In recent years, ‘human-centered artificial intelligence’ (HCAI) has emerged as a dominant framing device in contemporary AI discourse and policy. However, alongside its widespread acceptance, the phrase has received criticism. One type of critique levied against HCAI attacks its tendency towards anthropocentrism, claiming that by taking human well-being as the focus of moral considerations, HCAI is ill-equipped for addressing the harms that AI technologies pose to nonhuman animals and other elements of the natural world. One possible reaction to this critique is to adopt a broader framing device that takes the well-being of the entire biosphere, rather than that of humans alone, as its starting point. This constitutes a move from HCAI towards ‘bio-centered artificial intelligence’ (BCAI). While appealing, the move remains conceptually ambiguous. What does it mean to move towards BCAI? We illustrate this conceptual ambiguity by presenting two possible readings of what it means to move towards BCAI, with significantly different implications for dealing with the development of AI technologies.

The first reading understands moving towards BCAI in terms of substituting the general goal of AI development. AI should not merely serve human goals, meet human needs, and be aligned with human values, but should instead contribute to the flourishing of the biosphere as a whole. This understanding consequently brings nonhuman animals and the natural environment more to the foreground of ethical considerations, likely fostering greater awareness and clarity of the harms posed by AI. However, this reading can be criticized for not being ambitious enough. Critical research into the underlying infrastructures of AI shows that the development of AI technologies is deeply intertwined with ecological harm. Merely changing the goals of AI development is insufficient in light of its deep-seated ecological violence. In response, the second reading of moving towards BCAI calls for a fundamental rethinking of the computational logic underlying the development of AI technologies. By understanding AI as based on a reduction of the world to computational models, we argue that a move towards BCAI involves a rethinking of AI at its deepest most conceptual level.

This research contributes a conceptual understanding of what it might mean to move from ‘human-centered AI’ towards ‘bio-centered AI,’ as such shedding light on what is at stake in the framing of the ethics and politics of AI technologies.