Thursday, October 19, 2023

On getting into AI art making and the ethics of it

 I am writing this post due to my getting into making AI art after being interested in the possibilities and wanting to use it for game modding.

Now, before I continue, I want to make a few things clear. I deplore the use of AI creation of any sort for evil intent. I do not ever plan to sell anything created by AI, do not ever intend to claim it as my own work (as it's generated by a computer, not my own actual skill), and I only ever plan to do AI creations of any sort in compliance with all applicable laws and with all proper safeguards taken against it causing financial, legal, or moral harm to another person. If I was ever given a request to take it down by proper authority, I would comply immediately.

Now that I got that out of the way, let me first explain what AI art is for the people unaware, then go into why I wrote this post.

First off, AI art is artwork, soundtracks, animations, or other media done by a computer being fed examples of currently existing material, then using procedural generation to attempt to make its own artwork based on this data. In practice, it makes hardly anything one would call truly original. It's in fact derivative of existent art, often with the resultant creation being a chimeric fusion of actual effort.

Legally, it's not entirely clear yet if it's plagiarism, parody, or falls into something of a grey area. So far, there are no laws against my use of it I'm aware of, and I'm just a mere hobbyist at best.

I use Stable Diffusion, an open-source protocol. I originally got interested when I saw Bing's Image Creator, but these days it's so locked down, sanitized, and refuses to make any results of anything that could offend anyone (but given it's run by Microsoft, who lobotomized their AI Tay when the internet taught it to be an unironic Nazi, this is no surprise). Ergo, I decided to try making my own with open-source tools.

The art I do is still images. I favor a 2D anime-based art style, and Stable Diffusion, provided you have a proper checkpoint (base data for your chosen art style) and Loras (models and other data that serve as addons to Stable Diffusion), you can make some decent material, albeit with a lot of work.


In fact, I'll be blunt. While there are many people who rage (with some degree of truth to their credit) AI art is theft, the overwhelming majority takes a LOT of tries before you get results that look even remotely good. I've lost count of how many terrible results and outright cosmic horrors I've produced before the generator spit out something passable.

Stable Diffusion works on a tag and weight system. It uses tags much like "Booru" style sites. For example "1guy", "car", "road" would be a basic prompt to specify you want to see a guy with a car on a road. Of course, you need to further specify if the car is in motion, if the window is down, and all the other details if you want more specific results, else the AI generator result can be super random. It also requires certain topics be given high or low weight, to give certain details emphasis or deemphasis, as the case may be.

Again, let me be absolutely blunt. This is a piss poor substitute for actual effort by actual artists. I have paid artists for work before, and I found it to have far more love, care, and charm than even my best AI creation. AI tends to hate and not understand subtle flaws that give real art its charms. It often makes horrible mistakes like deformed limbs, bad proportions, incompetence at drawing light and shadows, and many other botched-looking failures at making an image. I am going to admit I'm a cheap man with no art creation skills to my name (drawing straight lines is nigh impossible for me and I have no idea how I didn't totally flunk art class). I'm sure even the crudest stick figure artist makes stuff way better than I could. I bow in humbled awe of anyone with actual artistic talent.

Now, I want to cover the controversy over AI art. There is a lot of rage it's going to put real artists out of work, it's crap knockoffs, it's stealing what other made money off of, and how it can be used for all sorts of evil intentions and in fact has been done already. I do not deny any of this, people have been caught selling games with AI art without saying so, it can be considered plagiarism of material others have done since it's all based on preexisting data drawn from publicly accessible sources, and yes, it's very arguable even for parodic purposes it could be called plagiarism in many legitimate senses.

Conversely, I want to say the following. The genie is already out of the bottle, it's too late to put it back in. This is a tool like any other that can be used for good or evil, and I agree it should be governed and restrained by the firmest sense of ethics possible. Whether it's detractors like it or not, it's not going away, so I advise focusing on restraining it's potential for abuse instead of cursing its mere existence, too late for the latter now.

Finally, I just want to say my conscience on this subject is generally clear, given my the above. I again regret and deplore use of AI for evil. I greatly enjoy audio parodies on YouTube of AI versions of our presidents playing Mario Kart, meme videos of Dagoth Ur from Morrowind saying the most goofy dialogue people can dream up, and enjoy meme parodies done in AI that blend art styles and genres one does not usually see. This does NOT have to be a tool for evil, it can be a means for harmless fun.

It's merely my solemn hope and prayer it can remain that way, or at least the potential for evil use can be mitigated as much as possible.



No comments:

Post a Comment

Asiago Pressato, Pecorino Calabrese, and Green Wax Irish Cheddar, my impressions

 It's been a while since my last cheese review, so let's make up for that, shall we? This time I ordered a pound each of Asiago Pres...