|
Lord Stimperor posted:Kickstarter: train SD on all known internet porn I'm mostly for tech anarchy, just let everything happen the wheel of progress will spin whether we want it too or not, but drat if you think about this for a minute there are some uncomfortable possibilities. There are porn stars (or even not stars depending on how good the AI is at recreating a person it doesn't have many samples of) and, they agreed to do specific things on camera. If I then make realistic pictures of them doing porn things they did not sign up for that just feels super, super wrong.
|
# ? Sep 11, 2022 13:02 |
|
|
# ? May 29, 2024 17:55 |
|
BrainDance posted:I'm mostly for tech anarchy, just let everything happen the wheel of progress will spin whether we want it too or not, but drat if you think about this for a minute there are some uncomfortable possibilities. I mean, that's not any different from a deep fake showing a politician doing a Nazi salute (like that would even be career suicide these days) or something, right? What's the difference?
|
# ? Sep 11, 2022 13:17 |
|
BrainDance posted:I'm mostly for tech anarchy, just let everything happen the wheel of progress will spin whether we want it too or not, but drat if you think about this for a minute there are some uncomfortable possibilities. code:
|
# ? Sep 11, 2022 13:37 |
|
BrainDance posted:I'm mostly for tech anarchy, just let everything happen the wheel of progress will spin whether we want it too or not, but drat if you think about this for a minute there are some uncomfortable possibilities. just wait until the true degenerates get a hold of the tech and make an AI trained on all the child porn in the world
|
# ? Sep 11, 2022 13:38 |
|
One problem with just dumping a shitload of images into the learning process is that generally images on the internet are not labeled. Booru-based image generation is pretty much easy mode since people spend countless of hours tagging every possible detail in the images. If you wanted to use the same power for good, you could take tagged photos on flickr for example, then try to detect untagged features in other photos and possibly suggest them to the users. "Does this photo have Vibrant Foliage in it?", possibly even adding an overlay to the photo where it thinks it's seeing the thing.
|
# ? Sep 11, 2022 14:59 |
|
strange feelings re Daisy posted:"Waiter, I'd like my food to be fresh and delicious please, no garbage scraps thrown in the microwave" Sir, this is an Applebee's.
|
# ? Sep 11, 2022 15:11 |
|
Wheany posted:If you wanted to use the same power for good, you could take tagged photos on flickr for example, then try to detect untagged features in other photos and possibly suggest them to the users. "Does this photo have Vibrant Foliage in it?", possibly even adding an overlay to the photo where it thinks it's seeing the thing. This seems like something google or whoever could do as a next generation of captcha or something pretty easily, if it became useful enough for them to want it. We'd get refined tags beyond anything we have now in probably just days with someone as big as google.
|
# ? Sep 11, 2022 15:19 |
|
Loab, nude, trending on pornhub
|
# ? Sep 11, 2022 15:52 |
|
Gadget hackwrench, inflated, ahegao, crudely drawn Ms paint DeviantArt
|
# ? Sep 11, 2022 15:57 |
|
train an ai on all ai generated images for double the ai
|
# ? Sep 11, 2022 16:05 |
|
or you just need the right fetish and you won't need to wait for the pornAI https://www.midjourney.com/app/users/308716887762534401/
|
# ? Sep 11, 2022 16:12 |
|
John Cena! NO!
|
# ? Sep 11, 2022 16:48 |
|
mobby_6kl posted:It just needs to learn the anatomy and various possible activities, it could be anonymized and the results wouldn't be recognizable anyway. Then you could use it with your own prompts, e.g. I don't know what you mean by anonymized but the AI is far too stupid to know which words/images are and are not a real person in order to self-censor about that stuff. And even if you removed all labeled celebrities/people from the imageset, people could still use it with inpainting, e.g. erase/blur all parts of an image but the face and then use the AI to generate a new image around it. But then again, Photoshop can already be used for the same thing and while there are a few creeps being weird with it it hasn't turned into a huge trend or overturned society or whatever. Burning Rain posted:or you just need the right fetish and you won't need to wait for the pornAI https://www.midjourney.com/app/users/308716887762534401/ Thankfully not visible to people without a MJ account. I'm going to assume this is a picture of a white woman buying wonderbread?
