The latest craze has been using ChatGPT to create action figures:

Prompt:
Make a picture of a 3D action figure toy, named “Miguel G.” Make it look like it’s being displayed in a transparent plastic package, blister packaging model. The figure is as in the photo, [Latino/He/His] style is very [light dark skin, bald middle aged old male with a mustache and goatee with a point, no sideburns]. On the top of the packaging there is a large writing: “[Miguel G.]” in white text then below it “[Superhero]” Dressed in a superhero (blue and gold) costume. Also add some supporting items for the job next to the figure, like [toucan bird sidekick, slingshot with small corn tortilla missiles]. The packaging design is minimalist, cardboard color with blue/gold highlights, cute toy style sold in stores. The style is cartoonish, cute but still neat, also put a mock toy brand logo in the top right corner (adapted from Reddit prompt)

But I was curious. “What impact does creating a ‘free’ action figure with AI have on the environment?“

So, I asked Perplexity as part of my earlier thread on water and CO2 emissions. Here is the result:

---

Water Consumption Impact

Direct Water Usage for Image Generation

When generating AI images of action figures using ChatGPT and image generation models (likely DALL-E or similar):

• Per-Image Water Usage: Based on available data, generating a single AI image consumes approximately 0.04-0.1 liters of water, primarily for cooling the data centers processing these requests.
• For a Typical Action Figure Creation Session: If a user generates 10-20 images to get their desired action figure design, this would consume approximately 0.4-2 liters of water.

Contextual Comparison

• This is roughly equivalent to flushing a low-flow toilet once (1.6 gallons/6 liters)
• Significantly less than the water used to manufacture a physical action figure (estimated 10-20 gallons/38-76 liters including plastic production and manufacturing processes)

CO₂ Emissions

Carbon Footprint of AI Image Generation

• Per Image: Each AI-generated image produces approximately 0.01-0.05 kg of CO₂ emissions
• Complete Action Figure Creation Process: A session generating multiple iterations (10-20 images) would produce approximately 0.1-1 kg of CO₂

Contextual Comparison

• This is roughly equivalent to driving a car for 0.5-5 miles
• Significantly less than manufacturing and shipping a physical action figure (estimated 2-5 kg of CO₂)

Economic Costs

Direct Costs to Users

• Free Tier Usage: If using ChatGPT’s free tier with DALL-E integration, the direct monetary cost to users is $0
• ChatGPT Plus Subscription: If using a paid subscription ($20/month), the marginal cost per action figure would be negligible if the user is already subscribing for other purposes

Hidden Costs

• Infrastructure Costs to OpenAI: Approximately $0.02-0.10 per image in computational resources, electricity, and infrastructure maintenance
• Water Costs: Approximately $0.0001-0.0005 per image based on industrial water rates for data center cooling
• Energy Costs: Approximately $0.005-0.02 per image in electricity costs

Total Environmental Impact Assessment

Creating a digital AI action figure through ChatGPT and associated image generation models has a significantly lower environmental impact than manufacturing a physical action figure:

Resource	AI-Generated Action Figure (10-20 images)	Physical Action Figure Manufacturing
Water Usage	0.4-2 liters	38-76 liters (10-20 gallons)
CO₂ Emissions	0.1-1 kg	2-5 kg
Raw Materials	None (digital only)	0.1-0.3 kg plastic and packaging

Conclusion

This is about what you would expect. A single Gen AI image generation? Not so much as a compared to the physical manufacturing process. But when you consider everyone doing it, the cost goes up quite a bit…”the cumulative impact of millions of users generating AI images does contribute to the growing environmental footprint of AI companies like OpenAI, Google, and Microsoft.”