308 | Alison Gopnik on Children, AI, and Modes of Thinking

Alison Gopnik sits down with Sean Carroll to argue that childhood is not a deficient version of adulthood but a distinct mode of intelligence. Rather than treating children as incomplete adults, Gopnik contends that evolution has shaped them as explorers—creatures whose brains are optimized for wide, noisy investigation of how the world works—while adults are optimized for exploitation of what they already know. The conversation weaves together developmental psychology, neuroscience, and philosophy to ground this explore/exploit trade-off in both evolutionary logic and the physical structure of the developing brain. The implications ripple outward to artificial intelligence, where current systems, despite their prowess, remain fundamentally limited by what children can do naturally.

The breadth moves across several terrain. Gopnik explains how simulated annealing—a computational strategy that begins with random searching then gradually narrows its focus—mirrors childhood development, from synaptic proliferation and high plasticity to the eventual pruning and myelination of adult brains. She details how children engage in active learning through intervention, not mere pattern-matching: they experiment on the world the way physicists do, and this capacity to discover genuinely new causal models exceeds what large language models accomplish through statistical extraction alone. The discussion turns to how children calibrate imitation vs innovation, adjusting their balance of deference to adult demonstration against their own observations depending on whether an adult appears confident or uncertain. Gopnik also extends the framework beyond humans, noting that elder orcas and other species solve the explore/exploit problem differently, and that intelligence itself is better understood as a constellation of trade-offs across the lifespan than as a single quantity one possesses in greater or lesser amounts.