Firstly, I am not taking away from some producers work and I know that there will be folks here that love AI pictures, that being said...

Personally I don't find AI pictures at all sexy. Some of the pictures are practically perfect and only the weird 'sheen' around it or odd focussing effect/colours (?) makes it obvious that its been generated. But the moment I notice it's AI/generated/photoshopped then all sexual aspects of the picture vanish, even if I had an initial flush before realising.

For me, if it didn't happen then it's not a turn on. It has to be real...which is strange as I can imagine plenty of scenarios that didn't happen that are still arousing, and AI is just the physical manifestation of imagination.

Who knows what the future will bring, maybe there will be a porn video so good it's arousing even though it's 100% computer generated. But until then I will have to hope that models will want to do this daft kink we all have.

Open to discussion, just a personal ramble.

I agree with a lot of what you said here about AI and pictures. As long as it is clearly labeled AI generated, I can appreciate it for what it is. Where I think the line needs to be drawn is consent. If the person in the image did not give their consent for their likeness to be altered I think that it should be treated similarly to revenge porn. While I know the discussion already got axed about a celebrity in this situation because of the background, I would like to see the celebrity go after the person that used the image of them without their consent depicting something that would fit in here that never happened.

I read an article recently that Stephen Fry did about AI and his impressions. Someone without his knowledge took all the books in the Harry Potter series that he narrated and used it to train AI and then a speech was generated that sounded like him. This is not a good situation.

The attitude of studio owners in Hollywood around this is they call someone in for an 8 hour shoot. Get enough data to train the AI and then their AI data can be put on file and used as they see fit forever with no reimbursement. Think of all the athletes and others that have sponsorship deals and other situations where that entity owns the rights to their likeness. Technically since AI is a likeness of the person, the company they have the sponsorship with can use the AI to do whatever they want. Even now on social media the families of dead celebrities are having to fight with entities using AI to claim they said something they never did. How do you copyright something that is who you are? Then how do you go after the people creating the AI content and prove it is not you. Sure now in its infancy we can tell the difference. Give it 5, 10 years will we be able to tell the difference? Will we be able to still detect the talent of the individual or will AI have mastered their nuances as well?

While I can appreciate some of the learning that can be done through AI, I think that images, spoken word, mannerisms etc should be off limits.

Completely agree with you. In fact, I was about to start a similar thread, having seen some recent posts on the forum in this vein!

For me, the AI pictures aren't true WAM: they're showing a concept, rather than an experience. And for me, in viewing WAM, I really want to see the lengths that someone's gone to in being on the receiving end. 'Generic picture of woman covered in custard', produced by AI, doesn't have the same effect. To me, it's no different from the Photoshopped images of a 'celebrity' taking a pie in the face. It didn't happen, and (based on what I've seen so far) is an obvious and artificial generation.

For me, good WAM is all about the ingenuity that someone's gone to in producing a scene, the level and complexity of the mess, the predicament in which they find themselves - and the certain knowledge that they're being completely covered in food/paint/mess

And when I'm doing a salon session, of course there's no room for AI!

I don't perv on Cortana

As I've said elsthread before, I think we need to have a hard rule that here, on UMD, no-one can post generated images, regardless of the method or tech used, featuring the likenesses of real-world WAM models, without those models' full informed consent, model releases, etc.

I believe we need to protect our own community as an absolete core requirement, and so this is a rule we should establish now, before it becomes a problem.

AI WAM is certainly a fascinating field, some of what's been posted to the AI group is astoundingly good. But from a turn-on POV, I wonder if it's more fun to generate images than to look at generated ones. Plus of course so far it's just single still images, as yet no-one had produced a convincing sequence of say a model wading into a pool or river, getting a bit deeper each image, never mind actual full-motion video. That probably is coming but not yet.

In some ways it's like going back to the days of taping mainstream WAM off the TV, where you'd have to freeze-frame the video at exactly the right moment and then imagine just how the person was feeling as they got engulfed, knowing you couldn't restart the tape because a few frames later it cuts away to the male show host or an ad break or something completely unrelated.

I think once completely convincing full-motion video is available, it'll be possible to create scenes as sexy as human modelled ones - but there will still be that "it didn't really happen" effect. I may be biased but I do think human-modelled, "yes we actually did this for real" content will always have a place.

I'm not totally against it, some of the WAM ones aren't bad at all, but when you realize they're AI images, that kind of takes away from it for me. Nothing will ever beat the real thing.

Outside of WAM, I've seen some hilarious random things generated like Matt Damon overseeing the birth of baby Shreks.

