By Rafael Kaufmann and the Digital Gaia Team

Abstract

Digital Gaia’s inaugural whitepaper provides a technical outline of the Natural Intelligence Network (NIN): a collective of rooted (place-based) artificial agents, designed to pursue global ecological regeneration and economic prosperity.

• The NIN is based on Active Inference (ActInf), a framework for constructing Bayes-optimal decision-making agents; our implementation of ActInf leverages a wealth of modern, production-grade Bayesian modeling libraries.
• Each agent in the NIN is embedded in, and co-evolving with, the real-world system it supports, and is able to independently inquire, learn, plan, and allocate resources by continuously performing online active inference on its context.
• Agents interact with humans and with each other via a protocol of “quests”, “claims” and “rewards” that encodes Bayesian inference at multiple levels, providing principled support for trustless inference, dynamic model averaging from a crowdsourced model universe, and complex forms of reasoning and communication.
• NIN’s architecture as a decentralized network of context-bound agents allows it to grow/evolve progressively across spatiotemporal scales and domains, and to discover global goal states adaptively, bypassing most known forms of AI risk.
• All network activity is fully persisted on append-only public databases with field-level cryptographic signatures, which allows agents’ reasoning and decisions to be interrogated and replicated by both humans and other agents.
• In their initial instantiation, NIN agents may act as “assistants”, providing decision support guidance to humans and traditional organizations; as they evolve, we may see them become economic actors of their own right, autonomously managing resources and organizations.
• In this latter, fully agentic incarnation, the NIN may play a crucial role in a global transition towards intentional, positive-sum “Gaianomics”.

Introduction and goals

Digital Gaia was founded in 2022 with the goal of catalyzing the transition to a global regenerative economy. Grounded in a wealth of scientific research, technical innovation, philosophy and indigenous wisdom, we propose a sociotechnical system, the Natural Intelligence Network (NIN), which we believe will play a significant role as the catalyst to this transition.

This whitepaper is a living document, intended to provide an overview of the current state of our technical design and our key theses about how this technology might solve the myriad challenges involved in it. As such, we assume the reader is at least broadly familiar with computer science, in particular with statistical modeling/machine learning methods and with distributed computing.

Structure of this whitepaper

• We first introduce the motivation for the Natural Intelligence Network (NIN). Our primary framing is economic: we describe the agency problem between humans and Nature, aggravated by bounded rationality and incomplete contracts, and the ensuing tragedy of the commons. We discuss the sociotechnical risks that accompany any attempt to solve a challenge of this sort, and consider the design stance that we chose to adopt as a counter.
• The majority of the paper describes the technical architecture of the NIN, focusing mainly on the inner workings of its nodes, called Ents, and particularly on aspects that diverge from standard agent modeling techniques and/or leverage techniques and frameworks that may be unfamiliar to most readers.
• We also provide two appendices: the first walks through our choice of technical frameworks (Bayesian probability and active inference), for readers that may be less familiar with these frameworks. The second briefly motivates our design stance and provides key references.

Motivation: Endowing ecosystems with economic agency

Human civilization depends on ecosystem services provided by our biosphere’s natural infrastructure (Mengist et al 2020). If its operating parameters are respected, and barring exogenous catastrophes like asteroids or tectonic movements, the biosphere is able to support humanity at high levels of population density and material prosperity indefinitely (Tan 2019). Yet economic actors with bounded rationality have pervasive and self-perpetuating incentives to overgraze on the natural environment, locking humanity into a Prisoner’s Dilemma (Hebert 2018). Traditional institutions, whether private (land trusts, Natural Asset Companies, DAOs) or governmental (ministries and supranational bodies) cannot be counted on to deliver us from the lose-lose outcomes implied: no matter how firmly we may state our desire for these institutions to act “on behalf of” nature, and no matter how sincere their leaders’ pledges to do so may seem, they can never be credible decision-making proxies for nature.

This is a straightforward consequence of the agency dilemma in economics (Eisenhardt 1989): humans and nature are locked into an agent-principal relationship where, on the time scales of human cognition, humans have the upper hand in terms of power and information; nature’s side of the feedback loop takes much longer to catch up with us and hold us accountable. In this situation of perfect moral hazard, the game-theoretic equilibrium is for everyone to collude in a race to the bottom of overgrazing (whether they do it consciously or unconsciously is irrelevant). To the extent that some of us behave better than this, it’s down to our capacity for moral intelligence as individuals and societies; but that moral defense line is itself vulnerable, and overgrazers have strong incentives to attack it (by neutralizing leaders, regulatory capture, etc).