A dream with error correction. When error correction is faulty due to being irrational, or asleep, or on certain drugs, all sorts of strange things appear.
I suspect that if we saw that a cat perceives the world and itself in much the same way as us, and acts on its observations, we would not hesitate long to believe it was conscious.
However, a computer can also be taught to perceive the world and itself, and act on its observations, and we would not call it conscious, because we think that it's still not aware that it's doing all that. But then we could teach it that. Etc. It's hard to formulate what's missing, but we humans feel conscious in a way that's more than just observation and introspection.
I suspect that once we no longer understand computers, we will eventually start to think of them as conscious too. Whether they really are will probably be impossible to know. 
Worth a watch if you have an hour to spare.
I mean, I get that this presentation is based on more empirical grounds and on material which is more fleshed out than the speculations of long dead philosophers. But unless a person is unfamiliar with indirect realism, there isn't anything inherently mind-blowing about the idea of a "controlled hallucination". And this is what this TED talk is about.