The sucky thing about AI is that it can take away from creators who specialize in an art. As a big supporter of artists, and art, that cheapens it a little bit for me. I'm always going to go to artists for PHYSICAL art created by a human's hands, but there are people out there who aren't willing to pay the price and wouldn't mind going the AI route to save some money, and that sucks for artists...but if I'm being honest, artists who are established and charge way more money than "unknown" average joe artists could probably stand to lose a little money.

I generate images, and I also suffer from the "didn't happen" affect. The experience needs to be real for me because I have a strong desire to experience what the model is experiencing. I can generate thousands of images, but at the end of the day I am watching real videos and photography.

I sort of don't like seeing AI images on the main messy forum or under People photos. I have concerns about AI images. I was worried that it might cause confusion or disappointment when viewers find that the image was fake.

AI images has potential for the social aspect. Here are some of my personal experiences with it:

1. It gave me a reason to post and something to contribute. I have seen this a few times where someone makes their first post to share AI images.

2. I've had several private conversations where AI images made it so much more fun. I've shared ideas I've had and created images of ideas others have had. This has given me confidence.

3. I made an image to describe an idea I had to achieve a specific look in a custom scene. I have also seen AI images used for the first time in the customs group.

It seems like a fun way to share interests. It can also feel like an extension of the imagination and be used very creatively. It can be used to create fantasy images which are not possible.

It is fun to share the images. I enjoy getting and giving likes and comments. I really like the idea of giving ideas. I would really hope that the images are inspiring, and not the opposite where people see it and are discouraged.

It is going to be interesting how this turns out. It could have potential, or it could make everyone sick of seeing all of the images. It could be good enough for people to try and sell, or bad enough for people to want to ban it.

I like AI images for the social aspect, but I think it is disappointing. I really want some real experiences myself hehe.

I think they can go on the Main page, but should be labelled as AI in the title...and to be fair, I think nearly all AI images that have been posted here have been labelled.

AI has a place, it can be used to plan a 'real' session, see if folks like it before spending loads of money and it failing.

Another thing, as long as the 'person' in the picture is not taken from someone in real life (deepfaking), then it is ethical porn! No fear of coercion, trafficking etc.

Update: have just discovered we do have rules preventing the posting of fake images based on real-world people, regardless of the method used, unless full consent has been acquired, well done MM.

https://umd.net/termsofservice/#deep_fakes

Hmm, I am not tech-savvy. I have just found out that AI generated art uses a lot of fresh water (for cooling I guess), hardware and tech...so yeah, clean water and paid producers/models or bad porn. Difficult choice.

I appreciate it as a form of art and creativity, but this is distinct from photos of actual scenes where you know the lucky people really feeling real sensations of real gunk on real skin.

I would love if sites like this had a tick box "don't show me AI images" option. I view most AI generated material as empty, low quality content that just clutters things up and obscures the genuine content.

That's coming as soon as I can find time to program it in.

I really don't know what I think about it.

Some of the muddy images I see look pretty nice but I can tell it's AI.

I've got actual drawings that I've done on my profile. I have not had any real interest in trying to create similar material with AI.

I am probably too much of a control freak to let a computer take over. Every stroke of the pencil and paint brush are under my direct control and I prefer my work to stand or fail on its own merit.

If I get so old that I have arthritis in all 10 fingers? Maybe then!

There is a lot that can go wrong with AI. It's absolutely head-spinning to imagine how this could go bad in the hands of the wrong people. I agree that we need to consider carefully what we do with it on forums like this one.

Nessie

If you think images have been created that way, use the report link and flag them, as that's completely forbidden in the ToS, because of issues of consent, and legality.

Deep Fakes and Artificial Intelligence
If you upload content that is photorealistic but actually generated or fake, that fact must be mentioned in an accompanying description or caption (a formal tagging and filtering system will replace this requirement later). The actual technology used to generate the content does not matter, whether it's AI, Photoshop, or a really skilled human artist.

If you upload content that features the likeness of real person but has been generated or altered to depict them in a fantasy scenario, that person must be you, your model, or someone from whom you've gotten consent to use their likeness for this purpose. The fact that it's a deep fake must be mentioned in an accompanying description or caption. We will need to remove any content that we feel may be using a person's likeness without their consent.

Photorealistic generated and altered content is subject to the same rules as natural content: Copyright laws still apply, subjects must be clearly portrayed as adults, and if the content depicts any nudity, the uploader must have a verified account.

Our rules about deep fakes will evolve as our community continues to discuss the morals and consequences of using this technology.

Note the "likeness of a real real person" rules apply to stock imagery and catalog models as much as they do to named individuals.

