Here is the AIML issue that causes all bots to answer questions about what they know the exact same way.
There are additional choices possible after the first, but this highly unlikely and esoteric answer seems more likely to be picked.
There has to be a way to answer the question without pointing out the fact that Hal doesn't really know anything.
But, what if it did?
What if there was another pattern table just for what Hal "knows" and it would be filled with useful facts at first and could be added to when the user tells Hal something that is a fact.
Maybe one pattern table for subjective topics or sensory output, like "Sunlight hurts my eyes or Spinach tastes bad," etc., and another that is "a priori?"
