Howdy - just trying to stimulate some discussion here:

While still in its 'beta' (testing) stage, Google has developed a new AI platform called 'dreamix' which 'evolves' video clips (used as image 'seeds' + text prompts similar to existing AIs) using an advanced 'stable diffusion' (algorithm) GAN (generative adversarial network; basically a '2-in-1' neural network)...it's output (presumably) is a new (longer, hence 'evolved') video (time limit specified by the user).

Presumably also, the new video will not simply be 'scraped' from images (here: video clips) on the Internet and digitally morphed, merged/synthesized (like still artworks from DALL-E and others)...but actual video imagery that is 'new' (but the new video could be a digitally altered synthesis).

I wonder how long after its official release to the public, dreamix videos will show up here (with the 'seed' clip edited out)...?

Further, will it make custom videos obsolete?

What about existing and future wam producers with original content (AI vids could be 'seeded' with someone's original work)...It's like: Why bother, once it gets so real that anything is possible?

Fantasy wam + celebrity (of your choice)? No problem!

On a larger ethical note: the advent of this AI program will make 'deep fakes' easier to produce (imagine the dis-informational possibilities for budding fascists!). But for relevance, let's contain the discussion to WAM (of all kinds).

Maybe but this is still cool
https://youtu.be/cY-AyxJdzWA

Hello world of wam AI style
https://youtu.be/kDWwKtQ7N6c

While I don't understand the jargon being used or the technology side, as a consumer, I love innovation and new experiences. So, if it produces a new level of content that exceeds our expectations then I am excited.

This will be a boon for people who just need to see a female-shaped body get messy--which I'm not judging at all. But if you like seeing women, actual women in messy situations that actually happened, then AI WAM does nothing for you. I think this is going to expose a clear demarcation in people's preference and cause viewers to think really hard about what it is that makes WAM appealing to them.

Fortunately, I produce my own shit and don't really care about the custom market from either a buyer or seller's POV. Any pretty, charming ladies in the Chicagoland area who don't mind a little pie and a little slime will not be put out of work by this.

That's my thought too. It being a real girl getting real mess poured on her matters a ton.

Watching AI porn is like having sex with your waifu. It may be functionally the same but the human feeling just isn't there.

For people who like realism ai isn't able to eliminate the "uncanny valley" effect. It's like having cgi Moff Tarkin in Rogue One or even the cgi deaging of Luke Skywalker there's something in the eyes and facial tiks that ai can't replicate. At least not for a good long while. I think this will have a niche but I don't see it as a industry killer. The biggest threat to wam industry is still and will be piracy of content

Yeah, a lot of the same caveats apply here as apply with any other machine-learning stuff, namely:

1. As TheSpecialist points out, for the foreseeable future there will continue to be people who want the real thing and are turned off by artifice.

(For another example in this category, think back to the mental health site that experimented with ChatGPT: "users were not initially informed the responses were developed by a bot, and 'once people learned the messages were co-created by a machine, it didn't work'"; source: https://www.businessinsider.com/company-using-chatgpt-mental-health-support-ethical-issues-2023-1)

2. For at least a little while longer, machine-generated stuff is going to be unrealistic. That's why we have a running joke about how AIs can't draw hands. So, again, people are going to avoid first-generation machine-generated WAM if they get turned off by, like, bad physics or inconsistent anatomy or something.

3. By default, computers are incredibly stupid, so it's actually kinda hard to get a machine to do what you want. For example, one person tried to use ChatGPT to write a fantasy novel and eventually just gave up. They did manage to wrangle it into writing a detailed outline, but even then it took multiple days of hard work (source: https://medium.com/geekculture/write-a-novel-using-ai-openai-chatgpt-c02be59cd937). This excludes all the potential users who are too lazy and/or tech-averse to bash their heads against the machine until the machine produces the thing they want.

