Good article! Mostly agree, except for cancer (I think biology research is in some way limited by capacity of keeping things in your head) and Trump win.
Especially on Trump, I feel like our society has no good models for population level behavior and information warfare. This is very hard to research, but I strongly believe there are some low hanging fruits there!
I don't think polling data would be enough, no, but I imagine a 300IQ being would be able to figure out a good way to datamine X and reddit and establish key influencers. For example it should be possible to extract the reach of an influencer by tracking certain speech terms and identifying where they came from (most recently a lot of RW people speak about green card holders as "guests"). This might provide more informative data than polling already, although it is heavily biased towards online people.
But I would not be surprised if a very smart person would be able to extract a signal from publicly accessible social media data.