Ok, I see what your concern is, with the hype around Soylent everyone’s opinion is skewed (even if they’re not among the fanboys).
You decided above that it wasn’t worth your time to try your own self-experiments with it. What if someone else were to take the time to do it? I like the concept but agree with the major troubles you listed above, and I have no experience with designing self-experiments. But maybe I’ll take the time to try and do it properly, long-term, with regular blood tests, noting what I’ve been eating for a couple months before starting, taking data about my fitness levels, etc. Of course, I would need to analyze the risk to myself beforehand.
How about a sentient AI whose utility function is orthogonal to yours? You care nothing about anything it cares about and it cares about nothing you care about. Also, would you call such an AI sentient?