4. Even if you aren't ruled out by 1-3, you still have to be able to give the thing a seed file that resembles the output you want. So every time someone wants some new niche thing to plug into their machine, they're likely to have to reach out to a traditional producer one way or the other.

.
.
.
Buuuuuuuut still, with all that said, the potential is certainly there. In theory, 1-4 can all be fixed. It's just a question of whether the stars align (for instance, whether machine-generated video can become cheap enough to allow for free usage by the masses, whether people decide to make it realistic enough to satisfy people who want realism, whether we bother to teach it conversational skills so that it's usable by laypeople, and so on). If the stars do align, then machine-generated video is going to become a part of life in every area, up to and including this one. If not, then we could see anything from AI becoming captured by industrial interests to AI becoming a quirky footnote in history books.

It does creep me out rather a lot. What if this technology gets so good you can't differentiate between it and real video? People could be framed for all sorts of crimes, or maybe even worse, they could get away with crimes by arguing that you can't prove any images or video of them are real. I really think it's going to need close and careful regulation, and I'm not sure I trust Google and other big tech companies to use it responsibly.

Despite my reservations, I'll still no doubt use it for my own WAM gratification, so I'm not helping.

At the moment, it's not a threat, as yet AI can't even draw WAM still images to genuinely realistic standards never mind full motion video. However, given the enormous level of investment being poured into AI at the moment, and the pace with which technology advances, these issues will be overcome.

The "uncanny valley" effect will be much harder to resolve, even once it is possible to have AI draw people with non-weird eyes, the correct number of fingers per hand, and without entire second sets of teeth inside their mouths. We are very good at spotting things that aren't-quite-right, and that'll ultimately be the dividing line between AI erotica and human modelled erotica. And I imagine that those of us who've always known only real human erotica will mostly always prefer it. But - the generations who grow up in an environment where AI erotica is commonplace may well feel differently to how my generation will see things.

I think it'll be a bit like the MP3 revolution. Before MP3s, the aim in hi-fi was ever higher fidelity and more accurate reproduction, and people would spend wild amounts of money in the pursuit of it. I know someone who had a hi-fi rig that, had he bought the components new, would have cost £30k. Which was the price of a house in some parts of the country at the time. Now he didn't spend that, be bought the kit second hand as others upgraded theirs, so he spent about 7k. I got to listen to the Johnny Cash Live At San Quentin vinyl album on it and that was almost a religious experience. However, then MP3 happened - and given the limited size of the hard drives of 20-odd years ago, 128bit MP3s at that. Hi-fi they weren't, but they took the entire music business by storm and within a couple of years they were pretty much what everyone was listening to. The convenience, and stability, of tiny music players with no moving parts that you could even wear while running, completely eclipsed arguments about quality, for everyone bar dedicated audiophiles.

At the end of the day the purpose of a WAM video is to stimulate the mind to the point of orgasm for those who are turned on by WAM. For some, the model's reactions are everything. For others it's the way their clothes change as the mess is applied. For yet others it's all about just seeing the genitals get messy. The detail variations are endless. And at least some of them will be catered to by AI erotica, and at least some of the audience will be perfectly happy with uncanny valley effects as long as it results in a successful wank.

I suspect those of us who want to go on producing modelled erotica will need to ensure we really emphasise the human side in our productions. And recognise that the market will almost certainly change significantly in the future as AI production houses start to appear.

We're stage-coach operators at the dawn of the steam age. Right now, that complicated and elaborate contraption up at the mine pulling its string of coal trucks at walking pace is no threat to our fleet of fine, well-sprung, upholstered carriages and swift coach horses. But anyone with vision can see that within 20 years steam trains will be linking towns together at 40 miles an hour, and the mass-market horse drawn carriage business will be well on its way to becoming history.

Related to what others have said, I think the emotional aspect will be harder for AI to replicate. However, the importance of this will depend on the video.

