Artificial intelligence may not be artificial (news.harvard.edu)

At a Berkman Klein Center event, Google’s Blaise Agüera y Arcas argued that intelligence — human and artificial — is not merely metaphorically computational but literally so. Drawing on his new book What Is Intelligence?, plus ideas from Turing, von Neumann and Lynn Margulis, he traced how brains evolved into massively parallel prediction machines and how evolution’s creative engine includes not just mutation and selection but symbiogenesis — cooperative mergers that increase computational complexity. He framed human cognitive leaps as products of social cooperation and collective computation, not solitary genius, and summarized his stance: “Life was computational from the start.” Agüera y Arcas reinforced the claim with technical examples from Google experiments in which a minimalist root programming language (eight instructions) and millions of random interactions produced self-reproducing, increasingly complex programs, paralleling hypotheses for life’s origins. The talk implies a continuity between biological and engineered systems: prediction, parallelism, self-replication and cooperative modularity are central design patterns. For AI/ML practitioners, that reframing highlights engineering directions (distributed, self-organizing systems; emergent behavior from simple primitives), and raises questions for evaluation, safety and governance as artificial systems converge functionally with biological computation.