In response to the rise of centralized AI systems, this research addresses the ethical, environmental, and epistemological consequences of mass data scraping and knowledge centralization. These systems often obscure data origin and ownership, amplify biases, and consume vast computational resources, reflecting a top-down approach to building so-called "general-purpose" intelligences. The central research question guiding this work is:
How can decentralized intelligence systems be designed to support sustainable, accessible and community-governed growth while ensuring ownership and control of data?
This research explores the potential of local, domain-specific data loops and no-code interfaces to govern and democratize AI development. With the approach of prototyping artifacts to combine deep understanding, experimentation and community engagement, the research aims to design and test a decentralized intelligence framework that enables community governance, explore the impact of decentralized LLM systems on data ownership, transparency and accessibility, and investigate the potential of sustainable, open-source and inclusive AI ecosystems.
Through investigation, this study contributes to a deeper understanding of the potential benefits and challenges of decentralized AI systems and supports the design of more sustainable, inclusive and collaboratively governed Intelligence and data ecosystems.
Keywords
Artificial Intelligence, Distributed Knowledge, Local AI, Open Source Governance, Participatory Infrastructure, Accessibility, Energy Consumption
In recent years, the field of artificial intelligence has undergone a dramatic transformation. What began as an open, assistive tool, accessible through lightweight writing applications which has rapidly evolved into a commercialized, centralized infrastructure. With the rise of large-scale language models (LLMs) developed by tech giants, knowledge is increasingly consolidated within closed systems that rely on mass web scraping, opaque training methods, and intensive energy consumption. This shift raises urgent questions around data ownership, bias, and sustainability.
Early reflections on these developments led to small-scale interventions in building neighborhood-based intelligences, but these proved difficult to sustain. The scale of even a single city exposed the need for more radical decentralization, both technically and socially. This thesis was born from that insight. It explores the development of a framework that enables communities to create, adapt, and maintain their own AI systems. Accessible by design, the platform leverages local computing, open-source technologies, and no-code interfaces to support collaborative knowledge-building.
The research was conducted within the context of universities and research institutions, but it also targets a broader spectrum of users,such as maker labs, small organizations, and individual practitioners. By making the system modular and customizable, it becomes a starting point for various use cases and contexts. The goal is to prototype a democratic model of AI that grows with its users, remains locally accountable, and supports sustainable knowledge systems.