Such system often deal with uncertainty quite well including random noise on their inputs. The output ends up a function of both logic and randomness, but can still be useful.
What I'm saying is that there's no logical point to the concept "should" unless you have some concept of free will: everything that happens must happen, or is entirely random.
This is not correct. Whether or not you have free will, stuff influences and is influenced by other stuff, so these arguments are not meaningless or worthless.
> This choice was either made 13 billion years ago at the Big Bang, or it is an entirely random process.
I had thought of this before, and what I had decided is that both of these are also independent of having free will. For example, if the initial state includes unknown and uncomputable transcendental numbers which can somehow "encode" free will and then the working of physics is deterministic, then it is still possible (although not necessarily mandatory) to have free will, even though it is deterministic.
Since there's no free will, outcomes are determined by luck, and what matters is how lucky we can make people through pit-of-success environments. Rust makes people luckier than C++ does.
I also have much less patience for blame than I do in a world with free will. I believe, for example, that blameless postmortems lead to much better outcomes than trying to pretend people had free will to make mistakes, and therefore blaming them for those mistakes.
You can get to these positions through means other than rejection of free will, but the most robust grounds for them are fundamentally deterministic.