Can strong AI emerge by Darwinian principle?

I find evolutionary computation interesting, but I have many questions. There seems to be an assumption that there’s some innate goodness of intelligence in an evolutionary context. This isn’t the case; intelligence just happened to win in our particular environmental niche.

If there’s nothing in the context of natural selection that would “direct” the evolution of strong AI, what would? In the context of artificial selection, the environment must be designed to fill that role. Here, if the environment is designed to select for intelligence, it seems we’ve arrived at a circularity. We still have to understand intelligence sufficiently to bias for it in the environment. So the environmental design doesn’t seem less complicated than the direct route.

I appreciate the argument about faster mutations, but this isn’t a closed search space. Success is inevitable, but only in the infinite, like monkeys typing out the works of Shakespeare. Dawkins wrote of this in The Blind Watchmaker. In his simple example, the selection was based on “the criterion of resemblance to a distant ideal target.” What’s that target for strong AI?