It's not that troubling because we should not think that human psychology is inherently optimized (on the individual-level, on a population-/ecological-level is another story). LLM behavior is optimized, so it's not unreasonable that it lies on a Pareto front, which means improving in one area necessarily means underperforming in another.

I feel quite the opposite, I feel like our behavior is definitely optimized based on evolution and societal pressures. How is human psychological evolution not adhering to some set of fitness functions that are some approximation of the best possible solution to a multi-variable optimization space that we live in?

LLMs, on the other hand, are much closer to being pinned to a specific set of objectives