Old Content:

Using AI to Determine Writers’ Intent

September 29, 2007 by Kyle Brady
Tags: AI, Code, The Internet
As required by the FTC, a Full Disclosure is available - this piece adheres to the Code of Ethics

---

An article over at Wired talks about how the government is supposedly developing (or perfecting) a form of AI that crawls the internet in order to find articles that "sound" like a terrorist.

Interesting. I wonder: if you write an anti-Bush post on your blog, but in a non-combative fashion, do you appear on the list of "terrorist" authors? How does it determine, exactly, the voice, tone, and intent of the author? Tone is sometimes hard to understand in written form, by humans, so how is an AI program going to do any better?

On the other side of the coin, could this be a step towards "true" AI? Meaning, artificial intelligence that actually learns, adapts, and acts based on a dynamic set of self-created methods and instructions... like, for example, the central computer in the movie "iRobot" with Will Smith.

Government funded projects have a habit of trickling down to the general public, after being declassified, and becoming large fixtures in modern technology and every day life. Just look at the microwave, satellite radio, GPS, regular radio, the Internet, etc. I can imagine, in the near future, a library you can use in your own code (much like the standard C libraries) that allows your program to learn, adapt, and rewrite itself as necessary.

And who says the American government is evil?

Old Content posts are leftovers from a less structured, less civilzed era that are kept for posterity.
Kyle can be found on Twitter and MySpace, or reached via email.