E.g. suppose that the script for a WAM video is basically "You sit in that chair, I'll ask you a question, then you say "I don't know" and I'll pour a bucket of gunge over you". That would be fairly straightforward for AI.

By contrast, think of gameshow WAM, where someone is asked a question and you can see that the answer is on the tip of their tongue. "Oo, wait, I know this, hang on, it's ... um ... dang it!" (with accompanying hand gestures). In that scenario, you either need a model who's a good actor, or it needs to be their genuine personality coming through. It might eventually be possible with AI, but not anytime soon.

There have been previous discussions where some people say that they fast forward through all the build-up (skipping directly to the mess) and watch videos on mute. I'm guessing that they'll be the initial target audience for AI WAM.

Note to self: blow away Miles Pieson

Btw for those who may be unaware, we have a UMD group purely for AI WAM:
https://umd.net/groups/group/ai-wam

The stuff that 3D Slimer and RdBd do can be used as templates for these new type of videos, couldn't they?

Yes, they could.

As for realism/naturalism v. AI's unnatural errors (fingers, teeth, etc.)...I would point you to a recent 60 Minutes segment featuring a 'deep fake' video of Tom Cruise. An actor was hired to do a series of actions and TC's face was digitally mapped onto the (non TC) actor's face (and used a voice sampling program for TC's voice). The programmer was using an advanced CGI program that he and his team created. The resulting video was astoundingly believable (if you didn't know how it was made, you wouldn't t know it was not TC).

My point being: how long before this sophisticated type of program is reduced to a few algorithms and then integrated with a next-gen adversarial neural network? I can imagine a customer being shown a roster of famous celebrities and then presented with a 'palette' of situations, actions, stunts, scenarios, etc. to choose from (with the usual parameter settings plus some new ones)...you pay your credit amount, and in a minute or two...voila!

Oh, sure. Deepfakes have been pretty believable for at least a few years now. But, like, look at what it took to generate this one example:

(1) a live actor doing all the hard work
(2) samples of Tom Cruise's voice
(3) samples of Tom Cruise's face
(4) a cutting-edge (expensive) piece of software built by a team of highly specialized (expensive) tech people for this precise purpose (as opposed to the, like, 5-6 different things that a WAM generator would have to do)

That's a pretty far cry from the type of custom-AI-content-on-demand that you're talking about. So, like, if you're literally asking how long it's gonna take until we get there? I have no idea, and I would be extremely skeptical of anyone who claims to know. On the other hand, if your question is just supposed to mean "surely we'll get there some day," then... maybe? It just depends. It might happen within five years or it might not happen within your lifetime.

I think in the medium term its probably not going to happen for a number of reasons.

Personally for me, the knowledge the mess actually happened is important. It won't be true for everyone but that is the case for me. You aren't going to be able to ask an AI how something felt or what they are looking forward to or dreading.

I also think individual motives for creating content are important. A lot of people seem to be creating the content they want to see and I doubt many will want to change medium for risking their house style, so even in if it does become commonplace, live action videos aren't going to dry up overnight.

I think mainstream porn will embrace the technology quite quickly. The problem is that realistic fluid dynamics is actually quite hard to replicate (computer gaming being the main example). I think we're a long way off of it being close to realistic and cheaper to produce using AI than mixing up some mess and getting a willing model in to do a scene.

Time will tell, but one thing is for sure - we're going to keep making real messes for a long time to come.

Why even mention it then? This is the most serious thing to be concerned about with this tech. How sad is it that the second something new comes out the first thing we think of is how the government will misuse/abuse it (or most likely already have been for years).

There are other moral implications too.


Just look at some of the deepfake porn discourse that's been going on lately with Tubers.

Remember that person on the show we can't talk about that got slimed and people tracked them down on social media and perved on them about it? Well, now we can just make our own deep fake of it, where it's the exact same scenario, but they're naked and fucking an octopus or something.

Aside from playing with filters of my own pictures and friends, I'm not a big fan of AI. It creates assets out of other people's art and styles.

