AI seems to be quite the hot topic lately, and while the pace at which I write is generally unsuited for current events and pertinent topics, there have been several different “coincidences” of ideas that have confluenced together to make me realize just what the depth and breadth of this whole thing is likely to be for me as a humble, small-time creator.
I’ve dabbled only a little in generative AI so far. When ChatGPT came out, I bantered around with it for a little while, asking it to do ridiculous things like rewrite the first chapter of Great Expectations as Temmie from Undertale (it claimed to not know how to do that but did it anyway), and to write a Bluey fanfiction (for some reason it kept having the girls rescue a lost puppy, even though I pointed out that would be like finding an abandoned baby). I tried out some free image generators too, but they just couldn’t understand that an animal that’s a combination of an avocado and an armadillo shouldn’t be a fruit and a critter fused together at the hip. Overall, I’ve decided to write the whole thing off and not worry about it at all, even though some of the tools I use have touted their integration with this or that GPT. I recently realized that things were getting more serious, though, when I was looking for a recipe for dairy-free clam chowder. That’s right, this story begins, like so many of mine, with food.
I can’t eat dairy or eggs, so in my quest for a vegan clam chowder recipe I found an article that was both hilarious and disturbing at the same time. (Okay, I get that it’s really nothing more than combining clams and a cream base with some vegetables, but I wanted some proportions or something at least) You can read it for yourself if you’re feeling courageous, or just wait until I pick it apart.
Okay, so the clam shells are cute, they’d be a little crunchy but you could take them out if you want. But it also looks like there’s no clam meat in them, which is odd…
It only took a little more scrolling down to realize something was terribly wrong.
Call me crazy, but I don’t want to eat anything called Pooom Ccconlut Owder, even if it was made by Cctinenihua. Also what kind of can has a spoon integrated into the lip?
And now we’re sauteeing the parsley, right in the pan with the clams and the… sweet potatoes? That would make a heck of a soup. Oh, there’s actually a recipe in here too, or something like one. My editorial cheek is in square brackets because they’re awesome. (the brackets, not my cheek, although… oh never mind)
To make a delicious coconut milk clam chowder, you need to start with fresh ingredients. [You certainly wouldn’t want to use spoiled ingredients.]
Begin by sautéing a mixture of chopped white and red onions, mixed peppers, and shiitake mushrooms in a large pot. [None of those things go in clam chowder.]
Add a veggie mixture of bok choy, cilantro, and lemongrass, and cook until the veggies are tender. [So clearly we’re making veggie stir-fry instead of clam chowder.]
Then, add your fresh shellfish, such as clams or oysters, and cook until they open up. [What? Oyster chowder? That would make a gross soupy mess and smell like… well, oysters!]
To thicken the chowder, add flour and stir until it is well combined. [I think this is the only thing they actually get right.]
Next, add a mixture of bottled clam juice and fish stock, and bring the mixture to a boil. [Since there wasn’t enough clam essence in the clams, or fish essence in the… fish you didn’t add.]
Reduce the heat, and add coconut milk, curry paste, and tomato sauce. [Oh, so we’re actually making seafood curry cioppino all of a sudden and no one told me!]
Simmer the chowder until it thickens, and season it with salt, fresh ground pepper, and lime zest to taste. [I have nothing more to say at this point.]
And then, to top it all off, the absolute crowning glory of this entire ridiculous pile of shenanigans, the true indication that things had gone horribly, horribly wrong.
After making that recipe which definitely wasn’t clam chowder, how about you relax with a tiny plate of canned clam chowder, and a side of sticky notes with more chowder on top of them, and a nice warm mug of Coconut Coltozi Cloouva a OIO Coconml NIILC. The perfect reward at the end of the day.
In case you haven’t figured it out yet, the entire article was written by generative AI, the images were spawned by generative AI, and it’s possible that the entire website is just some poor lonely algorithm’s fever dream of owning a seafood company. There are many other articles linked to in that one, and all of them have commensurate levels of nonsense. I didn’t even realize until I started constructing this article, but the alt text for each image (copied into the caption, since the alt text I used is much better) is actually the prompt that was used to create it. So that pile of chowder-covered sticky notes are “frequently asked questions cards.” That clears that right up!
