The point is that non-deterministic outcomes can be a solution to the problem.
We are at the stage that Donald Knuth is releasing literate programs with LLM assisted proof. They aren't stochastic parrots. They are a useful additions to the thinking mans arsenal of tools.