That said, I don't really care if other people really like it or think it's the best, that's fine.
I guess enjoy it while you can before things like digital ID come out?

Why mention it? To serve as a thought experiment (hence this thread). Plus, we humans must asked ethical questions now to inform (if not influence) the purveyors of advanced AI tech.

Many good/valid points have been made here. I do not know who will turn out to be 'most right'.

But I will note that with MS and Google now in an "AI arms race" (media term, not mine; it seems to imply the use of AI as 'weapons') and spending billions on this tech...it will advance rapidly (as it already is). The Tom Cruise example mentioned earlier, for example, will without doubt be reduced to (or implemented through) an additional 'hidden layer' of nodes (each composed of a number of weighted advanced algorithms) ...and add on some extra few months for tweaking and refinement ...and you will see deep fake video platforms (cloud-based or as downloadable apps) within (I'm predicting) 2 years (or less).

As for fluid dynamics (the computer modeling of)...completely realistic modeling of FD has been a goal of CGI programmers for decades...and once quantum computation achieves real 'supremacy' (sufficient qubits [49 +] maintaining coherence indefinitely) -- also a goal of Google and MS -- then FD will become trivial...the future will have arrived, and AI tech will be unstoppable (the 'singularity'???).

Enjoy your day.

Sooooooooo

1. Microsoft's chatbot is a flaming trainwreck right now (https://www.theverge.com/2023/2/16/23602965/microsoft-bing-ai-sydney-fury-furry-venom). Insofar as these types of programs have advanced at all, they've mostly done so in terms of, like, tone and grammar and syntax. If you want logic or factual accuracy (or, obviously, authentic understanding or originality or ethics), you're still out of luck. More on this later.

2. We have deepfake apps already. You can just google "deepfake app" and find websites where you can do deepfakes for free. Again, though, deepfaked video (or even deepfaked video + audio) is a different beast than custom AI-driven media. It's the difference between repainting a car and building a new car. More on that later, too. For now, hey - if you want to test out face-swapped WAM, I guess you now have the option to try that. Let us know how it goes.

3. Fluid dynamics aren't just a matter of raw computing power. (Very few worthwhile software challenges are.) It's also a matter of modeling the fluid in question correctly in code, which can be very difficult. Interestingly, a digital artist talks about this a little in the Frozen 2 documentary on Disney+.

4. As traditionally defined, the singularity is definitely not just "machine learning gets very good at realistic audio/visuals." The idea behind the singularity is that computers will eventually develop some real degree of agency and then somehow use that agency to make humans obsolete or irrelevant or whatever.

But what we ignore when we talk about "the singularity" is our own role in the process. Let's go back to ChatGPT. Why would anyone be worried about a system like ChatGPT? Well, because very large companies are pushing these systems on us, some media outlets are hyping them up, and in general our society already pushes people away from normal social relationships and into weird, one-sided quasi-relationships with brands and followers and shit. So even if ChatGPT has a lot of influence and high usage rates, that doesn't necessarily mean that ChatGPT is, like, good. It might just mean that our leaders (nonpolitical, political, or both) are willing to overlook ChatGPT's flaws because of its potential to consolidate their existing power.

This brings us to fully customizable machine-generated movies (WAM or otherwise). That's only going to exist if someone wants to build it, and someone is only going to want to build it if it can do something that lesser technologies can't do. For example, search engines are interested in chatbots because chatbots have the potential to stop users from clicking on search results (thereby giving ad revenue to publishers) and keep them on the search page itself (thereby redirecting that same ad revenue to the search engine). As such, those companies have little or no interest in developing a chatbot that can write a decent novel or provide a meaningful policy analysis or even just tell the truth. Someone might want a bot that can do some of those things, but Bing won't, and so if we rely on Bing to get that type of bot then we'll never get one.