Right at the same time that I started drafting this article, my employer announced that we all had to take a mandatory generative AI training course. As a work-from-home software engineer, this seems like it would make sense, except that I don’t work with generative AI products. I don’t even write code anymore, the platform that we use for application development is mostly drag & drop and just typing in variable names. But, that platform is also pushing its own gen-AI products pretty hard, so the likelihood that I’ll run into them is fairly high. These were the best (worst?) takeaways from the course material:
If you’re building and training your own generative AI models, there are certain things that you can use to mitigate these risks. When you're setting up your training data, make sure you use diverse and representative data sources. Make sure that you're constantly reviewing and testing the output of your generative models for bias. Catching biases early on can help ensure that you train the model to mitigate these biases.
Also, transparently disclose training data and algorithms. Making users aware of the data that you've used to train the model will promote trust and reliance on your model. Make sure that you're constantly assessing the quality of data that goes into training, and that you're implementing measures to detect and prevent fake content from being used to train your models. If you're using other people's work, make sure you obtain user consent before generating content. And finally, make sure that you yourself know your model’s capabilities and limitations, and you communicate this very clearly to the users of your model.
Is that something that people actually do? “Hey, what are you up to this weekend?” “Oh, I was thinking of building my own gen-AI model and making sure it's free of bias and doesn't propagate harmful information.” “Sounds like fun!”
In addition to these commercial use cases [translation, summarization, etc], ChatGPT can also be useful to augment human creativity. Its ability to generate creative and thought-provoking pieces of literature and poetry can be used for a variety of purposes, or as a starting point for creative exploration.
I'm sorry, what? Every time I try to get it to tell a story, it gives a regurgitation of the three-act plot structure with the theme or topic just pasted over the top. Ask it to tell you a story about anything and the thing starts of in a peaceful, happy existence, then either finds something new that makes it question its happy existence and sets out on a quest to make sense of its life, or just wanders away for no reason, then learns lessons and makes friends and comes back happy. Just like every other story ever.
I could go on. I really could. But it wouldn’t help you or me or anyone. If you’ve used the internet in the last few years, you’ve seen the problem. It’s everywhere and it’s only going to get worse before it gets better. But how could it possibly get better? By people standing up and making a decision about it.
Before factory farming, food was grown at a smaller scale. People worked the land by hand and with draft animals, providing for their own needs first, then those of their communities. As technology changed (note I didn’t say improved) and more food could be grown on more acres, the quality of that food started to decrease. We went from nutritious vegetables grown in healthy soil to what are basically carbon sinks fertilized with artificial chemicals and supplements.
Before generative AI, content on the internet was created at a smaller scale. People wrote their own blog posts and did their own voiceovers, providing for the needs of their fans and audience first, then those of larger communities and sites. As technology changed (note I didn’t say improved), and more content could be created in less time, the quality of that content started to decrease. We went from mom food bloggers sharing their latest gluten-free biscuit recipe to algorithmic horror stories like the one I detailed above.
Cheap factory food has pervaded our culture just as thoroughly as garbage content has. Nearly every grocery store is full of carbohydrate molecules that will keep you alive, but leave something to be desired. Fast food is cheap (or used to be) and keeps you hooked with fat and salt and sugar. But we know there’s something better out there.
The organic food movement has gained a lot of traction in recent years, motivated by people’s desire to eat things that haven’t been covered in pesticides and herbicides and any other kind of -cide out there. The funny part about this, to me, is the fact that organic food is set aside with its own label and designation. The opposite of organic is usually synthetic, but the food still comes out of the same ground and isn’t made of chemically-fused biopolymers (yet). Anyway, when organic foods first started appearing in grocery stores in the 90’s, they were this new and exciting thing that were an improvement over “regular” foods. In actuality, “regular” foods were a degradation of the way that food had been farmed and cultivated for millennia. So if anything, non-organic foods should be the ones with labels and certifications and such, since they are they deviation from the true path of agriculture.
