AI, Aesthetics, and the Carbon Behind the Curtain

It’s raining in today Florence. That slow, steady kind of rain that makes the city feel like it’s breathing under the weight of its own history. From my home office, a few blocks from where Michelangelo brought David out of flawed marble and into the world, I’m watching people hand their faces to machines.

Not metaphorically. Literally. One after another, while I research for a new program I am writing, I scroll past videos and portraits and selfies that have been fed into AI filters to be turned into cartoons, avatars, Barbie-fied. Action dollified. Over the past month people have been turning themselves into smooth plastic action figure versions of themselves, caught somewhere between nostalgia and pathology. There are millions of these images now. Faces stripped of pores, limbs narrowed into factory perfection. Each one served with that familiar shrug, It’s just fun. I know what I am doing.

That’s always the line, isn’t it? It was just fun when we handed over our faces to Snapchat’s baby filter. Just fun when we let FaceApp guess our age. Just fun when we gave TikTok permission to track every blink and pause to tune the algorithm to our moods.

Just fun is how surveillance got in the door.

And now it’s back again, dressed as Barbie.

But here’s the part no one’s saying loud enough. The part I said to two nineteen-year-olds from Byron Bay who were staying with us this week. They’ve been raised with the right language. Offsets. Regenerative farming. Microplastics. They know how to compost. They’ve been taught that climate responsibility lives in daily choices, in recycled packaging and carbon footprints. But no one had ever spoken to them about server farms. About what it takes to make a single AI-generated image. The water it consumes. The fossil fuels burned. The energy grid behind the doll filters and Lensa portraits. The carbon debt tied to each flicker of novelty.

So I told them.

And when I laid it out, the training cycles, the GPU clusters, the constant data churn, the look on their faces said everything. It wasn’t confusion. It wasn’t guilt. It was grief. That quiet, heavy recognition that they were trying to do the right thing in a system that had left them blind to its real mechanics.

And the betrayal in that moment was real.

Because they were trying. They were told composting and everything else they had learned was enough. That caring was enough. And no one mentioned that every time someone uploads a photo to be Barbie-fied, they were contributing to the same ecological collapse they were taught to fight.

That’s the cruelty of the current moment.

AI has been framed as a marvel, a tool for play, an inevitability. But what it really is what it depends on is energy. Unseen, unchecked, and growing fast. A study from the University of Massachusetts Amherst in 2019 estimated that training a single large AI model can emit as much carbon as five cars in their entire lifespans [https://arxiv.org/abs/1906.02243].

And it’s getting worse.

The newer models are hungrier. The demand is higher. The content, because that’s all this becomes, is being generated at a rate that dwarfs anything we've seen before. And while we feed our likenesses into systems that turn us into dolls, the real world burns. Quite literally.

Water sources are being tapped to cool AI data centres [https://arxiv.org/abs/2304.03271]. Electricity grids are being overloaded to support the always-on, always-generating machine. This is not theoretical. It is happening now. Quietly. Invisibly.

Back in Florence, the rain hasn’t stopped. The streets are slick with the kind of wet that makes the stone shine. The statues carved by hands that belonged to men who never saw their work as disposable are drenched. Michelangelo was twenty-six when he began David. He worked for over two years on a single piece of marble that had been deemed unusable. Botticelli painted Venus for the Medici family in a time when myth and theology bled into each other like pigment into plaster. These artists weren’t perfect. But their work held weight. Real time. Real stakes. Real cost.

What we are doing now has none of that. We are generating images that look like art, feel like art, but are as light as vapour. Not because they’re digital, but because they’re instant. Because they require nothing of us except a face and a prompt. And we are calling that creativity. We are calling it culture.

But culture is built on memory, discipline and tension. Culture cannot be copy-pasted. It cannot be scaled at the speed of trend cycles. When you dress up a machine in the aesthetics of the Renaissance, or Barbie, or 90s cartoons, it doesn’t make it meaningful. It just makes it familiar. And familiar is dangerous, because it numbs us to cost.

And make no mistake, there is a cost.

Every time someone makes an AI version of themselves, they are participating in an industrial-scale operation of unseen environmental damage. Not because they mean to. Not because they are careless. But because we’ve allowed this technology to grow in a vacuum of accountability. We’ve taught a generation to be fluent in tools but illiterate in consequence.

So no, this is not a campaign against fun. But it is a call to stop pretending fun is neutral. Fun has always been the delivery system. That’s how they got us with filters. With quizzes. With harmless little games that mapped our personalities and stored our preferences. That’s how they trained us not to ask questions.

Now AI is training us to believe that creation should be frictionless. That it should feel good, instantly. That effort is a flaw and efficiency is enlightenment.

But Florence stands as a reminder that anything that lasts demands more of you. It demands time. Skill. Doubt. Care. The things that make us human, not perfect.

We owe it to the next generation to tell them this. Not to shame them. To inform them. Because if they care about the planet, and I believe they do, they deserve to know what’s underneath the cute. They deserve to know what it is doing to the only world they have.

Because it’s not just aesthetic fluff. It’s power-hungry infrastructure. Some estimates suggest a single detailed prompt to a large model like Stable Diffusion or Midjourney can use between 5 to 20 times more energy than a Google search [https://arxiv.org/abs/2211.02001]. That might sound small. But scale it up. Multiply that by trend cycles of what ChatGPT knows about you in an action toy vibe, by reposts, by every “just trying it out” user. And remember those servers don’t live in the cloud. They live in real places. Often in drought-hit regions. In Utah. In Arizona. In Western Europe. Running 24/7, cooled by water that could be sustaining crops or ecosystems, powered by grids still dependent on fossil fuels.

So if we’re going to use these tools, and we will, we have to start treating them with the same seriousness we apply to everything else we’ve been taught to ration. Like water. Like fuel. Don’t just hit "generate" like you’re flicking through outfits for a doll. Approach it like turning on every light in your house. Ask yourself if it’s worth it. If you could sketch first. Think first. Plan your prompts instead of hammering the model with randomness. Map what you actually need, not just what might go viral. Respect the cycle. Because the energy to run this tech is coming from somewhere. And one day, it might be coming from your own home. That same home where you turn off the lights when you leave the room. Where you time your showers and carry your tote bag to the shops. Where you fight to live with less.

This is no different. In fact, it's worse when it hides behind entertainment.

We can’t just teach our kids how to use AI. We have to teach them why and when and at what cost. This is digital responsibility, not just digital literacy. Because the planet can’t tell the difference between a viral Barbie filter and a serious research prompt. It just feels the heat.

We have the chance to shape a generation that knows better. That still creates, still experiments, still plays but mindfully not mindlessly. Not endlessly. Not with the lights blazing and the tap running.

Do what the artists of Renaissance Florence did. Know the weight of the work before you begin. Honour the process. Turn off what doesn’t need to be on. And never confuse automation with intention.