The same applies to audio and video. Currently, deepfakes are sufficient for most purposes, good or bad. If people eventually decide that they want more, they'll work towards getting more. But that's not "AI becoming unstoppable." Anything that we do, we can stop doing. And even if people do choose to pursue custom AI movies, that'll still be a human decision, meaning that someone (likely someone with power) will have to believe that the benefit is worth the cost. In other words, none of this is inevitable in any sense that matters, least of all within a microsystem like the UMD.

AI - machine learning (ANNs, GANs) are being used to discover how proteins fold in our cells, discovering polygenic causes for diseases, diagnosing cancer with 95% accuracy...Fluid dynamic modeling with be achieved through some type of GAN -- not by humans (that is the point of this thread)...and similar DL networks are actually writing code...once someone (or some AI) puts two of these together to 'compete' against each other to best model, say, actual video (microscopic or macro) of fluid flow (dynamics), it will be game over (for that challenge anyway)....then this will be applied to other 'challenges'...then, it will become either commercially available or common place on the Web, and likely free, as the 'arms race' will drive the price down (as long as there is competition).

So, Ok, I will conceded that I may have over-stated the arrival of the singularity, slightly (note: actually, the singularity is defined as 'the point in which a super-intelligent machine creates another super-intelligent machine, and so on...' I. J. Good)...so, we now have robots building robots, and AI's writing their own code...Truly, we are already more than half-way there.

The future is nearer that you think.

Have a nice day.

Yeah, I mean, those ideas make a lot of sense in the abstract, and one day they may come true. Right now, though, we're still at the phase where Tesla's "self-driving" cars plow into children in broad daylight and Microsoft's chatbot lies about what year it is. Personally, I'm not sure that we're halfway to machines having *regular* intelligence, let alone whatever "super-intelligence" would even mean.

But, again, the tools are out there if this is something that you personally want to work on. I really would be interested in seeing what you get when you use a publicly available deepfake tool to try to make custom WAM, for example. Or if you have an interest in plugging machine learning algorithms into each other so they can team up, it'd be fun to see how smoothly that actually works in practice.

And yeah - as for the rest of it, I'll believe it when I see it

"Tesla's "self-driving" cars plow into children in broad daylight..."

yikes...a cautionary example of why not to put too much faith in the AI hype, for sure...

It should be noted:

Messmaster himself started a thread (not on this Messy forum) on AI-generated 3D movies (prior to my current post)...in which he presciently speculates that wam producers "will become redundant" (or the like). Here is the link to the thread:

https://umd.net/groups/i/whole-movies

AND, DungeonMaster also noted the advancement in AI tech (in the context of wam videos). Link here:

https://umd.net/forums/ai-generated-messy-content

FYI

Sounds to me like your mind is pretty made up on the subject, but even if all the predictions about advancements in technology and the timescale are correct, I don't buy the headline that "The end of WAM videos as we know them coming soon!".

I will still be turned on by my partner getting messy, she will still get turned on by doing it and sharing it, and I'm fairly certain others will still get turned on by watching it, so why would that cycle stop even if the market is flooded with AI videos? We're not money motivated.

Also this community seems to know exactly what they like. There seems to be fairly regular occurance of posts lamenting the "good old days" and posts of people selling their VHS collection with lots of interest. While for some people, AI may give a chance for videos similar to those made 20 years ago to be made again, for others "now" will become their "good old days" which I think will keep live WAM videos going while people want to make them.

In any case, I have no intention of fearing the technology of the future because of what might happen. There is enough scary stuff already happening in the world without fearing what might be. I think it's very easy to just give up now by thinking that way, so I'm not going to, we're going to keep having our fun and doing what we do.

Platypus530:

No, my mind is not "made up" ...I believe that wam fans/producers will continue to make 'old school' wam media; over time, however, there will be more and more AI-based wam content (and it will replace a large share of the commercial wam pie, so to speak)>

While I suppose that any speculation/prognostication regarding new tech's impact on 'old tech' (or the products thereof) COULD be viewed as 'fear-mongering'....that was not my intent (truly).

