Energy Requirements Undermine Substrate Independence and Mind-Body Functionalism (www.cambridge.org)

A new argument challenges the philosophical idea of substrate independence—the claim that minds and computations are implementation‑neutral—by foregrounding the hard constraints of energy. Unlike abstract Turing machines, real brains and computers are energy‑bound mechanisms: energy is limited, costly to acquire, consumed by sensing, processing and acting, demands efficiency (e.g., perceptions per calorie, inferences per joule), and produces qualitative tipping points (e.g., starvation or system shutdown). The author argues that because information processing depends on material energy flows (illustrated by neuroscience points like mitochondria increasing energy per gene), claims that cognitive states can be realized equally across arbitrary substrates, or that minds can be uploaded or simulated without regard to physical energetics, become much less plausible. For the AI/ML community this reframes feasibility and evaluation: energy–information tradeoffs are intrinsic constraints on architecture, deployment, and claims about machine consciousness or moral status. Recent biology linking information to metabolism and computer‑science work on inference‑to‑energy ratios push toward hardware‑software co‑design, energy-aware learning objectives, and new metrics (inferences/joule) rather than purely abstract capability comparisons. Practically, this limits indiscriminate “multiple realization” thought experiments, stresses the importance of energy-efficient algorithms and specialized substrates, and tempers speculative proposals like cost‑free mind uploading or substrate‑agnostic moral attributions.