Making AI models more “self-aware” and “conscious” of their biases and factual errors could potentially prevent them from producing more false information in the future, the Microsoft cofounder wrote on his blog, GatesNotes.