Thursday, November 19, 2009

Artificial Intelligence a la Blogspam

Drawing back the curtain on this thing we used to call weblogging and have since shortened a tad...

I have recently been seeing something I've heard about but hadn't personally encountered before -- overnight, this blog receives strange comments which are sometimes related to the posted content, sometimes not.  There seems to be some artificial intelligence involved, augmented with human participation, although sometimes even that doesn't rise above the level of what AI could manage on its own.

Usually these comments contain an advertising link, and they almost never contain any interesting or relevant new content.  There's something inhuman about them in most cases, and as they don't generally add anything to the discussion, I tend to reject them.

But I have seen a few interesting patterns in these attempts to look and sound like a human being genuinely trying to add to the online conversation.  My guess is that these comments are computer-generated with some human assistance, but that even the human being isn't really engaged in reading or comprehending the post.  It appears to be a factory operation designed to push persistent ad links onto the Web as inexpensively as possible, but on occasion one of these submissions reads coherently enough to challenge Turing Test criteria.  Usually, though, there's something "off" about them -- the more they say, the less sane they sound to a real human being.

These are the types of blog spam posts I have catalogued to date:

-- "I read your post and I agree!" and "This is very interesting!" comments, which are either completely generic or work "game" into the content somehow

-- Summary of the post itself, with bonus flattery directed at either the author or some product I have mentioned

-- Seemingly coherent responses that give away their generic nature by, say, waxing rhapsodic about a "well built device" or "very fun game" I have been busily tearing apart in the body of the post

-- Comments that quote directly from OTHER posts on my own blog, repeating material clearly unrelated to the topic at hand, presumably hoping to fool me by sounding like me

-- Comments that assemble random Google information related to something in my post, creating statements about, say, Plexiglas headquarters because someone's name I've mentioned has initials similar to those of the company's CEO, even though the last name differs

I'm sure there's a whole shadow advertising industry out there devoted to creating these kinds of comments, which means bloggers must be equally devoted to keeping the flood at bay.  I have actually let a few of them through, in cases where whatever generated the comment actually ended up saying something interesting, or demonstrated some glimmer of comprehension.

At any rate, it's interesting to observe the process in action, and I regard these attempts with a certain degree of affection.  I like to think that these are Eliza's grandchildren, cast out of the theoretical realm, trying to eke out a living in the Internet age.

No comments:

Post a Comment