Ask pg: Why MUST good startup ideas look like bad ideas initially?

You have said many times that "good startup ideas have to look like bad ideas initially."

The reason that you suggest in your essays that the idea MUST look like a bad idea is because if it didn't then other people would do it. What leads me to doubt whether this is true is the idea that it ignores innovation. Therefore, what if the reason that other people dont do it is simply because they don't know how?

For example, if a startup were able to create a piece of software that could take a block of text and produce a summary (not keywords and such, but a real summary that a human would write if asked). I would think that that seems like (and IMO is) a good idea. Here, the problem is how do I do this rather than should I do this.

A historic example rather than a theoretical one could be Googles' basic concept of page-rank. It seems like a good idea (this is not just hindsight, IIRC there are even stories about Larry Paige explaining the concept at the time and people thinking that it was genius). Just, how many people at that time thought of it? Even if people did think of it, it still is a technically difficult algorithm when you take into account the sheer amount of data that it must process. So again, the problem is not should I, but rather, how do I.

Do you see any error in my logic?