Creative Writing

Generative AI is a divisive topic depending on who you speak with. One recent critique I heard was that it’s extremely bad for the environment, but no evidence has ever been presented with this statement. That made me slightly sceptical of that specific talking point so I decided to ask the AI to defend itself. This was my first time using ChatGPT’s deep research and I’ve got to say, I was mighty impressed, it spent a good 12 minutes, checking through 23 sources and refining the information with the queries I’d set and ultimately this is what came back: Electricity Use: AI Image Generation vs. Everyday Activities   Introduction Generating an image with an AI model (like ChatGPT’s image generation or similar generative models) consumes electricity – as do many common household tasks. This report compares the electricity usage of creating one AI-generated image (in kilowatt-hours, kWh) to several everyday activities, highlighting the environmental impact (energy use and equivalent carbon emissions) of each. Below, we examine each activity’s typical energy consumption and, where available, provide an estimate of its carbon footprint (CO₂ emissions). A summary comparison table is included for quick reference. (Note: Carbon emissions are approximated using a conversion of about 300–500 grams CO₂ per kWh, depending on energy source. Lower values assume cleaner electricity, higher values a fossil-fuel mix.) AI Image Generation (Single Image) Generative AI models require computational power. Running a model inference to produce one image typically uses only a few thousandths of a kWh. Recent tests on various AI image models found an average of about 0.003 kWh per image generated (​theverge.com). This is on the same order of magnitude as the energy needed to charge a smartphone once. In fact, one AI-generated image can consume almost as much electricity as fully charging an average smartphone (​theverge.com). In CO₂ terms, 0.003 kWh corresponds roughly to 1–2 grams of CO₂. Why so low? Individual AI inferences are brief computations, often on efficient hardware. However, note that the training of such models is extremely energy-intensive (hundreds of MWh ​theverge.com), and widespread use of AI at scale means these small per-image costs could add up globally. Boiling a Kettle of Water Boiling water in an electric kettle is a common high-power task. To boil about 1 liter of water (from room temperature to 100 °C) in a typical 2–3 kW electric kettle uses roughly 0.1–0.15 kWh of electricity​ (theguardian.com). For example, one experiment measured about 0.112 kWh to boil 1 L in an electric jug kettle (​theguardian.com). This is about 30–50 times more energy than generating a single AI image. In terms of CO₂, one kettle-boil might emit on the order of 50–75 g CO₂ (depending on the electricity source). Boiling smaller quantities or only as much as needed can save energy. (For comparison, heating water on a stove or microwave is generally a bit less efficient. A microwave might use ~0.14 kWh to boil a cup, see microwave section.) Charging a Smartphone Charging a modern smartphone fully (from 0 to 100%) consumes a surprisingly small amount of energy: on the order of 0.01–0.02 kWh. The U.S. EPA estimates around 0.012–0.019 kWh per charge for an average smartphone (​theverge.com​epa.gov). That’s only 12–19 Wh – a tiny fraction of a kWh. The carbon footprint per phone charge is correspondingly tiny: roughly 5–10 g CO₂. This means that generating one AI image (≈0.003 kWh) uses roughly one-quarter of the energy needed to charge a phone​ (theverge.com). In other words, one AI image = 1/4 of a phone charge in energy. Both are very small in absolute terms. Playing Video Games (Console or PC, 1 Hour) Video gaming can draw significant power, especially on modern consoles or gaming PCs. The exact usage varies by device and game complexity: Game console (modern): A current-generation console like the PlayStation 5 or Xbox Series can draw around 150–200 watts on average during gameplay, which is approximately 0.15–0.2 kWh per hour (​gamesradar.com). Demanding titles on a PS5 have been observed to push power use to ~200 W (0.2 kWh/hour​ gamesradar.com). Older consoles (PS4, Xbox One) or less intensive games might use a bit less (100–150 W). Gaming PC: A high-performance gaming PC (with a powerful GPU) typically ranges from 200 up to 500 watts under load (​energysage.com.) This equates to 0.2–0.5 kWh per hour of intense gameplay. An average gaming session on a mid-range PC might be around 250–300 W (~0.25–0.3 kWh/hour), whereas a top-tier rig can approach 0.5 kWh in an hour of play. Over an hour, gaming easily uses two orders of magnitude more energy than an AI image generation. For example, 1 hour on a console (~0.2 kWh) could power ≈70 AI image generations. Carbon-wise, an hour of console gaming might emit ~60–100 g CO₂ (if using ~0.2 kWh), while an hour on a high-end PC (0.5 kWh) could emit ~150–250 g CO₂. Using energy-saving settings, limiting 4K graphics, or enabling efficiency modes can reduce a gaming system’s power draw. Rendering a 3D Object (Graphics Software) Rendering 3D graphics or animations in software like