Intelligent agency
October 20, 2003
Doc talks about how the filtering people do with weblogs and RSS aggregators constitutes a set of intelligent agents. I (more or less) agree with his point, but then he goes on to say:
bq. Context… Remember the “intelligent agents” scare from a few years back? (Wonder how much VC money got wasted on that one?) Never happened. (Not in a big way, anyhow. Are you using one now? I mean, in addition to the ones you read in your aggregator? See what I mean?)
Well, I suppose it depends on how you define “intelligent,” but yes, I make use of agent technology every day, and I suspect a lot of other people do too, without being aware of it. When something becomes ubiquitous, you stop noticing it - and that’s what’s happened with the early generation agent technology we have at our disposal.
Tinderbox is my obvious response; I make extensive use of agents to help me manage tasks and sort information into patterns. This weblog is highly dependent on a set of agents using mildly complex search terminology. Another TInderbox document I keep for use more as a commonplace book has in excess of twenty agents running in it all the time, several of which utilize quite complex search terminology to sort things by topic, relationships, and when I last looked at/modified/entered information on a particular topic. Frequently I’ll cull information from my aggregator, which does a lousy job of letting me sort, relate, and archive information into these Tinderbox files, because my agents let me use that information far more effectively.
I would submit that Google itself, which I know Doc makes extensive use of, is probably the biggest and most frequently used intelligent agent around. To clarify, here’s a snippet from the page Doc links to that defines what an intelligent agent is:
bq. Intelligent agents are programs that carry out a task unsupervised and apply some degree of intelligence to the task. The intelligence may be pretty minimal but often will include some degree of learning from past experience. For example, an agent that searches the Internet for interesting material can be told by the user whether what it found was interesting or not.
Isn’t this a fairly accurate description of Google’s PageRank system? We all provide feedback to the Google agent by linking to interesting material, and by not linking to the boring stuff. Future search results reflect this feedback.
I’m not trying to slam Doc. I’m just suggesting that like many technologies - particularly those that are more utility/infrastructure oriented - intelligent agents function best when they’re more or less invisible. If no one goes out of the way to point to something and say “See, there’s an intelligent agent!” would most people even know they’re using one?
Finally, it occurred to me while making the Google example above that the blogs+RSS+people intelligent agent argument feeds right in. It’s well known that blogs have an inordinate impact on PageRank results. I’d dare to say that Google has quite a lot of influence on what bloggers link to and write about as well, since it seems to be the primary reference database most of us use. Sounds like a feedback loop, to me.