|
# ? Sep 11, 2022 17:12 |
|
Burning Rain posted:or you just need the right fetish and you won't need to wait for the pornAI https://www.midjourney.com/app/users/308716887762534401/ lol, I noticed that guy generating his images.
|
# ? Sep 11, 2022 17:31 |
|
Comparing the different finetuned models: base stable diffusion 1.4: waifu diffusion v1.2: trinart2 115000:
|
# ? Sep 11, 2022 17:46 |
pixaal posted:Full training sets are still close to a million though right? Is that likely? I didn't think they were increasing VRAM that quickly? Edit: Well I guess the RTX 3090 has 24gb of RAM, I thought it had less. Nevermind. WhatEvil fucked around with this message at 20:52 on Sep 11, 2022 |
|
# ? Sep 11, 2022 18:23 |
The AI pornpocalypse and deep fake stuff is going to be a major issue eventually. People already have enough trouble spotting photoshops. Now imagine you can take any picture of anyone anywhere, press Nudify in the latest app and it knows there is a human figure in the photo and replaces their body with a nude match. Catfishing and scams for sure. Blackmail and abuse. Its going to get wild. As far as the actual hardware is concerned, if there is a big demand for AI generated content in general, then the hardware will trend in that direction and slapping a bunch of RAM on something is pretty straightforward if theres a market for it. May see actual dedicated AI hardware in consumer grade equipment in the next few years.
|
|
# ? Sep 11, 2022 19:30 |
|
RPATDO_LAMD posted:
No, just page after page of bodybuilder women making out When ranking random midjourney uploads I did also come across one definite attempt at doing child porn and reported that, so that's happening too
|
# ? Sep 11, 2022 19:39 |
|
RPATDO_LAMD posted:I don't know what you mean by anonymized but the AI is far too stupid to know which words/images are and are not a real person in order to self-censor about that stuff. RPATDO_LAMD posted:Thankfully not visible to people without a MJ account. I'm going to assume this is a picture of a white woman buying wonderbread? mobby_6kl posted:Someone's going to be out of work soon!
|
# ? Sep 11, 2022 19:51 |
|
Now do the laughing dog from Duck Hunt
|
# ? Sep 11, 2022 21:18 |
|
computer... enhance
|
# ? Sep 11, 2022 21:50 |
|
MrYenko posted:Now do the laughing dog from Duck Hunt It kind of flipped him around and made him more friendly.
|
# ? Sep 11, 2022 22:14 |
|
New monster hunter looking like rear end
|
# ? Sep 11, 2022 22:20 |
|
Grabbed this from someone else "ancient electric guitar ensnared in vines in a dark forest, dramatic lighting, surreal, 8k"
|
# ? Sep 11, 2022 23:38 |
|
I did some img2img on certain image that people should recognize. Windows XP background
|
# ? Sep 12, 2022 00:17 |
|
IShallRiseAgain posted:I did some img2img on certain image that people should recognize. Windows XP background Very cool! What did you use for the prompt?
|
# ? Sep 12, 2022 00:27 |
|
WhiteHowler posted:Very cool! What did you use for the prompt? 3: A fantasy landscape, Dark Souls, Silent Hill, George Innes, Greg Rutkowski 4: demonic wastelands, Stephen Gammel, Gustav Dores, Zdzisaw Beksinski
|
# ? Sep 12, 2022 00:36 |
|
anyone know where i can send my resume to Riot Games?
