Decentralising the Ledger: Why We Are Off-Boarding from Commercial AI
By optimising compact open-source models to process our specific narrative terrain, we achieve two things: complete artistic and data sovereignty, and a drastically reduced environmental footprint.
We have just published a rigorous, numbers-driven lifecycle assessment (LCA) over on our research node (artoffaceless.org). The essay breaks down a mathematical equation that corporate tech platforms rarely want to look at: the true environmental cost of the global AI boom versus the total footprint of an average Western tech household of four.
The findings are stark. When you factor in short hardware obsolescence cycles, ultra-pure silicon processing, and the massive data-network multiplication effects, today’s macro-scale AI infrastructure generates an environmental toll equivalent to over 135 million affluent consumer households.
While personal green choices like keeping smartphones longer are highly commendable, they operate on a completely separate plane from the massive energy draws of centralised hyperscale data centres.
You can read the full, unedited technical paper and review the mathematical matrices here:
Lloyd Lewis
Our Infrastructure Strategy: Localised, Recycled, and Sovereign
This data isn't just an abstract academic exercise for us. It directly dictates the next phase of our studio infrastructure strategy here at artoffaceless.com.
If we are to critique "Cognitive Colonisation®" and push back against corporate algorithmic control, we cannot remain dependent on the centralised, power-hungry servers of commercial LLM providers. For a creative collective, paying monthly subscriptions to multi-billion-dollar tech cartels is both an ethical contradiction and an infrastructure security vulnerability.
Because of this, we are actively developing and testing alternatives to commercial LLM access.
Moving forward, our goal is to migrate our workflows to smaller, highly efficient, fine-tuned models hosted entirely within our own sovereign sustainable architecture. Instead of contributing to the massive carbon footprints of new corporate data complexes, we are leveraging decentralised compute power. We are setting up internal server environments that utilise local PC processing capacity and recycled, refurbished GPUs.
By optimising compact open-source models to process our specific narrative terrain, we achieve two things: complete artistic and data sovereignty, and a drastically reduced environmental footprint. We are trading the corporate cloud for private local nodes.
This infrastructure shift mirrors the broader restructuring we executed earlier this season when we moved our core community spaces off centralised platforms to our independent network hubs. We build the terrain; we run the hardware.
Please subscribe to Art of FACELESS for upcoming technical logs detailing our progress and technical information on our local GPU array setups and localised model workflows.
— Art of FACELESS Studio Team