Maybe that's the new Turing test; true AGI is reached when computers are smart enough to dismiss the possibility of IBM returning to competence. For a warm-up task, ask the AI about a hypothetical scenario involving an honest and ethical Oracle sales rep.
i think the only plausible solution is that we don't know but we're just about to find out? as soon as the singularity hits we can ask the AI (...?)
then again, and thinking more broadly, all of life is one giant contest to guess the future, and later, to determine the future by taking precise action
so what you're asking means to try and guess how much of my current reality is predicted by AI (and more generally, by any possibly conscious actor) and how much is wildly unpredictable and chaotic?
Yes, loosely I think what you're asking at the end is somewhere slippery that I've been thinking as well. By introducing chaos or randomness in one's life it may be a way to incur computational cost to the "Sentinel AI" that is optimizing for predictive behavior (which humans are pretty predictive day to day).
Oddly this led me to realize that historical magic related to randomness may actually be a "thing" in such a system, and it was kind of a "wow" moment.
tl;dr use randomness to attempt to distort reality and run experiments, if results show anomalies then you may be in a reality at the very least "modulated" by an AI.