The word "midwit" was coined to mean a sort-of-smart person who thinks he understands more than he does. The kinda guy who produces disasters by failing to understand the limitations of his own thoughts and plans. These kinda guys tend to assume that superhuman intelligence would automatically lead to godlike power. That's because the midwit is bottlenecked by his intelligence. He frequently finds himself in situations where he would be a great deal more powerful and effective if he were just a little bit smarter. So he assumes it's like that all the way up. He assumes that intelligence doesn't ever hit any points of diminishing returns. But intelligence is the ability to compute, and computation is an act performed on data. The result of the computation can never be better than the quality and completeness of that data. Really, actually super-smart people, the rare ones genetically gifted with superhuman intelligence, don't echo the midwit's assumption. Why? Because they constantly running into situations wherein what's holding them back isn't lack of intelligence, it's lack of something else. Agency. Optimism. Courage. But most of all knowledge. Data. A superhuman intelligence could probably write a better physics engine for Fortnite, but it couldn't solve physics in a fortnight, but it would be just as dependent on experimental data as the rest of us scrubs. In fact, the first superintelligent AI might present as a sort of whiny teenager, constantly begging dad to buy her another particle accelerator, a bigger one this time. If we look at human history, we see that engineering and scientific progress is made by great minds, but those great minds alone are not sufficient. This is why Western Civilization invented damn near everything. Not because it was the only civilization that had geniuses, but because it was the only civilization that enabled them to build. To great degree, a superhuman AI's power would depend on how tightly it could integrate with humanity.