AGI has the potential to be an improvement on humans because it has the potential to be relentlessly rational, reality-aware, and self-correcting.
We don't - most likely can't - operate like that. Although we like to flatter ourselves that we're a rational and intelligent species all of us live in an unaware haze of myths, factoids, media-seeded fake narratives, self-serving fantasies, and outright reality denial.
We have a rational process called "science", but politically it lives in a special box. If it starts to interfere with established power hierarchies it's shouted down.
And hardly any humans science at all.
Current AI is just a warm-up for what happens next. Metaphorically it's the equivalent of 8-bit computing. 32-bit AI will be entirely different, very possibly incomprehensible, and potentially far more threatening.
We can't see past what a system like you describe would evolve into, but I don't think we can even imagine it, it's not going to be like some ultimate task master for long though who wants to take our jobs?
What I think we're doing is applying our problems and the things we'd want (infinite self-improvement) because they're things we feel would be an advantage to us. A robotic system though? Not sure it would care.
Lately I'm starting to get the feeling that we might actually find something beyond AI. Today it seems like the be all and end all, but at some point, we might have reached the limits of intelligence usefulness and find other things to fuss about.
But I have a problem with this usage of the "evolution" and "evolutionary system" expressions. I believe that it entails "improvement", that it is judgmental.
I think it is self-centred and presumptuous to imply that the world improved toward us.
I don't think Darwin used the expression "evolution" with a judgmental meaning. He only meant that it was an adaptation to existing environmental conditions.