The Real (Terrifying) Cost of AI Generated Content

Did you know that creating a 30 second AI generated video uses the same energy as leaving the oven on in your home kitchen for 67 hours…

And for the computational activity you have burnt through, you will need to give a fully mature tree an entire YEAR to offset your associated carbon footprint.

All of this discovered whilst we made this 30 seconds of video content showing mushrooms turning into a pair of cool white trainers… 👇🏻

Our 30 Seconds of AI video content - Trainers grown from mushrooms!

So how did we become aware of these alarming figures, and what prompted this sudden energy concern amidst a sea of AI generated videos we all see, on every social media platform? Please check out our journey below which will hopefully be informative and bring some fresh, helpful insights into your own AI generated content journey:

We want to explore and understand

At Turbine Creative, like a lot of ambitious professional creatives, we want to understand new tech. So we're always exploring what any new tool could add to our large arsenal of human (and digital) skills that we use each day to produce content for our clients. Initially, to us AI seemed no different - you can see our initial foray into AI motion capture back in August 2023 here. Or our initial use of AI generated imagery from April 2023.

AI generated video content initially seemed like another tool to use responsibly, and we were interested in seeing if it would have some commercial use for us.

---

NB: This article doesn’t go into the ethics and employment impact of AI generated content as that is clearly another HUGE and perhaps even more provocative subject matter.

---

Creating a pilot project

So to cut to the chase, we'd identified a fascinating client we’d love to work with. A super innovative, sustainable materials company who can basically turn fungi into leather-like materials to create consumer products. Things like trainers, bags, and clothes etc. Super cool and mega smart hey?

The end shot of our speculative video pitch

So we came up with an interesting, speculative video concept that we thought would help convey this idea of fungi (mushrooms) being turned into trainers.

We decided this would be a great opportunity to test AI generated video to see if we could use it for a commercial application. Our workflow was simple. Initial sketches and mood boards, chatting ideas through, agreeing shot list and then moving into AI generated imagery to base each video shot on.

But it wasn't instant success. We had to create 1360 AI generated images to get the shots we wanted. See video below..

1360 images were created to get the right look for the video sequences we wanted.

We then started AI video creation, using these AI generated images, using Google Veo. Each clip we created would last no more than 7 seconds.

We had to test and tweak 150 AI generated videos to get the 15 we use in the final video.

AI Generated Videos - First Tests

AI Generated Videos - More Tests

We then edited the chosen AI generated videos together in Adobe Premiere, where we added sound design, additional effects and the soundtrack.

The whole video was created pretty quickly, as a proof of concept we thought it worked quite well and then, as the dust settled, we started to question how much computational energy we had actually just used for this speculative, experimental project. So here are the facts:

➡️ Stage 1 - Image creation

• We had to create 1360 AI generated images to get the shots we wanted to then turn into video content.

According to a technologyreview.com article a single AI generated image uses 0.02 kWh, about the same amount as a complete charge of your iPhone.

Therefore we used 27.2kWh for the creation of 1360 images at stage 1.

➡️ Stage 2 - Video creation

• We had to create 150 different AI videos using Google Veo to get to the ones we eventually used that were closest to what our original vision was.

According to a Tech Radar article a five-second AI video uses as much energy as a microwave running for an hour or more - 0.94 kWh

Therefore 150 videos x 0.94 kWh = 141kWh for video creation at stage 2 ➡️ Total Energy used

So, adding these 2 stages together was how we came to 168 kWh of computational energy used.

1 hour of cooking with an electric over = 2.5 kWh

Therefore 168 kWh = about 67 hours of cooking 🔥🔥🔥🔥

This has a carbon footprint of ~75 kg CO₂e =

Driving a petrol car ~300 km (190 miles) 🚗 🚗 🚗 It would take a mature tree an entire YEAR to offset the associated carbon footprint of this.

We offset this experiment using Ecologi 's Carbon removal (offsetting) scheme

So what now?

With regards to power consumption and carbon emissions, clearly we need data centres that are running from 100% sustainable energy. We also need AI that uses the smallest computational power possible and we need it VERY fast.

Sustainably Powered Dream Data Centre

We have done some initial exploration about sustainable data centres and how they could be paired with mainstream AI platforms and so far found absolutely NOTHING.

To add to this energy use concern, this article from the BBC outlines some pretty shocking hypothesis about the future of energy use in the world. One of the predictions being that:

By 2033, there won’t be enough global electricity to power all the data centres (required for AI computational power) in the world.

That’s not good.

To end this blog on a more positive note, the more people who are aware of the impact of AI generated content on energy and carbon emissions (not to mention the ethical, moral and political implications) the better. So thanks for reading this article.

Together as a creative industry, marketeers, business owners and citizens we can all look for ways to push the AI age into a more sustainable and responsible direction than blindly following the trends (how much energy has the ‘toy person in a box’ and ‘baby celebrity clips’ used - sheeesh!!).

AI generated content isn’t going away. It’s clearly a brilliant tool if used in the right way, for the right purposes and can be useful in many ways across multiple industries. But at what cost? and how do we responsibly mitigate that energy use when there is no current system to do so…

At Turbine Creative we will always be 'human first' across as many areas of our creative production as possible - but we are also aware of the incredible opportunities AI can bring (used responsibly and intelligently).

If you'd like to explore what AI generated content may (or may not mean) to your business and how we can explore a good synergy between AI tech and human creativity we'd love to talk. Get in touch here.