jeremy_goodrich - 3:26 pm on Aug 21, 2002 (gmt 0)

It seems like the key to this IBM patent is the fact that it is extending the ability of a search engine through P2P style tech instead of basing an engine entirely on P2P, which is what makes the most sense, because to be commercially viable, in the end there will be a central repository of the data for the engine.

Multi agent systems (like this one) are the next big thing in search imho, because it pushes not only the meanial tasks (eg, fetching documents, storing them in a DB) but the slightly higher order tasks, such as coordinating the whole thing, onto a group of well programmed and fuzzy :) bots.

What I wonder is how powerful, compared to what we have now, would be the engine that actually made good use of such a system, the combination of Google and for example, Napster. It would be pretty incredible to see that, because the participation in Napster and similar services was tremendous, because people got a lot out of it.

If that same participation was engaged in building an incredible SE, it would have to profile people to serve the *most accurate* results, because in most research I've come across, different users will want different things when searching for the same two word phrase.

Mind boggling stuff...something this powerful might actually be able to approach cognition of sentence particples...that would be very interesting / maybe even do syntactic analysis, instead of just basic thematic analysis like some engines do now.