Sense-making in Embodied AI. Towards Autopoietic Chemical AI

Experimental Epistemology (EE) – i.e., the branch of cybernetic epistemology founded by Warren McCulloch [1] to experimentally explore “embodiments of mind” – still has significant potential to be expressed in AI. This is the basic premise of our work, which recognizes in one of EE's most interesting affiliations, namely Humberto Maturana and Francisco Varela’s [2] autopoietic approach to the description of the living organization, a framework useful to improve contemporary Embodied AI's modeling of natural cognition. Our main goal is to incorporate the autopoietic theoretical model of the biological organization in EAI’s synthetic models of natural cognitive systems, in order to artificially generate forms of autonomy and sense-making similar to those of living systems. The core novelty of our research program relies on the hypothesis that, to be effective, this operation cannot be realized in software or hardware, and requires wetware modeling. We plan to use Synthetic Biology (SB)’s techniques to develop wetware models of minimal living-like systems, such as closed chemical reactions based on chemical organization theory and active inference (i.e., the free energy principle), to test whether they can implement sensory-motor loops arising from self- production and agent-environment interactions, and generating minimal forms of autonomous (chemically-)embodied cognition.