“You asked her what color a house was and she said, ‘It’s white on this side.’”
  “That’s right.”
  “She didn’t assume that the other side was white, too… and a Fair Witness wouldn’t.” 
  -- Stranger in a Strange Land (1961)

An LLM is an abstraction machine, it mashes together anything that is nearby in a high dimensional space. Its statistical model is its source of truth. For a Fair Witness AI reasoning needs to supplant statistics. Which I'm guessing can get weird fast. LLMs are really good at being suggestible. For this we need the opposite.