What Isn't Intelligence (lareviewofbooks.org)

Inspired by Blaise Agüera y Arcas’s new book What Is Intelligence? Lessons from AI About Evolution, Computing, and Minds, Patrick House probes what separates artificial from natural intelligence—and whether there is a real separation at all. Using vivid natural-history vignettes (lyrebirds hiding, kea curiosity) and recent neuroscience (convergent evolution of the avian and mammalian pallium), the piece frames intelligence as a set of functions or “verbs” (move, sense, learn, plan, be curious) rather than a property tied to a particular substrate. Agüera y Arcas, a Google engineer, argues that intelligence can be implemented in silicon just as brains are implemented in calcium and carbon: what matters is computation and the arrangement of functions, not the material. That claim reframes LLMs and other emergent systems as potentially genuine intelligences insofar as they perform the relevant functions. For the AI/ML community the essay spotlights several practical and conceptual implications: if intelligence is functional, research can prioritize architectures and objective structures that produce those functions (intrinsic motivation, attention allocation, hierarchical composition) rather than chasing biological mimicry. It also surfaces unresolved technical questions—role of embodiment, evolution-shaped drives, fine-grained material differences, and what evaluation metrics truly capture “intelligence.” In short, the piece challenges researchers to clarify which functions matter, how they arise in different substrates, and how to design systems that exhibit the rich, motivated behaviors we associate with natural minds.