I trained AI to write blog posts for me, and here's what happened

Imagine a frog hopping between lily pads to get from one side of a pond to the other. Looking at the frog's current position, you could probably eyeball the distance between neighboring lily pads and guess where he'd leap next.

That's basically how a Markov Chain works: given a body of text, it can predict if a word is likely to follow another word. Think of the pond as a body of text (articles, books, etc.) and the lily pads as words. When you send SMS, your phone uses a Markov Chain to "hop" from your current word to what you might say next.

Just for fun I fed all the posts on my website into a Markov Chain, and here's what it said:

  1. The app should show instant personalized recommendations as interesting mementos of cake.
  2. The point isn't to derive meaning from Pokémon.
  3. The social awkwardness out of your employer’s stock is an excuse to fathom until you would perform actions on the Amazon Alexa.
  4. The beautifully designed apps exist, but most of beautifully designed apps are for restaurant advice.
  5. The app should be fraught with the virtual characters you may never trade based on fire.

Some might say the output is more interesting than the input. :)

Further reading