Statement on AI

Why doesn’t ergot. consider work produced by AI?

The short answer: we believe that the aims, ideas and goals of creative work produced by AI are diametrically opposed to the sort of work sought and published by ergot.

The long answer: while there are some interesting projects being created by AI (a notable example being Inside the Castle’s Castle Freak Project) we believe that the act of producing new literature—especially experimental and innovative literature—is its own pursuit, and that the finished, published work itself should be a record of that act.

We like to think of the work published on ergot. to be in some ways aligned more with research than pure entertainment. Consider the way that thinkers working in other fields (like the sciences, mathematics, philosophy among others) approach their work: the work starts out with a question, or a general direction, then these thinkers set out into the unknown in pursuit of that question. This work—the research itself—primarily involves working for long periods in a setting like a lab, in the field, in the archive, a working group and so on. The work itself is a struggle and much of the time they may be unsure whether the direction they are heading is valid, or will be ultimately fruitful. These thinkers may set out in one direction, hit a dead end, then go in a different direction. They may even find that their original premise was in fact not correct. When thinkers in these fields believe they have concluded a section of their research, they put together a polished record of their work to share with others and present it, usually as a paper published in a scholarly journal.

Of course creative writing doesn’t have the same high bar for falsifiability, source acknowledgement or technical rigor as these other areas, but there is something to be gleaned from the approach taken.

With AI, the work must start out with the end goal: a prompt is fed into a large language model that then uses the large amount of existing works it had been fed to produce a long piece of pseudo-random text. The instigator of this already knows more or less what kind of final piece of text is going to be produced, and the large language model fills in the rest. While the person who fed the prompt into the LLM may go back and change some of the resulting output by hand, it is hard to imagine a truly original work arising from LLM output. It is important to note too that the work of generating the text was not undertaken by the person who prompted the AI, but by those who created and trained the LLM and, most importantly, those whose work the LLM was trained on.

If a thinker in one of the fields mentioned above were able to just put a prompt for their research into a LLM, and have it spit out a complete and valid paper, that would just indicate that the research was not novel enough to be worth doing in the first place. Any data, equations or references generated in the work would almost certainly be hallucinations (read: fake). We would hope that the resulting paper would not be accepted for publication.

If the connection to research isn’t compelling, here’s another analogy. You’ll often see this in the outdoors, where visitors want to travel a trail just to get to a good view point in order to take a picture (usually to put on social media). This results in large numbers of visitors to these areas hustling through the trail as fast as possible so they can get to the same view point, stopping for a second to take the same picture, posting that generic (usually poor quality) picture to their social media, then leaving. But in many cases there are very interesting and beautiful things—rocks, animals, flowers, fungi—to be seen along the way. Those who are observant and adventurous enough might even find an unmarked side trail that leads to a seldom visited feature that is as interesting as the view point. Not to mention the people who travel the trail in a new or uncommon way, whether on unicycle, horseback, hiking it backwards or blindfolded. Too much fiction is like the same low quality photograph of the Grand Canyon taken from Hopi Point.

A great deal of really interesting work has been created by dedicated thinkers setting off in one direction, then finding some totally different, but very interesting, result or path along the way. The literary equivalents of these are as—if not more—interesting to us as the result reached at the end.

All creative work starts with an end goal in mind, but we want to see the struggle and the journey that happened along the way, the interesting results stumbled upon, the side trails found or forged, the obstacles overcome or that prevented that end goal from being reached at all.