When Chimeras Dream

The fiction here opens with first lines from famous books in the public domain. What follows builds on text suggested by the open-source large language model GPT-Neo. In most cases, GPT-Neo built on a pairing of those first lines with a short plot synopsis. The catch? The synopsis and opening lines come from different works, producing a menagerie of narrative chimera.

There’s been a lot of talk about language models like this putting writers out of a job. Some folks even worry they may be sentient. The reality? They complete patterns based on prompts. Feed them some text, and they produce more following the same pattern. Their output is more like free association or a dream than plotted stories, hence our title—When Chimeras Dream. Given the nature of these outputs, it seemed fitting to use them as prompts to build on. In this way I was able to take advantage of the age-old observation that it is easier to rewrite than to write. That is, the tech wasn't an author, it was a creative catalyst.

If you’re interested in how to use models like GPT-Neo, don’t worry. The fiction is followed by some facts—like a primer on how one “teaches” algorithms to “understand” what a document is talking about, paired with discussions of AI and the law. Yes, this site is a Trojan horse. So dear reader, choose your path. Come for the fiction and stay for the facts, or take them à la carte. Also, despite being a cyborg attorney, not all of the tales here revolve around the law or technology. Hopefully, a few are just good yarns.

On this site you will be able to read books in the series for free, but purchasing Kindle versions will be much appreciated since the author is donating his net proceeds to support cornerstone projects at the LIT Lab. Be warned, there are facts, and some math, hiding after the fiction. Additional stories and facts will follow in subsequent volumes.

^† David Colarusso is the Director of Suffolk University Law School’s Legal Innovation & Technology (LIT) Lab where he teaches classes like Coding the Law.