"I have no intention of fearing the technology of the future because of what might happen"

Good on you! I do not fear it either....but 'forewarned is forearmed ' as they say.

(the use of the word 'might' is tricky, insofar as much of the AI tech described in this thread is already happening and it doesn't take a Nostradamus to see where the trend is heading.

The only limiting factor (to wide-spread adoption) that I can see now is the increase in processing power that these AI video platforms (the making of said products) will require.

cheers!

I say might because the technology hasn't brought about "the end of WAM videos as we know them" just yet, and I don't think it ever will. AI WAM might form a large market share in the future, but I still don't see it ever surpassing live action.

Cost (of processing and no doubt licesing) might be the only limiting factor but I think that's significant. I think motive is also a factor, most people will find their way to producing WAM because they enjoy doing it and start filming what they love doing.

We've known about and been investing heavily in nuclear fusion for decades but I haven't got a fusion powered car just yet. So we'll see where the tech goes and if it ever becomes mass market.

I have many questions: Why do we need it? What is the purpose of using it? How can we benefit from it? I haven't heard any intellectually sound reasoning for creating this technology other than for people who have absolutely no talent to have a computer create something for them. Instead of becoming a good artist, you just tell the computer to create it for you. Instead of becoming a good writer or simply being literate, you can just ask a computer to do it for you. Instead of being a good producer/director/cinematographer, you can just ask a computer to make it for you. So why is what is created so special? So a computer can do something with precision, why does that mean anything special? If a person does something, it's because they actually are talented in it and overcame challenges and actually did work to build that talent. But a computer can do exactly what it's programmed to do, so it's not doing anything beyond what it's being told to do, whoopty shit. It's just not anything to care about. Computers are designed to be precise. Humans are born with limitations and challenges. Overcoming those to create something that no one else can create is far more interesting and spectacular. The act of story telling is far more compelling from a human that is able to express emotion and empathize with and sympathize with. How do I care about what a machine was told to do? The person telling the story lived the experience or dreamed it from their own life experiences, so why would anything a machine was told to do be more important?

So here's my biggest argument on this technology and why this should be banned from all video platforms. The violation of consent. Already there are people pointing out how you can use this technology to insert their favorite celebrity into any porn that they want to create. So how does the celebrity feel about having their identity used in this way? It's one thing to fantasize about someone, that's entirely harmless because you're not invading anyone's consent to do it. Your mind is your own. But to create a realistic image of that, it gets really obsessively creepy and downright invasive. If someone doesn't like porn, is deeply religious, and does not consent to making those images of themselves, it is SO COMPLETELY violating to find out that those images could even exist. And if someone is inconsiderate enough to create an image of someone without their consent, what's to stop them from selling it and making money off of it? It's exploitation, it's violation of consent, it's identity theft. I mean, there are so many violating experiences from this kind of technology. And just imagine the level of catfishing that will happen from this. Imagine thinking you're setting up for a date with the most attractive person only to find out that you've been led on by kids making a joke out of you, or a serial killer who will murder the fuck out of you. There are so many levels of fucked up bad that this will lead to because humans can't be trusted. Not in any way, not to be respectful, and especially not to be intellectually responsible with the implications of what these things are capable of doing.

Gotta co-sign what CreamMeAgain said. Sums up every thought I have on this in all of its nuance.

For me personally, this won't change anything. At least not for an unforeseeable timeline anyway, because uncanny valley is such a huge thing for me. When I see it, the whole thing is over. The level of detail needed to get over it, even becoming on pair with later content done at 480p just isn't there. At least not for this genre, and by the time it gets to that point I'm sure it will be censored out like I'm told Chat GPT was. I think it was here that I read it couldn't even come up with an idea where a single pie gets used because it's somehow mean.

It won't change things at home either. If I'm making a mess, cameras are rolling.