amounts of water to cool them down. GPT-3 training used 5.4 million litres of water. GPT-4 consumed tens of millions of water. Information about the training amount of current model GPT-5 isn’t disclosed, but one thing is sure - the amount of water used will soar persistently with each new, more complex and capable generation. There are one billion messages being sent to chatpgt every single day. And that’s just one AI bot. Gemini, Claude, Deepseek, also need to train their models. Although per-query water use is tiny compared to the training phase - roughly tens of millilitres per short text response.
In the fourth generation, to generate a 100-word email, it consumed about 500ml of water (17 oz). So writing a 400-word email or re-doing a 100-word one will consume as much water as an average human needs per day ≈ 2000ml (½ gallon)
Generating one 1080p image is said to use between 2-3L of water.
Let’s say you want to generate a 10s long 1080p 60fps video. (Assuming it generates frame-by-frame)
60 fps × 10 s = 600 frames
600 frames × 2 L/frame = 1,200 L
1,200 L = water needed for 600 people for a day.
Even if there was some crazy optimisation reducing the use down to 50% it’s still approximated to be 300 people's daily water amount needed for some dumb AI video.
Conclusion
I’m not your parent, and I won’t tell you what to do with AI. Use it however you choose. I just want to share my perspective — to help you understand the impact it can have, if you haven’t thought about it yet. Be human, do human.
Enjoyed the article? Subscribe to my RSS feed :D
Further reading: