Some issues with Probabilities and Bayesianism for Unified Models of Cognition or Artificial General Intelligence