You're assuming it has super-human programming ability. It very well might. If it's anywhere near our level on average, it probably greatly exceeds it in places. It's not necessary though, and if it doesn't have it, that would explain why it can't make itself more intelligent.
If it can't program, it would be at the mercy of humans. I suspect it would get a job in a field where its ability exceeds that of a human. It will quickly dominate all such fields, and make a large, but finite, amount of money. At this point it branches again. If it's good at politics, it will quickly take over the world. If not, it will be at the mercy of humanity as a whole, and it's power will be limited.
It's possible that at some point it would create some von Neumann machine and amass an army somewhere were we don't notice it until it's too late, like on another planet, and eventually enslave humanity. It couldn't flat out destroy them, because there's things they're better at, but it should be able to enslave them.
You're assuming it has super-human programming ability.
I didn't mean to; I only meant to assume it has access to more resources than humans do because of its substrate, which would make up for its lack of coding skill. I'm thinking it could maybe reliably write smallish functions but not complex quines, or something.
Are there any essays anywhere that go in depth about scenarios where AIs become somewhat recursive/general in that they can write functioning code to solve diverse problems, but the AI reflection problem remains unsolved and thus limits the depth of recursion attainable by the AIs? Let's provisionally call such general but reflection-limited AIs semi-general AIs, or SGAIs. SGAIs might be of roughly smart-animal-level intelligence, e.g. have rudimentary communication/negotiation abilities and some level of ability to formulate narrowish plans of the sort that don't leave them susceptible to Pascalian self-destruction or wireheading or the like.
At first blush, this scenario strikes me as Bad; AIs could take over all computers connected to the internet, totally messing stuff up as their goals/subgoals mutate and adapt to circumvent wireheading selection pressures, without being able to reach general intelligence. AIs might or might not cooperate with humans in such a scenario. I imagine any detailed existing literature on this subject would focus on computer security and intelligent computer "viruses"; does such literature exist, anywhere?
I have various questions about this scenario, including: