What would it mean to train AI on an ecosystem? Not on human knowledge and values, but on ecological data streams like rainforest sensor arrays, drone footage, bird calls, water-flow sonics and satellite canopy maps. Would we be building a model whose “corpus” is earth’s living system itself? Guided by experiences of reciprocity (nutrient cycles, pollination networks), competition (resource fights, predator/prey balances) cooperation (flocks, mycorrhizal networks), resilience (a forest reorganizing after a storm), what would alignment look like in this paradigm? How could this AI interact with human culture? Is this even a technical possibility? And what philosophical leaps could be made from there?
Does the AI become a translator between human planning and ecological feedback, like a voice for the rain forest in policy or design? Or does it become a coordinating exo-organism, a nervous system for humanity, mapping “nutrient flows” in civic life and optimizing them in the way forests optimize light and soil nutrients? Would it decide that humanity is uniquely threatening earth’s balance and impose correctives that are contrary to our existence? Or could this AI re-wild us to be earth stewards at a 21st century scale?
This talk will explore these questions, highlighting scientists and computer engineers beginning to consider such alternative frameworks, but also mapping out adjacent thinkers, communities, and threads of thought from fields like philosophy, art, indigenous knowledge and activism. Using her own ecologically-inspired AI application being built locally with engineer Ryan Meyers as a case study, Martin will conclude with a discussion of the kind of work we can do now – in our current capitalist/extractive circumstances – to move ourselves, and our AI, towards conditions that are conducive to life.
