With climate denial on the wane, former denialists have transitioned to portraying sustainability as sacrifice.
The individuals and corporations trying to slow efforts to electrify, to shift to renewables, and to better our relationships to the built environment, transport, and food, are attempting to convince the public that a greener future means giving up comfort, expediency, and pleasure. That’s a powerful argument to the weakly engaged.
For people who don't confront these topics every day, it's enough to hear these ideas a few times to be left with images of range anxiety, bland tofu, and an electric coil where the gas range used to be.
At best, to the casual observer, the conversation around climate change might seem like a grey world of actuarial minutiae like carbon offsets and regulation and taxes. Not inspiring stuff.
For the 2023 AWS re:Invent conference Sustainability Showcase, we wanted to explore how generative AI could help us envision sustainability transformations as something desirable. We wanted to create images that answer the question:
"What if we made different decisions with the same resources?"
Using the foundation model APIs at Amazon Bedrock and the AWS's new Inferentia and Trainium compute instances, we created a series of high-resolution 'start plate' images representing recognizable real-world conditions from the status quo. We then built a layer cake of diffusion inference techniques including fine-tuned models, custom LoRAs, textual inversion, and layered ControlNets to gradually inpaint, in real time, a plausible 'after' scenario in its place.
Where SDXL was the visual medium, our hand-rolled Claude chatbot, running on the Bedrock backend, was our creative partner, churning out hundreds of ideas to explore.
In the end, we created dozens of these scenes, including:
A festive table laden with meat-heavy dishes ➝
A sustainable feast of beautiful seasonal vegetables and grains
A typical American gas station ➝
A biophilic charging oasis with shaded seating, lush landscaping, and a cafe
A typical American 'stroad' ➝
A solar-harvesting protected bike lane adjacent to light rail, making a walkable community
A dense monoculture industrial greenhouse ➝
A vertical aquaponics facility growing a variety of crops
A diesel bus, idling in a cloud of smoke ➝
A sleek electric bus charging wirelessly through its bus stop
A series of flat, barren urban rooftops ➝
A connected series of lush green roofs shaded by solar pergolas
The physical embodiment of the installation was a larger-than-life high-resolution LED display creating these transformations in real-time throughout the run of the re:Invent conference.
There are some fast-moving GIFs on this page showing condensed examples of these transformations in motion, but, as always with our work, the real thing plays out at a more contemplative pace, inviting the viewer to closer inspection.
Technical effort aside, much of the labor in this project was dedicated to understanding the texture of the foundation models we worked with — sustainability has an aesthetic, and that bias in the training material can make some AI-generated portrayals of sustainable outcomes feel unreachably utopian, or, disappointingly, bourgeois and expensive. We didn't want 'only for the rich', or 'green fantasy', or replacement, or 'distant future', we wanted 'today, only with different priorities'.
For more on the conceptual foundations of the project, check out this long-form interview about the making of Epoch Optimizer with TouchDesigner makers Derivative.
Thanks very much to AWS for the faith and support in seeing this project through, and for putting sustainability on the stage at re:Invent.
All images on this page are AI-generated with the exception of the photos of the installation itself.
_______________________________
AWS: EPOCH OPTIMIZER
Client: AWS
AV Support: Fuse
Special thanks:
comfyanonymous,
Dr.Lt.Data,
Ferniclestix,
Elburz Sorkhabi,
Tristan Valencia,
Eric Alba,
Dennis Schoeneberg
Built with:
AWS
TouchDesigner
Stable Diffusion
