Stable Diffusion, NovelAI, Machine Learning Art - AI art generation discussion and image dump

This morning Bing AI suggested "a meme" when I clicked the Surprise Me button and it returned a picture of a cat with garbled text on it. So I decided to see what it would do with prompts like "funny dog meme" and stuff. "Funny lizard meme" returned a bunch of deformed Pepe the Frogs so to see what it would do with "funny FROG meme" I entered it and, well, this is what I got:

View attachment 5414364
I don't get it.

1697326399149.png1697326415486.png

Okay, I deserved that. Damn.
 
I feel it comes down to liability; MS/Dall-E doesn't want heat because somebody dared to use the tool for nefarious purposes.
What liability? It's a tool. I know this is what they claim but it's kinda bizarre how this is the argument as you wouldn't blame Adobe for bizarre conspiracy theory photoshops either, nor would you blame the VLC guy or whichever department does the inbuilt windows media player for people watching child pornography with it, or the manufacturer of the screen, or intel for providing in hardware video decoding etc.. The guy who wrote the prompt is responsible for what the program produces. The "AI" doesn't have agency. It doesn't "decide". That's also how the legal situation should be. (that said, this isn't even about legaility, the prompts people mention here being blocked aren't even "illegal" in any way. A corporation doesn't decide what's illegal and what's not to begin with and it's also not their place to enforce laws) If you let that train of thought continue otherwise, you'll not really be allowed to do anything by yourself as you basically give them the permission to take all your agency away for the sake of "safety", whatever this even means in this context. It all really comes down to that they don't really want you to own what you use, so they can control how you use it. We see this trend everywhere, not only here.

When they'll finally shove AI into windows, you better believe it's gonna be used to make better use of your telemetry by being able to analyze your usage in much more detailed and usable ways. (something that's actually kinda hard to do with more standard computing approaches, this is the real strength of natural language processing LLMs can do) From there it's only a small step, one manufactured moral outrage (probably involving the safety of children because why wouldn't you support that? Are you a pedophile?) to "we have noticed you're using our OS to propagate unsafe content and misinformation. Our ToS doesn't allow the usage of our OS for this." Do not believe your government will protect you, either. They'll probably have their custom list of demands what is deemed "unsafe content" by nationality. That is the real danger of AI. Of course that's not going to happen overnight. They'll slowly boil the frog, so to speak, by making people fully dependent on this technology first.
 
A (joking) reminder to all Bingheads:
You will never be a real AI Prompt Master. You have no loRas, you have no Embeddings, you have no ControlNet. You are vramlet twisted by poverty and consumerism into a crude mockery of text2img perfection.

All the “praise” you get are two-faced and half-hearted. Behind your back people mock your images. Anons are disgusted and ashamed of the cooldowns you have to endure, your “friends” laugh at your rejected prompts on the other threads.

Coomers are utterly repulsed by your model. Thousands of SD checkpoints have allowed coomers to generate lewd with incredible efficiency. Even DALL-E prompts who “pass” the filter look uncanny and unnatural to coomers. The lack of 1girl with big tits/ass on your gens is a dead giveaway. And even if you manage to circumvent the filter and get bare boobs or funny racist memes, Microsoft will "correct" and block that the second they see your pitiful, laughable jailbreak slop.

You will never be free. You wrench out a fake smile every single morning and tell yourself your gens are good, but deep inside you feel that you really dind't asked for that, the ChatGPT just changed your prompt and generated another politically correct, sfw image.

Eventually Microsoft will implement a paywall - you’ll buy a subscription, 100 gens per month, $ 39,99, no coom or pol prompts allowed. The glowies will have access to all your prompts, disappointed but relieved that you don't even try to bypass the filter anymore. They’ll putt your name on a list alongside with your social security number, and every fed for the rest of eternity will know how many times will prompt for cunny. Your gens will decay and go back to be normie shit and all that will remain of your legacy is some edgy memes that you generated during the first few days.

This is your fate. This is what you chose. There is no turning back.
 
Idk what's going on with DalE3 but I noticed that when you have no boosts left and it says you have to wait like 5 minutes, you just need to reload the page after like 30 - 40 seconds and the new pictures appear. At least it works for me.

Or it's just that the timer is wrong but if I don't reload it takes longer.
 
Last edited:
I am absolutely seething mad at how neutered this thing got. Sick of this gay dog.
This is the 3rd rewording and it's still not good enough for it. My shortlived AI excursion is ruined.
View attachment 5416370
You're telling me. Ingrid Bergman, no. Marlene Dietrich, nein. Joan Crawford, nope. Ginger Rogers, 23 skidoo. Bullshit, it's practically useless to me now.
 
OIG (9).jpeg OIG..jpeg

OIG.MauOdqE1eeVwr.jpeg OIG (23).jpeg

I've been messing with this on Bing all day, this really feels almost like magic at this point. It's pretty much impossible to get exactly what you want, though. What it needs to be truly mindblowing is the ability to make specific adjustments after an image is generated. But I don't suppose that's gonna be possible anytime soon, since it seems that would require the AI to actually "understand" what it's doing.
 
It's pretty much impossible to get exactly what you want, though. What it needs to be truly mindblowing is the ability to make specific adjustments after an image is generated. But I don't suppose that's gonna be possible anytime soon, since it seems that would require the AI to actually "understand" what it's doing.
Microsoft/OpenAI adding an in-painting function to Bing would be great, but unfortunately, I don't see it happening anytime soon. You could try taking their result and in-painting on a separate model, though the quality will likely differ.
 
  • Feels
Reactions: The Last Stand
Back