Building knowledge from parts
Google has done a great job of decomposing knowledge and making it accessible but is it possible to build knowledge from parts in an automated fashion.
The first questions two ask are 1) Is it possible to build useful knowledge for components in an automated way - (Rooter a machine generated would tend to suggest otherwise:-)) 2) why would we want to do this.
For me the why is the information overload caused in large part by change. A tool to categorize and filter information intelligently would be a better Google. Navigating down to page 15 of a Google search to find a nugget is not very productive. Though compared to the alternative it is a major step forward. Being able to compare and contrast becomes even more interesting an intelligent Google Fight (thanks to Mark Baker for the pointer) may be an interesting starting point. Rather than distill knowledge provide the ability to contrast and compare POV - something our political process badly needs.
How about parallel news tickers showing pros and cons on a viewpoint to help decide on a course of action. How about an uncertainty measure - we all know the uncomfortable feeling of not having enough information to make a decision - does it help to quantify. We live in an imprecise world is it possible to measure/estimate the degree of ambiguity?
So search is great but it is just the start - we have parts now what are we going to build to create new knowledge and understanding?