What do I mean when I say AI?
AI is getting thrown around everywhere these days. Programs that have been around for years are suddenly being called “AI”. The reality is there are two primary forms of program that are the true AI of the current age. First is Large Language Models and the second is Generative AI. I’ll get into that a bit later, but for now, a bit of background.
Noise Engines
For years now, AI systems have been a bit laughable. Whatever they generate is just a bunch of gibberish. No one would use the language programs as anything but a joke. Heck, the ongoing meme was feeding an AI 1000 hours of something and having it spit out ridiculous scripts. For the generative AI, it would take a prompt and drop a chaotic miasma of randomness that might vaguely resemble something. I like to call these early AI noise engines. Noise engines are never going to generate anything polished enough to be of value. Or more accurately, nothing of value as more than a seed for other ideas. This early version is great to form a scaffolding to generate ideas for artists without actually being anything more.
Generative AI and Large Language Learning
Those noise engines were just the early versions of what we would come to know as the more advanced AIs of the present. The AI learned by going over existing works, trying to predict what was a readable/artistic work based on the words associated with it. The AI doesn’t think, so much as just guess what should happen based on what it has dealt with relating to the prompts. To further this development, the developers needed data. Lots and lots of data.
They obtained this data by scraping everything from the entirety of the internet. Because the developers were doing this as a non-profit venture, the law allowed this. It was an academic matter, so the law allowed looser usage. Of course, what even some of the programmers who helped develop it have come to realize is that the companies backing the projects were entirely unethical. The moment the programs were capable of quality work, they suddenly shifted to a for-profit model. Even if they never scraped another bit of data, they’d already benefited from piles of copyrighted works. Works they would never have been allowed to use had they started with the understanding they would be doing it for profit.
Ethical vs Unethical
Arguably, some of the later bandwagon participants have tried to build their systems on ‘ethical’ methods, but it is quite hard. How can they compete with an internet worth of training these early AI already managed? Some companies have done deals to allow for scraping of their sites, much to the chagrin of users. DeviantArt and Adobe have both fallen victim to their greed to scrape for AI. It has pulled many artists away from them. Lawsuits abound for artists and writers whose copyrighted works have clearly been used without their permission. Companies have even been forming that specialize in protecting your work from being scraped by AI without your permission. It’s incredibly difficult to find an AI that was ethically trained.
Using these AI is already sketchy, but then acting as if you are a writer or artist just by using prompts to get something acceptable is disingenuous at best. I think there is an argument to be made for using it to create an idea that that you develop yourself. Not a finished product or something you touch up, but something to spark creativity. If you paint an entirely new picture and use the AI to get the layout idea worked out, sure. If you ask LLMs to help you flesh out an idea, then sit down and write a book yourself, I accept that. Do I like it? Not so much, but I do get it. I grow more wary when you start adding art from AI or entire passages to your works. The democratizing of art, the allowance of budget-less entities, and all such arguments hold very little weight for me unless you’ve managed to somehow find a truly ethical AI. I have no interest at all in buying an AI written book, nor AI generated artistry. These are obtained on the back of real artists and at the expense of those who took the time to build their skill sets.
When I will and won’t use it
So when will I use AI? I admit that when this all started, I tried out some of the AIs. I generated some images, asked it about what it might suggest on handing some of my stories. I’ll be honest that it is incredibly tempting to work through prompts to generate the art needed for my TTRPGs. I just can’t bring myself to do so however.
What I will do is use those older noise engines, where they still exist. Granted, they have become ever more rare, but so be it. One of the art pieces I have considered using in the final version of Weighting Styx and Stones was a noise piece that I went on to modify after feeling inspired. The cover I had planned for The Platinum Chain had an element I used AI to work out, but have chosen to redo through traditional artistry. A few story ideas were spawned from those days tinkering with LLMs. That said, I don’t have any intention to further work with the more advanced AI unless I manage to find one that is both quality AND ethical.
My stance on the use of my work to train AI
So where does that leave me for the use of my copyrighted works to train AI. Let me be clear, I disallow the use of any of my works to train AI without my expressed written permission. If such permission is granted, it will come at a steep price. If my works are scraped without my permission and I discover it, you accept the resolution outlined in my legal documents. In short, don’t. Even automatic systems that scrape the internet are making the conscious choice to plagiarize my works. There is no quarter to claim a lack of foreknowledge when one actively chooses to unethically scrape from my work.