The recent proliferation of AI-generated content mirrors this decline. We have more and more stuff to read/watch/play/listen to than ever before, but is any of it actually intellectually nutritious? Organic foods used to be relegated to specially-labeled sections of the produce department in big grocery stores, but now entire specialty businesses are thriving on selling only organic-certified produce, meats, and anything else that people need. Fresh Thyme, Natural Grocers, and Sprouts Market are just some of those that I’ve encountered, I’m sure there are more. When you go there, you know what you’re going to get and it fits in with your expectations and ideals about food.
So, we need to do the same for content.
I recently had front-row seats to a historically defining debate, at least in the history of Catholic fiction writing groups. By now you’ve probably noticed that I’ve been involved with LegendFiction in varying capacities over the past several years, and if you’ve looked at any of their materials you’ve seen that they rely very heavily on generative AI. It’s always been something of a side note, a window-dressing kind of thing that they use to decorate the place with but nothing more. Until, of course, now.
This is the post that started it all. Read it if you like, and/or watch the video, but one of the main points is that generative AI draws from existing works of art and creates new ones, just like human artists do. But that isn’t really how human artists create art. Maybe they do during their formative years, copying shapes, postures, tones, colors, etc. (plots, characterizations, descriptive language, humor, etc.), but once they start to develop their own unique style, they take all of that data they’ve ingested through studying other people’s art and synthesize it through the unique, God-given process of their co-creativity. They apply their own personal life experiences, specific tastes, where they live, what they ate for breakfast that morning, and so on, and use it to create something that is truly and utterly unique. Which is something that a computer program, lacking all of those things, can’t do.
I promised a debate, so here’s the other side. I joined up with the Inkwells & Anvils community much later, only the beginning of this year, but have already been enjoying it tremendously. What a coincidence that they would also release a document outlining their stance on AI just a few days after LegendFiction:
https://www.inkwellsandanvils.com/post/art-artists-ai
It’s basically the exact opposite. I think it also serves to show that not only do we need both sides in this debate, but we need stances to be clearer and more visible. When you walk into a Fresh Thyme or another natural grocery store, you know right away because of all the signage and marketing materials that they used to get you through the door in the first place. When you walk into a standard grocery store (un-natural?) you see the big national brands that you associate with the homogeneity of the industrial food system.
Imagine what things would be like if you were able to tell within a few seconds whether a creator or an artist or a writer used AI for their end products, not by having to guess, or by something looking a little off, a little too real or unreal. If the artist was completely forthright with their stance on AI and made it be part of their branding and marketing. “100% human-painted artwork” or “completely AI-free,” or even “created entirely from text prompts.” That would certainly change how people approached new things, not with a mindset of fear and uncertainty, but with confidence that they know what they’re seeing and it’s what they want to see.
And, since I certainly know the importance of eating one’s own dog food (though I wouldn’t recommend it, it’s pretty awful), I am prepared to make my position known right now.
As of today, author Zephyr Thomas and illustrator Michelle Thomas, operating under the label ZMT Books, promise to our audience that we will not use AI-generated artwork or text in any capacity. Every photograph you see on this publication was either taken by Zephyr or someone else and credited accordingly. Every piece of digital art was either digitally drawn by Michelle or someone else and credited accordingly. And none of the text content was produced by AI systems, except where indicated for the purposes of satirical roasting.
So there we are. Maybe someday there will be a certification program to prove that your content was or was not created by a human or by another process, and I’d proudly wear that badge. Until then, if you are a creator, consider making a statement like this part of your thing. If you’re a creative consumer, encourage your favorite creators to do the same. We’ll all get through this together.
Want more timely analysis of current events? You should probably look for that somewhere else. But if you want family-friendly entertainment in multiple forms, subscribe and feel confident that this was all written by a real person.