Earth as Substrate

A parallel pattern is visible in the work being done with AI for biodiversity monitoring and ecological sensing. Traditional ecological survey is labor-intensive, geographically limited, and slow. The convergence of edge computing, machine learning, and satellite sensing has begun to change what is possible. The 2025 NCEAS-led global assessment of automated biodiversity monitoring catalogued over 250 active digital systems tracking living organisms worldwide—acoustic sensors capturing species vocalizations, vision systems analyzing camera trap data, drone-mounted classifiers identifying vegetation and biodiversity indicators, satellite-based AI detecting deforestation and habitat change at planetary scale.

The technical capability is moving faster than the institutional capacity to use it well. The same NCEAS study noted that taxonomic gaps remain severe—particularly for insects, microorganisms, and many marine species—and that biodiversity-rich regions are systematically underrepresented in monitoring infrastructure compared with wealthy regions. Data exists in volumes that vastly exceed analytical capacity. Decisions that depend on this data are often made too late to matter.

What is most interesting about this field, from the perspective of the argument the book has been making, is the question of whose intelligence the systems serve. Indigenous communities, who represent roughly six percent of the global population and steward approximately eighty percent of remaining terrestrial biodiversity, are increasingly deploying AI-assisted Earth observation on terms set by Indigenous governance structures rather than by external developers. The COP16 biodiversity summit featured extensive programming on integrating traditional ecological knowledge with AI-based monitoring. Projects in Goa and the Indian Himalayas have begun combining satellite data with Indigenous knowledge of forest fire patterns, fuelwood pressure, and invasive species spread. The pattern is one of intelligence serving stewardship rather than extraction—the same data infrastructure that extractive industries use to optimize resource removal being repurposed for the opposite work.

This is fragile and uneven. Many AI-for-conservation projects remain technology-centric, designed by developers rather than by communities, and serve as little more than legitimating cover for extractive operations that continue alongside them. But the existence of genuinely people-centric deployments demonstrates the substrate principle in action: the technology does not choose. The conditions surrounding the technology choose.