A piece on Cnet discusses some of the issues behind photo recognition bots, in particular through a discussion with the executives of Riya, a photo startup. There is also a rather introductory discussion of language search bots, and the desire to move beyond “keywordese”- a word Barney Pell uses to describe the funny language we have to learn in order to get keyword search to really find those things for which we are learning. There are also a few comments from Esther Dyson, by way of a startup placed in the story in which Esther has (go figure) invested.
Tom Mitchell, chair of Carnegie Mellon’s machine learning group reckons truly language smart search bots, able to “read” the web, will run rampant by 2015. That’s only about eight years from now. I don’t know what I think about such a prediction, but it would seem a reasonable chance. Stats based systems will probably make the first go at it, but I wonder whether more “heuristic”-like bots will have a better chance. That seems to be the way in which we learn language, first by simple words, then by simple phrases that have a meaning. Once these words, and networks can be placed in context, and connected to each other, well, it would seem you have the capacity for knowledge.