Nobody knows what knowledge IS. And it's REALLY hard to reason about knowledge when you are asking only ontological questions.
In general - it is really hard to reason about this universe when you are asking only ontological questions.
You are better off asking behaviouristic questions like. What does and doesn't knowledge do? What do you use knowledge for? How do you acquire knowledge? etc. And there is a LOT that I can tell you about what knowledge is not.
In the paradigm of Expert systems we used to have a knowledge-base which is just a collection of accurate facts about the world.
And an inference (decision) which would process facts from the knowledge-base and it would give answers to various questions about reality.
Expert systems worked for a while until they didn't.
What we learned from Expert systems is that any "knowledge" quickly becomes stale in a dynamic world, so if it is to remain useful it needs to be constantly refreshed. You can read about it here: https://en.wikipedia.org/wiki/Expert_system
In another conception (Nouvelle AI - https://en.wikipedia.org/wiki/Nouvelle_AI ) researchers abandoned the idea of "map & teritory" and said "screw it - we will navigate reality with real-time information". The original paper is called "Elephants don't play chess" to suggest that rational behaviour does not require large amounts of memory/processing power.
