I'd argue consciousness is a boring byproduct of evolution. You survive by predicting the actions of your competition. You predict your competition by introspecting yourself. Definitionally, consciousness.
If yes, one essentially lands in IIT [0] (according to some, implying that even a light switch has a 1-bit micro-consciousness).
If no, then one wonders what utility qualia brings to the table, compared to simply optimizing the algorithm to work without subjectivity. (Would ML training be more effective if we could somehow add "yuck" and "yum" experiences? How would we tell, assuming we aren't accidentally doing it already?)
For another perspective, Peter Watts' novel "Blindsight" explores the idea that consciousness is merely a temporary local optimum, and eventually selected against, in favor of life forms which are intelligent without being conscious.
[0] https://en.wikipedia.org/wiki/Integrated_information_theory
Maybe Blindsight is right; maybe there's a future life that has no consciousness and thus is always maximally entropic.
Or maybe it's a zero-sum game of simultaneous moves between entropy and some other force, or spatially-separated moves such that relativity cannot order them, making nondeterminism the optimal strategy. But now I'm making stuff up.