Hello all how are ya? :3 I'd like to preface my submissions with a journal entry! I'm a computer scientist, coder, and all around huge computer nerd. I suck at drawing but I've always loved the artwork and content that came from the furry community. In some of my hardest times I accredit the support from members of the furry community which I'm forever grateful for <3
I first got into "AI" when I was accepted as an early beta tester for OpenAI's GPT-3 beta build and I became instantly hooked. As a coder/computer scientist the transformer(LLM/Large Language Models) algorithms to me was the most fascinating bit of technology to be released in quite a while.
While delving into research on "AI" (LLMs) the mystique slowly began to fade. I've come to realize.. this isn't "my definition" of "AI" at all. Under the hood; clearly this isn't sentient at all. It's just a few cleverly stacked calculus algorithms in a softmax(the real calculation that allows LLMs to work/watch 3brown1blue on youtube for more info) and that's it. Humans finally discovered the best word guessing algorithm in the world.(For now)
Concluding my research on LLMs suddenly I found out about "Stable Diffusion."(v1.4) Wow! Now we have an open source "AI" model that can generate Photos/Art/Drawings. Thus began my research on SD1.4. During this time my love for computer science and the furry community began to collide. As my research progressed I obviously leaned heavily into furry related generations.
With LLMs the same with SD.. the mystique began to fade, but into fascination this time! Again.. this isn't "my definition" of "AI" at all. This is really just a convolutional neural network on top of Perlin Noise algorithms. I know.. I know.. you're saying "What the f**k does that sentence even mean?" Well I'll tell you!
Very simply put. In the 70's at Disney Studios this clever guy that went by Perlin developed a mathematic algorithm to quickly texture objects in animations! With a some googling you'll see "Perlin Noise" just looks like a bunch of random static like on an old tube television. But its not actually "random." The pixels are laid out in a specific fashion to create textures on like a canvas. This is exactly how every Stable Diffusion generation starts. As calculated (latent) noise!
Like LLMs; Convolutional Neural Networks(CNNs) just guess.. Instead of guessing the next words it guesses the next pixels(surrounding pixels/watch 3brown1blue on youtube for more info). Just a bit of more clever calculus on top of math we already figured out. It's not sentient, it's not evil, it's not going to kill us. Just humans getting slightly better at math and code.
Onto my main point.. As much as I love the furry community I found SD as a way to contribute and give back. I'm not trying to get rich, and I'm not trying to take anyone's job. I see coding as a form of art and expression. To me looking at the source code for Bitcoin is like how an art connoisseur looks at the Mona Lisa.
I've spent over two years learning A1111/ComfyUI/Python with many "AI" models from SDCheckpoints/Loras/Controlnets/Esrgans/Etc and the code tying it all together to make what I'll be posting. I spent my first year making "aislop," learning the craft, and getting better. I don't like "aislop" and I think everyone should respect manual artists by becoming experienced at CNN usage before posting.
If you continue to disagree with "AI' and saying it will take artists jobs I have a word of advice. You're not going to become rich making $5 sketches. When I talk to real professional artists about AI they could care less. Why? Lets take fursuit artists as an example. They draw/color/digitize/3d model/order material/tailor/manufacture/sell their product for decent sums of profit... "AI" will never take this from these artists. Successful artists always have a multifaceted approach for turning their passion into a career and you should too! Please stop hating on other people hobbies when you should be putting that energy into getting the skills that will push you passion into a career.
