Let me say something clearly, because the Firewalkers gets misread sometimes: we are not anti-AI.
AI is helping diagnose cancer earlier than any human clinician can. It is modelling climate systems with a precision that is changing how we understand what we are facing. It is giving people with disabilities new ways to communicate, create, and participate in the world. It is translating languages, accelerating drug discovery, and — yes — helping communities like ours fight back against the very infrastructure that is being built in its name.
We need AI. The question has never been whether to use it. The question is whether we use it wisely — and right now, the answer is often no.
The cost we are not counting
Every time you use a large AI model, somewhere a data centre draws power. Somewhere water is consumed for cooling. Somewhere a planning decision was made — probably without community consultation — to put that infrastructure on land that used to be something else.
The numbers are stark.
These are not arguments against AI. They are arguments for using it differently — more intentionally, more efficiently, and with a genuine accounting of what it costs.
The false choice we keep being offered
The technology industry presents us with a binary: either embrace AI fully and accept its costs as the price of progress, or reject it and fall behind. This is not a real choice. It is a negotiating position dressed as a philosophical truth.
There is a third path. It involves asking harder questions about which AI we use, how we run it, what we run it for, and who bears the cost when it goes wrong.
Compare what we have with what we could have:
The AI we have built
- Hyperscale data centres on ancient land
- Billions of litres of freshwater for cooling
- Electricity grids pushed to breaking point
- Communities consulted for 12 weeks then forgotten
- Proprietary models that capture and centralise knowledge
- Energy costs externalised to the grid and the earth
The AI we could build
- Smaller, efficient models that run on existing hardware
- Renewable-powered infrastructure with real accountability
- Distributed computing that shares the load
- Community ownership of the tools that serve them
- Open-source models that keep knowledge in the commons
- Energy costs counted, disclosed, and minimised
What green AI actually looks like
Green AI is not a slogan. It is a set of real, practical choices that anyone using AI — from an individual to a global corporation — can make right now.
Run AI locally
Tools like Ollama, LM Studio and AnythingLLM let you run powerful AI models on your own computer — no data centre, no cloud, no energy cost beyond your own laptop. The models are smaller but for most everyday tasks they are more than sufficient.
See our local AI guide →Choose efficient models
Mistral, Microsoft Phi, Meta LLaMA and Google Gemma are a fraction of the size of GPT-4 and perform comparably on most tasks. Using the right-sized model for the job is one of the simplest and most effective things you can do. You do not need a 175-billion-parameter model to draft an email.
See our model comparison →Demand renewable infrastructure
The major AI providers are increasingly committing to renewable energy — but commitments and reality are not the same thing. Ask your provider where their data centres run, what their power usage effectiveness score is, and whether their renewable energy is genuinely additional or accounting fiction.
Use AI intentionally
Not every task needs AI. The Jevons paradox tells us that efficiency gains are often eaten by increased use — we get more efficient and then use more. The answer is intentionality: use AI for the things it genuinely does better, and don't use it for everything just because it's there.
Support open source
Open-source AI models keep knowledge in the commons, prevent monopoly consolidation, and tend to produce smaller, more efficient architectures because the community that builds them cares about running them. Hugging Face, EleutherAI, and the open-source community are building the AI the world actually needs.
Advocate for accountability
The AI industry does not yet face the same scrutiny as other heavy industry for its environmental impact. That is changing — but it needs people willing to ask the questions, demand disclosure, and hold organisations accountable for what their infrastructure costs the world. That is exactly what the Firewalkers is for.
The Jevons paradox — the risk we must name
In 1865, the economist William Stanley Jevons observed that improvements in the efficiency of steam engines led not to less coal being used, but to more. Efficiency made coal economically viable for more applications. Use expanded to fill and exceed the gains.
We face the same risk with AI. Green AI models that use less energy are essential — but if their efficiency simply makes AI economical for ten times as many applications, the total environmental cost rises.
This is why green AI is not only a technical question. It is a governance question. A values question. A question about what we actually want from this technology and what we are willing to pay for it — in money, in land, in water, and in the irreplaceable fabric of the living world.
What the Firewalkers believes
We believe AI can be a profound force for good in the world. We use it ourselves — including to build this platform. The irony is intentional and important. The tools that are reshaping the world can also be turned toward its protection.
But we also believe that the current trajectory — hyperscale, extractive, built at speed without community consent — is not inevitable. It is a choice. And it is a choice we can change.
The Firewalkers is building the community, the tools, and the voice to make that change possible. Not by rejecting technology. By demanding that it earns its place in the world.
Use AI to protect the land — not consume it.
Our Good AI guide has free tools, sustainable model recommendations and AI prompts for community action. No technical knowledge required.
Explore the Good AI guide Join the movement