That's coming. Uploaders will be able to mark any content with something like an AI tag (just like the gender and explicit tags) and then visitors will be able to filter it out of their sitewide results.

But I'm working on other stuff along with that. We need to be able to mark more content as explicit so we can properly filter that against your prefs (e.g. forum posts and pics, blogs, ads, etc). Might even have explicitness levels. Anyway when I do that, I'll also add AI tagging to everything.

The update will also include the new sex vs gender system I've been talking about, meaning all content can be tagged with those things separately, and then your filter prefs can work with that.

I feel that all content should be able to be tagged and filtered by all these criteria, so it's more efficient to knock them all out together.

When this is launched, it'll all go into a revamp of the sorting / filtering preferences layout. Instead of sprawling bubbles and checkboxes and links, they'll go inside of pulldowns. Mobile users will see a button for a new filter menu next to the hamburger button.

So that's why AI tagging is taking so long: It'll be part of an update of the whole content preferences / filtering system and its responsive layout. I expect to get it done by the end of October, hopefully. Then I can go back and retroactively tag all the existing AI stuff, and all will be good in the neighborhood

There are 2 ways it can work:

1. A new image is created using Generative AI.
2. An existing photo is altered with Generative AI (or photoshop or anything).

Some of us are using Stability AI's Stable Diffusion. It is a model that is trained on billions of photos that have text captions which makes it possible to generate new images using a text prompt. There is open source software and online tools used to generate images. It is also possible to take photos and alter them.

It is possible to:

* Alter a stock photo or any photo of a real person.
* Generate images which resemble real people if the model is trained on that person's photo and name.
* Train models on the photos and name of a real person to generate images of them.

A model is a file. Most people do not use the base model and will use checkpoint models and LoRA models. These are fine tuned models. Some are fine tuned to produce art or anime, while others are fine tuned to produce photo realistic images. The checkpoints and LoRAs are privately trained. This is the biggest problem because it would be very difficult to tell how the model was trained. If it was trained on thousands of stock photos of a single person, the generated results will be biased and likely look like that 1 person.

I usually prompt for "woman, brunette, naturally beautiful face, soft smile". This should generate a random brunette that is not meant to resemble any one person. I do not take stock photos and edit them. I think they resemble stock photos because of how bland their faces look, with eyes wide open, a generic facial expression, and very little facial features.

There are a lot of concerns about Generative AI with copyright and IP rights. Getty Images is suing Stability AI. I don't think the images posted here violate copyright law, but I'm no expert. There was one recently deleted post which had some altered images of some altered real photos. There are some CC licensed photos which allow derived images if there is attribution, but this still violates the ToS because it requires consent from the person in the photo. I guess people have to trust that the person submitting the images are following the rules and honoring consent, and report them if not.

Hopefully this addresses your concerns about these images and their resemblance to stock photos

Actors are literally striking for this reason. The writers just ended theirs with (justifiably so) better control over what they can determine is misuse of AI in the industry. My guess are actors are demanding the same in addition to residuals. Again, justifiably so.

https://www.cnn.com/videos/business/2023/10/04/exp-sag-aftra-strike-artificial-intelligence-gavin-mueller-intv-100402aseg2-cnni-business.cnn

There is a new preference that I'm working on that should filter synthetic content out. Can you try that?

I can totally understand. I can imagine it feels like spam now since it is getting much easier to producer better images. A lot of people are sharing images which they have generated because technology is improving.

There is a new "synthetic" tag and there is an option under your profile preferences to block. This was just added . I think it might be ready to try out now. I feel assured that this will be addressed well and not become a problem.

I can also understand why people are sharing the images that you are seeing. They want to share ideas and form connections. The most recent example at the time I am posting this is https://umd.net/forums/how-dare-you--ai-made-concept-art-clean where it has nearly 2K views and 22 replies. I love how the first reply mentions Penelope and Zoey. Imagine how cool this would be if it inspired real content. It is also fun to imagine that this is helping people form friendships or relationships.

Thanks - that seems to work well. I can still see some older posts, e.g.
https://umd.net/forums/saturdai-night-pie-pictures
However, I assume that they'll gradually get bumped off the front page, so it's not worth the hassle of retrospectively tagging them.

Thanks for the feedback I've been manually updating older content on the wall / galleries / pic archive, / etc. I won't get them all as my priority is getting all the forms working so new stuff can be tagged and filtered properly. With everybody's help, all this AI content hopefully shouldn't be an issue.

Does the filter work on blogs and stories?

Yes, but still working on that. Eventually I plan to have it affect all content and associated pictures.