Adventures in AI Text Generation

After my success--or "success"--with the AI art generation in previous posts, I thought about the text generator I'd been messing with a couple of years back (not my online generators--those work more on the basis of Mad Libs than true text generation). Decided to poke around and see what of the new generation of text generators available to the public was like, and I found one based on the GPT-2 neural network set up as a Google Colaboratory notebook by Max Woolf, saved a copy of it for myself, and set to work.

The text-generation model has been trained to start with on a bunch of text from the intartubes. You can use it as-is, or you can fine-tune it by feeding it a bunch of data and letting it chew over that data for a number of iterations, calculating probabilities.

( So naturally I scraped the Zoe Chant website and fed a bunch of shifter romance descriptions into it. )

A quick update...

...on the giant dump of generated shifter romance book descriptions I asked a neural network to give me (see previous post).

( During one iteration, it got stuck on boots )

Adventures in AI Text Generation: Blog Edition

I have done what nobody everybody was clamoring I do: I have downloaded three years of my Dreamwidth posts and trained an AI text generator on them!

( Dare you see what lurks beneath? )