How likely is it that the Singularity is an unsolved problem, even for a superintelligent AI?

I have seen people arguing, that the Technological Singularity could be an unsolved problem for a superintelligent AI, just like there are hard problems for humans in math like the Riemann Hypothesis, the P versus NP problem or the Collatz Conjecture. If the Singularity is an unsolved problem even for a superintelligent AI, that would mean, that the AI couldn't self-improve indefinitely and it would probably take thousands of years until it could improve itself enough to become omniscient.<br><br>What do you think about this idea and how likely do you think it is that the Singularity is an unsolved problem?