What about AI?¶

My AI stance¶

Since LLMs are (with reason!) a divisive topic, here is my stance on them, presented clearly and plainly. Hopefully this will allow people to make an informed choice whether to use YaVTT or not.

LLM coding: I approve it with caveats¶

To me, coding is a tool. Whether I am coding a data pipeline or a scraper or a virtual tabletop, the code is a way to achieve something that works.

At the same time, LLMs are also a tool that consistently speeds up other work. However, when unattended, they tend to generate sloppy results. That is why I have a quite thorough testing process (automated and manual) that hopefully catches any incongruencies or plain errors.

My final take-away is: I approve LLM coding assistance and LLM agentic coding (it is a great tool, after all) as it allowed me to create YaVTT; I disapprove irresponsible use that leads to instability. Over time I started using more and more constraints in my LLM usage and now it follows a well-defined pipeline for each feature I want to add or bug I find. I also test the program thorougly as best as I can.

Art/text generation: I disapprove it¶

To me, art is not a tool. Art is something that to be classified as such, must carry intent. Intent is something LLMs, obviously, do not and cannot possess. What LLMs do is statistically predicting things. That makes them good at practical works that have a well-defined objective, but bad at anything that actually require thinking because LLMs do not think.

This is why YaVTT features a terribly drawn logo, quirky stick people, and why the Darkness Falls style looks like a site from the 2000: because I do not believe AI is capable of accomplishing these things, and pretending it may be able to is an insult to what drawing or creatively writing is.

In fact, every guide, README, text, etc... in YaVTT and friends is written by ME. All the docs here are written by me. All the text in the websites is written by me. It's not that I do not trust LLMs to write user-facing content; I may let it write something and edit it. But I prefer writing everything by myself, so I can be 100% everything carries MY own voice.

My final take-away is: YaRS has no LLM-generated content, no LLM-generated art, no LLM at all was used to write its rules, its items, its spells, its abilities. Same goes for the setting of Darkness Falls that is based on it. Also no guide has been written with LLMs.

The ethics of AI¶

LLMs are 100% based on stolen work. However, under capitalism, no consumption is ethical, sadly. That is why I try and avoid using LLMs when not needed, and use open weights models. And, at least, I use LLMs to create open-source programs that anybody can benefit from, with absolutely no intention of earning anything from this effort.

It must be said that "AI" is not just Large Language Models. AI is being used in many fields and it is helpful in all kinds of research. But that is another story: it was just to give some context.

Hopefully I was able to express my ideas about LLMs and AI in general. If you choose to use YaVTT: great! Happy to have you on board :) If you are put off from LLM use, I respect your opinion and maybe we can even discuss about it: contact me here: info at yavtt.eu if you want!

Now onwards, to new adventures!