"Distinguish frozen weights from plastic human learning One of the most practically misunderstood differences between LLMs and human cognition is memory persistence. The speaker points out that once training ends, an LLM's weights are frozen: "during that conversation, you're doing Bayesian inference, but then you forget — the next time a new conversation starts with zero context, you don't retain any learning that happened in the previous instance." Humans, by contrast, "remain plastic all our lives." For anyone designing AI-powered workflows, this means you cannot rely on an LLM to accumulate operational knowledge across sessions without explicitly engineering a memory layer — every session effectively starts from scratch. Source: Vishal Misra, No Priors (Martin Casado), 'How LLMs Actually Work: Bayesian Inference, Causality, and the Path to AGI'"