We introduce a relatively simple model of cognitive dynamics based on Rate Distortion Control TheoryRate Distortion Control Theory, abducting results from statistical mechanics to study analogs to ‘Fisher Zero’ phase transition phenomena. In some contrast to much recent venture capital-based speculation, we find that machine cognition cut off from (very) rapid, high-bandwidth, real-world ‘sensory’ feedback will spend most—or all—of its time in a sensory-deprived ‘hallucination’ mode, a finding parallel with an increasing number of formal studies. In particular, ideas of ‘machine consciousness’ and ‘artificial general intelligence’ seem singularly challenged by these considerations. Consciousness-as-we-know-it involves rapid acquisition of, and response to, massive levels of incoming sensory information under draconian time constraints. Indeed, this accelerated exchange seems the basic purpose of the underlying evolutionary adaptation. Without the centrality of such large-scale to-and-fro sensory/action interchange, ‘consciousness’ and ‘artificial general intelligence’ become, as Peter Hacker contends, meaningless philosophical oxymorons.