A dumb AI making an encycolpedia-like Database by extracting information from (syntaxed) posts, interpreting them as attributes for objects.
The web has too much information, and doing online research you only get the views of a few people.
'The Database' would store how things are described online at a low lelvel.
This would not be comparable to the comprehensability of wikipedia at first.
But using this information intelligently, e.g. automatically composing articles in a subjects, has potential to be extremely simple, versetile and comprenehsive. - It would start a long term AI project.
One cold visualise the different (contradicting) views, how many support them (and hence the statistical significance) together with the credibility.
The information could be extracted from posts following a specific syntax.
A simple solution is making it for open Evernote Notebooks, where researchers could write their findings, and everything would get automatically indexed together with its source.
Mapping up all this knowledge and all the connections would have endless possibilities, and it might be beautiful.
Later development could implement machine understanding of anything from blog posts, scientific articles, to say direct input from large online population doing measurements on the effect of different foods on wellbeing.
Sterlin ps commented
what is the difference between googlenow and siri ?