The 30% Spam Bot

I just found the following spam comment on my blog:

My partner and i don’t know about you men and women nevertheless for me personally the structure of your website is very important… I would personally state virtually just as much as this great article alone. In addition I’m a genuine pot intended for Youtube video lessons… or maybe, as a matter of truth, Any advertising content in any respect. Therefore, every time We get a identical write-up My spouse and i solely hope the actual embeds a number of world wide web training video anywhere. Currently, a substantial amount of weblog managers ordinarily tend not to… I cannot picture exactly why?… Perhaps this might be distinct with the topic issue… even so My partner and i continue to believe that it would created for virtually just about any content, because it might regularly be great to discover a heated along with warm and friendly deal with and also perceive the voice at the beginning go to. Regardless, I reckon that I’m a touch off of issue the following?.. Proper… The idea looks like which! Hahah… Appreciate enriching the net using this posting!

For anyone who enjoys language, syntax, or word games, this is pretty hilarious. Bits and pieces of the wording make sense in isolation, but when strung together, the thing is clearly nonsensical. This must be auto-generated. It really makes me wonder about how much effort goes into crafting the algorithms that piece together spam-text.

In a total coincidence, I just finished reading the Cory Doctorow story “Pester Power” (available in free eBook form here as part of his With a Little Help collection) – which is about someone who writes a genetic algorithm to evolve an AI by spamming forums and chat rooms until it can pass a Turing test. The idea is kind of clever, and it might actually work…which is frightening, because I figure spammers, ad agencies, and marketing firms must already be on it…evolving evil AIs whose entire purpose is to sell us things or entrap us with malicious links…

2 thoughts on “The 30% Spam Bot”

  1. It may not be autogenerated:

    perhaps the spammer creaed a perfectly readable spammessage in mongolian. It goes through a translationengine. This is what results. No matter its genesis, it fails.

  2. This is a common example of markov chain generation. First a process crawls over a large corpus of English-language text, and creates a database of the most likely words to follow a given word (or occasionally pair of words, or even letter if you’re interested in coining new terms). It then uses this database to generate a stream of new text.

    What results is statistically likely English that is almost always devoid of useful meaning. It’s a fun party trick, but it gets old quickly. See the x11r5 bot on twitter and for an example of this being used to stimulate conversation, or head over to for text auto-generated from IRC conversations.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.