Pavlov’s dog and LLM
Blog: For Practitioners by Practitioners!
Martin Milani posted: “Human intelligence created language to express thoughts—but language itself does not create the thoughts it expresses. Perceptual learning (as seen in neural networks and LLMs) represents a lower-order form of intelligence, rooted in simple pattern recognition, correlation, and basic classification. It’s how both humans and animals learn to detect patterns—like Pavlov’s dog correlating the sound of a bell with the expectation of food, a classic example of conditioning or “training” driven by simple perceptual cues. This is the foundation of how LLMs operate—they identify statistical patterns in data and predict likely outcomes based on past examples. While impressive in scope, this process is not true thinking, and certainly not reasoning. LLMs do not “reason” in the human sense; they simulate and mimic reasoning by retrieving and recombining memorized patterns that resemble logical processes.” Link