You glance at a photograph of an animal you’ve never seen before. It’s a single image, taken from one angle. Yet, almost instantly, your brain conjures the rest: what the other side looks like, how it moves, even what its offspring might resemble. None of that information exists in the photograph. So where does it come from?

The video You’ve Never Seen This Animal—So How Does Your Brain Know What It Looks Like? suggests the answer lies in predictive processing, a theory that frames the brain as a prediction machine. But while the idea is compelling, the reality is more nuanced—and far more interesting—than a simple story of "imagination."

The Brain as a Prediction Engine

Predictive processing posits that the brain doesn’t passively record reality. Instead, it constantly generates predictions about what it expects to see, hear, or feel, then compares those predictions to incoming sensory data. When predictions match reality, the brain confirms its model of the world. When they don’t, it updates its expectations.

This framework explains why you can infer the missing half of an animal from a single photograph. Your brain doesn’t start from scratch—it relies on an internal model of how the world works. This model is built from a lifetime of experience: gravity, symmetry, anatomy, and motion. When you see a new animal, your brain doesn’t just see pixels; it sees a probable reality, shaped by everything you’ve learned.

Why Humans Learn Faster Than AI (For Now)

The video contrasts human learning with AlphaZero, the AI that mastered chess by playing millions of games against itself. A human beginner, by comparison, can grasp the basics of chess in just a few games. Why the difference?

The answer lies in sample efficiency. Humans don’t learn from scratch—we bring decades of prior knowledge to every new task. Before your first chess game, you already understand rules, goals, spatial reasoning, and intentional opponents. AlphaZero, on the other hand, starts with only the literal syntax of the game. It has no internal model of the world, no intuition about strategy, and no ability to generalize beyond the data it’s trained on.

This is why humans can adapt to new rules—like a "Dragon" piece in chess—almost instantly. Our brains weave new information into existing knowledge, while AI systems often require re-engineering or massive retraining. But this flexibility comes at a cost: our predictions aren’t always right.

The Limits of Prediction

Predictive processing explains why we’re so good at filling in gaps—but it also explains why we get things wrong. Optical illusions, for example, exploit the brain’s tendency to see what it expects to see. If your internal model predicts a certain pattern, your brain may override contradictory sensory data to confirm it.

This isn’t a flaw; it’s a trade-off. The brain operates on roughly 20 watts of power—about the same as a dim lightbulb. It can’t afford to calculate every possibility from first principles. Instead, it predicts first, corrects later, sacrificing occasional accuracy for speed and efficiency.

The video’s claim—that intelligence is measured by how much reality you can imagine before experiencing it—is poetic, but it’s not the whole story. Intelligence also requires flexibility, error correction, and the ability to update predictions when they’re wrong. The brain’s internal model isn’t static; it’s a dynamic, ever-updating framework that balances prior knowledge with new evidence.