Forgive the analogy, I know its leaky, but I believe it to be approximate. I liken the mindset that leads to advances to something similar to the Rete algorith

Forgive the analogy, I know its leaky, but I believe it to be approximate. I liken the mindset that leads to advances to something similar to the Rete algorithm, with human knowledge representing the alpha nodes and with human intellect representing the beta nodes, and human memory and computation power encompassing the whole graph. Once you match knowledge with a potential use of that knowledge, you create innovation. This can happen forwards or backwards, driven by new knowledge matching with existing applications (the research lab model) or with new applications matching with existing knowledge (the startup model). The thing is, the rete graph requires enough working memory to store the graph. It really only works if knowledge and application can be co-located. And that means that polymaths have a huge advantage: their graphs are much bigger, so any new information has a much higher probability of matching up with something existing in their heads. This also explains why the greatest scientists and inventors tended to make their most significant discoveries later in life: their brain's graph had grown and become more useful and more connected. The thing is, by using IPC with distributed memory and computation abstractions, you can expand that graph to be represented across multiple humans. That works too. But we must remember that our abilities to innovate across a multi-human graph absolutely rely on our speed and quality of communication, and it must navigate past filters that represent our own philosophical world view. And while that really puts a damper on our collective innovation abilities, it doesn't impede it and we shouldn't be discouraged by it. TL;DR: The world doesn't need polymaths, per se. We can innovate by collective minds communicating with each other. But it's not as good, and so it's still really beneficial to have polymaths around.