Spammers vs. Turing Test: Inching Along

Most of the dozen or so spam comments I delete every day consist of little more than gibberish. At best, a spam comment will have a poorly worded paragraph or two touting pharmaceuticals, handbags, shoes, or other junk, with absolutely no relation to the post. It’s easy to tell they’re generated by a script: keyword-heavy verbiage, bogus usernames, junk websites, and so forth and so on. Boring, is what they are.

Recently an interesting comment appeared in response to that post on KG-UV3D audio levels which Akismet tagged as spam:

The microphone and radio matching capabilities are terrific. Adjust the wide-range input level for optimum drive to the built-in microphone amplifier [...]

Fluent, idiomatic English that started out pretty nearly on-point for the post! The rest of the comment sounded like advertising copy, though. Well written ad copy, but ad copy nonetheless. Feeding a representative chunk into Google produced a link to the description of the W2IHY Two-band Audio Equalizer on the Official Website.

Now, as it turns out, Julius lives up the river from here and I’ve met him several times. I also know he’s not spamming me, because the URL associated with the post points to some weird-ass Angola gold mining fraud that’s all too familiar from previous spammage. Oh, and the IP address resolves to a Tor server.

As I observed there, eventually the spammers will become bright enough to hold an intelligent conversation and then they’ll be provisionally human. Depending on what they want to talk about …

About these ads
  1. #1 by Aki on 14-September-2012 - 11:46

    “Bogdanov Affair”

    “The Bogdanov’s [sic] papers consist of buzzwords from various fields of mathematical physics, string theory and quantum gravity, strung together into syntactically correct, but semantically meaningless prose.”

    http://en.wikipedia.org/wiki/Bogdanov_Affair

    Let’s do some “computer science”…

    “SCIgen – An Automatic CS Paper Generator”

    http://pdos.csail.mit.edu/scigen/

    • #2 by Ed on 14-September-2012 - 12:13

      strung together into syntactically correct, but semantically meaningless prose

      That covers most of the incoming junk!

      I’ve read stories describing how we might train an AI by having it write junk mail; when the recipients no longer tag its messages as spam, then it’s effectively human. The evidence so far says they’re not even close to the mark…

  2. #3 by Frans on 15-September-2012 - 17:11

    I assume you’re familiar with http://xkcd.com/810/

    • #4 by Ed on 15-September-2012 - 18:00

      Absolutely!

      It’s actually how we raise people: train ‘em up right and they tend to be more useful later on.