Remembering the Future

I have been talking to Bryan about what it would take to integrate intelligent agent technology with weblogging. It seems that weblogs are creating a tremendous volume of metadata that could be mined to produce a number of useful automated services.

However, upon further reflection, I realized that we already have a lot of automated agents mining metadata. It’s called SPAM. So if we don’t want to add to the exponentially growing mass of spam, which now includes weblog comment spam, what do we do? What is the legitimate role of an intelligent agent?

I think this plays into a general theme that has come out of the diagnosis of the AI crash in the 1990s. At least for the present, machines can’t produce an autonomous intelligence, but are an extension and projection of human intelligence. The appropriate role of an intelligent agent then, is to compensate for the limitations of human cognition to allow human intelligence to expand into areas that were not previously possible.

Spam is one such extension, although most would say an inappropriate extension. I was telling Bryan that my weblog got little attention, because I didn’t advertise it by commenting on other people’s weblogs. It was mostly an extended memory, that I used to keep track of things. As he put it, I wasn’t part of a community. So, how could agent technology be legitimately used to extend my ability to join a larger circle of communities?

The obvious answer is that an automated agent shouldn’t be commenting on other people’s web log entries, unless it is invited. That invitation must be extended by a human. Then if someone else finds the results of that agent useful, they can subscribe to the agent generated web log. A human with a stable identity can advertise for an agent, but an agent with an ambiguous identity should not be advertising for a human.