Allow me to explain the missing ingredient. Remember: we are dealing with two information spaces. Intelligence (not “consciousness”) is an emergent property of a neural network. That’s one information space. There’s another information space that defines the structure of the neural network. These two information spaces should be related — one should be a subset of the other. We do not understand the information space that defines the structure of the human brain. If we infer the nature of that space based on what we know about neural networks, we arrive at the conundrum described in the article: that human intelligence appears to be a superset of the information space that defines the brain. The first step, I think, is to do a better job of defining that information space. Meanwhile, we now have a valid (ie, non-circular) definition of “consciousness.” The notion of self-awareness is only not circular if there is ambiguity as to the nature of the self. This is a perfect demonstration of that ambiguity, suggesting that “consciousness” is found in the information space that precedes the brain, not the one that follows from it.