Life-cycle assessment of biohybrid neural computing systems: methodological challenges and open questions

Abstract

Biohybrid neural computing, as the use of living neural tissue and adaptive cultures for computation via microelectrode arrays, has increasingly been framed as a potential sustainable alternative to silicon-based artificial intelligence (AI). The rhetorical foundation of the claim contrasts the human brain's approximately 20watt power budget with the megawatt scale of frontier machine-learning (ML) training. However, the claim has traversed peer-reviewed literature, commercial announcements, and public policy discussions while the methodology to evaluate it is still taking shape. The field is still working out what to count as a functional unit, where to draw system boundaries, which silicon comparator to match against, and what a complete biohybrid LCA would even look like. The distinctive feature of biohybrid computing is that it constitutes a single functional object spanning wet-lab and dry-lab subsystems simultaneously, and neither the established biotechnology LCA tradition nor the information-and-communication-technology LCA tradition is equipped to assess such an object on its own terms. This perspective argues that the field may arrive at its ecological claim before developing the methodology required to evaluate it. We articulate the methodological gap, pose six open questions for rigorous assessments, name five principles that should hold even while those questions remain open, and outline the community infrastructure the field may build before any credible substitution claim. We do not propose a finished framework as that work belongs to the community. Here, our contribution is to name what we believe is missing, and to argue that the biohybrid computing credibility depends on treating its ecological claim as a hypothesis to be tested, not a premise to be defended.