|
# ? Sep 12, 2022 02:14 |
|
Posts like these two are putting an idea into my mind of making a Myst-like or a VN using SD images. Maybe something akin to the old ICOM games like Shadowgate and Deja Vu. Not actually going to do it, but it kinda makes me want to. "gateway_between_life_and_death,_oil_painting_by_James_Gurney"
|
# ? Sep 12, 2022 03:21 |
|
I really love Mediterranean architecture and design, and with this prompt (which I grabbed from Reddit), I can now create it whenever. " Modern Mediterranean luxury house interior, modern, realistic, cinematic lighting, finely detailed features, perfect art, trending on artstation, painted by greg rutkowski, makoto shinkai, takashi takeuchi, akihiko yoshida."
|
# ? Sep 12, 2022 03:28 |
|
Question for people who know something about how all this actually works, what does this mean? https://www.reddit.com/r/StableDiffusion/comments/xboy90/a_better_way_of_doing_img2img_by_finding_the/ They talk about finding the 'noise' that would have made SD make the original image, making precision customization possible right? I feel way too dumb
|
# ? Sep 12, 2022 04:41 |
|
TheWorldsaStage posted:Question for people who know something about how all this actually works, what does this mean? that sounds about right
|
# ? Sep 12, 2022 05:39 |
|
1950s sepia photograph of a nasa astronaut holding a duck and a grey alien. Not sure that qualifies as a "grey alien", but what I am sure is, this aliens *better*.
|
# ? Sep 12, 2022 08:10 |
|
leveraging img2img and SD upscaling to brew up some future hypercars edit: can't forget the history lesson. today we will learn about the lesser-known "monster truck crusade" VectorSigma fucked around with this message at 08:51 on Sep 12, 2022 |
# ? Sep 12, 2022 08:23 |
|
I just tried using stable diffusion to make some textures for some light graphic design work and I'm really impressed. Here's some examples for "leather texture, 3d render". In this context "3d render" appears to be a magic incantation that makes the results cleaner. "3d render" also causes some funny business like this to happen occasionally
|
# ? Sep 12, 2022 08:57 |
|
How do I get started on, I guess, text to text AI stuff? Like this https://twitter.com/wvschaik/status/1489720700898529280 I was talking with my wife about AI and kinda nerded out about the potential and she was like "well, can an AI write my daily report for me? Cuz this is a pain in the rear end" Obviously this isn't a serious thing and we're not actually gonna start having an AI do her job, but I thought "if it works like SD but just, for text, might actually work" So I gonna write a python script that just chops up chat logs into hour blocks and see if I can pass each block to an AI, tell the AI to summarize it, then stitch it back together into a timestamped daily report. But I don't know if text AI stuff even works like that.
|
# ? Sep 12, 2022 09:28 |
|
BrainDance posted:How do I get started on, I guess, text to text AI stuff? text AI has been capable of writting your wife's reports for years. you need to keep up
|
# ? Sep 12, 2022 09:32 |
|
BrainDance posted:How do I get started on, I guess, text to text AI stuff? Follow these instructions to get started: https://huggingface.co/docs/transformers/model_doc/gpt_neo edit: With 8GB of VRAM, I can also use the 2.7 billion parameter model by using half precision: Python code:
Wheany fucked around with this message at 10:41 on Sep 12, 2022 |
# ? Sep 12, 2022 10:01 |
|
Rahu posted:I just tried using stable diffusion to make some textures for some light graphic design work and I'm really impressed. Here's some examples for "leather texture, 3d render". In this context "3d render" appears to be a magic incantation that makes the results cleaner. If you have the newer voldy version I linked earlier in the thread, you can turn on tiling and it will generate infinite tiling versions
|
# ? Sep 12, 2022 10:30 |
|
|
# ? May 29, 2024 17:55 |
|
That sounds neat, I'll see if I can get that working. I'm a little limited on which port I can use right now because I'm on a M1 mac so I've mostly been sticking to the lstein fork.
|
# ? Sep 12, 2022 11:59 |