Last week I talked about a point that is often misunderstood: for AIs, truth does not exist.
Today, I take the reasoning a step further. Because there is an even more deeply rooted misconception: believing that an LLM is a knowledge base. It is not. A language model generates probable sequences of words, not verified facts. In other words, it recites fluently, but it never cites.
That is exactly what I develop in my new podcast: why this confusion persists, and how to tell parametric memory from explicit memory, so that we can finally combine the two properly.
Find the full transcript of this episode, along with all the sources and plenty more, at: An LLM is not a knowledge base