Blender is a computationally intensive task, often maxing out CPU/GPU resources for extended periods. A mid-range desktop workstation can draw around 600–800 W when actively rendering (​blog.render.st). This means roughly 0.6–0.8 kWh per hour of rendering time for a moderately complex scene on a typical setup (​blog.render.st). In practical terms, a 10-minute render could consume about 0.1 kWh, and a full hour of continuous rendering might use ~0.7 kWh. That’s equivalent to 233 AI images (if each image is 0.003 kWh) or about 5–7 smartphone charges per hour of rendering. The CO₂ emissions for an hour of rendering could be on the order of 200–400 g CO₂. High-end workstations with multiple GPUs could use even more power, while simpler renders on laptops use less. (By comparison, idling or light 3D modeling without rendering is far less demanding.) Watching Online Video (Streaming 1 Hour) Watching an hour of streaming video (Netflix, YouTube, etc.) not only involves the electricity to

It’s been 9 months since I last took on a personal creative project, and that’s simply because life got too busy. I transitioned from freelance to full time in order to better secure a mortgage and help improve my family’s living standards, which has meant a lot of personal sacrifice and less free time in general. On top of this, our house move dragged on for 11 months and the amount of admin to make sure it didn’t fall through was essentially a part time job. The good news… We got the house, after nearly falling through a dozen times we got the home of our dreams (at least within our budget). But it turns out that our new residence was by no means perfect. For one, it was at the upper end of what we could afford, in our starry eyed nature we overlooked some of the minor quirks that quickly became major issues that needed fixing. So my creative aptitude quickly moved from digital to DIY. I’ve been on the roof, scraping wall paper and polystyrene off the walls, painting, ripping up carpet, laying flooring, lifting heavy things which has lead to injuries… Lots of them… All while taking care of my work responsibilities and family. There is an end goal to getting this home in shape, and that is to get back to creating content! But the list of things needing to be done to get to that point is constantly increasing and I’ve found that I haven’t had the environment or time to push forward. But then ChatGPT’s new image features came out, and I saw a glimmer of hope. I’ve made some comics in the past and with the new ChatGPT update, I wondered if it was possible to create a new IP based on the art style of my past projects. After a couple of hours testing out the tools, I fed ChatGPT a character sheet from my animation Wayward Valley, some backgrounds from Frigsbys Ferries and a dozen face shapes from various digital art projects from my past. I told it my inspirations and my vision for the comic and that I wanted 2 characters ‘Leaf and Core’, a sentient rock and plant navigating the post-apocalypse trying to find meanings. And ChatGPT, took my old artwork as a base and produced something new. Now, it’s by no means a perfect workflow. I decide on a location and theme for the 4 panel comic and use ChatGPT to refine the ideas, there’s some clever prompting to keep the characters mostly consistent (although the panels are often photbashed and drawn over in Photoshop). I do feel like there’s a lot of me in the writing, it’s my attitudes and observations about the world that come through, and although ChatGPT will pitch in with line suggestions, I find that I’m putting more thought into that aspect. My wife says the comics aren’t funny but she laughs anyway because she has the context of who I am making them. In reality, I just wanted an outlet to make stories with the limited time I have to create. I have no incentive to get big numbers with Leaf & Core, or turn a profit, it’s a way I can explore ideas and bring them to life while adapting with AI, it’s probably closer to art therapy right now than anything else, there’s no pressure and I’m creating for fun. A much needed break from DIY after an injury has forced me to stop working on the house. There will be those who see Leaf & Core and poo poo it because AI has had a hand in making it. I think that mentality is a little short sighted and sad, as those I’ve shared the comic with have seemed to enjoy it. Me adding value to the world doesn’t stop other people being creative. It’s quite the opposite with these tools. Anyone can use them. But here’s the kicker, just like having an iPhone doesn’t make you a filmmaker, having access to AI doesn’t make you a creative. If you want that, you have to go create things. Anyone could look at Leaf & Core and do their own version of it for the next year, but why would they? When it’s easier and better for motivation to put your time into something that’s personal to you? We’re entering a world where we can make anything we want. It still takes time and commitment to see a project through. With that said, I will be posting the comics on Instagram But if you are someone who has stumbled upon this blog at the right time. You can see the first 9 comics here