I only have to be correct for my own use case (with nearly 28 years experience being myself) and google has to be correct for everyone with (at best) 18 years of search results, of which you have to back out what the user is searching for. e.g. If I look for George Washingtons birthday I could look for information about the person george washington, the actual query george washington birthday, where I might find the data wikipedia:george washington, or an event that could contain it preseidents day.
I know both what my query is, and what a successful answer to my query is (at least the format of an acceptable return). Google has to infer that a query like presidents day and then a query like george washington are related and a successful result was the resource: wikipedia, because it contained george washington's birthday.
The point is:
* these examples are contrived. However, a person can be given the tooling to organize their own data & potential data.
* a person will have insight into what they want, almost certainly better than google.
* if they are wrong/the resources hasn't been cached for them, well, the personalized engine can just simply default to google if no results exist and return them
* privacy.
edit: not only this but having every site I visit (if I wanted this and private & personally encrypted) available and searchable is hugely valuable. Google receives takedown requests, resources move, or I forgot how I received them. Caching/downloading/storing video, images and web results and having them remain private and searchable is pretty much reason enough for me to tell you to take my money.,