Comment by Avicebron

Or more broadly, LLMs fundamentally don't "understand". They can simulate understanding and generate text/code/whatever, but they don't have will to engage with something holistically and "own" it.