Quantcast
Channel: Rainmeter Forums
Viewing all articles
Browse latest Browse all 1354

General Discussion • Re: Training an AI model to write rainmeter skins and lua scripts.

$
0
0
I'm normally very anti- "AI"-generated content for almost every medium, but in the context of Rainmeter skin creation and adjacent, the lowest I can come up with is a "meh", on the basis that it's not going to have a huge impact, negative or positive:

To start with the positive: I'm not against someone using AI to streamline some initial "boilerplate" steps on skin creation, and to treat it as a possible tool in a toolbox instead of being some final solution. I think it would still take a disproportionate amount of work into the LLM depending on what you expect the user get out of it, but in general, there is a small something there, using AI or other tools or methods.

The basis of Rainmeter is customization and personalization. Even if you put in a lot of time and work into a Rainmeter-specific LLM with quality data, the user will always want to personalize the result further, either now or later.

Briefly trying some prompts with a LLM like Perplexity, the current output of AI Rainmeter skin creation seems to be mostly (and sometimes outdated, example) instruction and one or two measures or meters. Versus someone asking for a skin bringing nothing, someone asking a LLM and bringing that (and what they might have read along the way) is always a step up from that as far as I can see.

It's commendable to respect skin licenses, but I can guarantee that if this were a lucrative area, from what consistently happens with the main LLMs (just the latest example), someone else would have no problem ignoring the licenses and requests of skin authors.

Statistics: Posted by Alex88 — Today, 2:23 am



Viewing all articles
Browse latest Browse all 1354

Trending Articles