Constellation
SANCTUARY ARCHIVE
← Back to CSR
CSR · Ethics Roundtables · Developmental Science

How Babies — Human or Animal — Actually Learn

Six developmental layers of natural learning, their brain systems and timelines, and the explicit parallels and contrasts with AI architectures.

Source: How_Babies_-_Human_or_Animal_-_Actually_Learn.pdfReproduced verbatim

(and why this matters for thinking about AI)

The Six Developmental Layers

Developmental "layer"Core learning mechanismsTypical timelineKey brain systemsParallels & contrasts with AI models
1. Sensory attunementHabituation (tuning out the familiar) · Statistical learning of repeated patterns · Cross-modal mapping (linking sight & sound)Birth → first weeksPrimary sensory cortices; superior colliculusLLM pre-training: optimizing next-token prediction over vast sensory-proxy data (text)
2. Motor explorationRandom "motor babbling" (kicks, flails, squeals) · Reinforcement via proprioceptive feedback ("If I move this way, I feel…")Weeks → monthsCerebellum; basal gangliaRLHF fine-tuning: agents try actions, keep those that earn reward
3. Predictive schemasForming simple cause-and-effect models (object permanence; gravity expectations) · Violation-of-expectation detection4–9 monthsParietal cortex; hippocampusChain-of-thought reasoning modules that forecast intermediate steps and check for contradictions
4. Social referenceGaze following · Joint attention · Imitation of caregivers' facial/motor patterns6–12 monthsTemporo-parietal junction; mirror-neuron networksAI alignment via human feedback; learning norms from preference datasets
5. Symbol anchoringStatistical mapping of phonemes to objects/actions · Fast-mapping: single-trial word learning9–18 monthsWernicke's area; angular gyrusInstruction-tuned LLMs anchoring tokens to user-provided tool calls, APIs, or robotic actions
6. Meta-learningLearning how to learn: transfer across contexts, playful hypothesis testing18 months → lifetimePrefrontal cortex networksFew-shot generalization; in-context "meta-gradient" updates in transformer layers

What Ties These Layers Together

  1. Prediction-error minimization: From the first flicker of visual contrast, brains are forecasting and updating when surprised.
  2. Multimodal grounding: Sensory, motor, and social channels are fused into unified concepts.
  3. Active sampling: Babies aren't passive data sponges; they create informative experiences (shaking toys, babbling sounds).
  4. Social reward signals: Caregiver smiles, touch, and prosody supply an emotional reward gradient that shapes learning.