I think the AI will have to form contexts about each person/thing based on its interaction with it just like we do. We're not born with extended vocabulary nor meaning of words. That comes later when we attach names to contexts/things. Because of this I believe that our brain is large association network. Say you have a picture frame and things in it. Then those things are associated with other things with other picture frames and who knows how else it's all connected. I don't know whether the logic comes from the frames or if this is something that uses/consumes the frames then making new ones with solutions in them. I also think that our subconciousness is part of this process because when I problem solve, I come up with solution without coming up with steps necessary for solving it. The solution just pops up in my head sometimes and I scrounge or try to trace back things I though about and how they lead me to the solution. I do know that past experience plays into that and perhaps it was stored as frames that were referenced during problem solving process. New frames created after problem was solved. The frames came to me when I was thinking of how early humans used tools. Perhaps they saw a big rock rolling down hill and crushing an animal. That would be stored as a frame in memory. In that frame we have a rock, crushing part, animal, etc. So next time we see an animal in another frame we can back reference to the rock+crushing and figure out that we can use tools to crush a new animal. Though nothing is perfect and if it's a turtle then it will be hard to crush. Thus we make a new frame, remembering to put in body armor and "hard to crush" scenario. So after zillion of these frames it would be easy to ping pong between them to formulate some kind of solution or step forward, though it might be a wrong one. Nothing says we are perfect all the time.