Loab, the AI Ghost in the Machine

[BEEPING] This week's episode is brought to you by Nate and Sob. Thank you, Nate and Sob. It's with the generous support of listeners just like you that we here at the whole rabbit get to keep launching our swarm of semi-sensitient nanoparticles directly into the Pacific Jet Stream. Where are these imperceptively small robots infiltrate the bloodstream of every living creature and begin analyzing it for its monetary value so we can make better and more informed negotiations with the Anunnaki overlords when they eventually arrive with the motherships to process and sell everything on this surface to pay off the debt to space bank. I hear they've got quite the tab. If you'd like to help assist the human race in extra galactic negotiations, or just here the second half of this episode, then please visit www.patreon.com/theholerabbit or your monthly donation of just five bucks that not only gets you the extended version of this episode, but our entire library of episodes. Going back before the turn of the decade, but that's not all. I'll send you a five-by-five-quality vinyl sticker for our big bunny baffinette cover art alongside the secret keys to our Discord server where we coordinate our ongoing investigation into the spiritual powers known of various cheeses and different firmness. If you enjoy the free show, but would rather save your five bucks for a double-sized can of corn-beath hash, then please consider giving the show a like, subscribe, or share on your social media platform of choice. On the flip side, we do have a $20 tier where you can both support the show and on top of all the other benefits, get a monthly tarot reading from yours truly. Every month, they do more and more and these so get them while they're hot. On this week's show, we're joined once again by Mr. Tim Hacker, the creator and host of the Crypto Chronicles podcast, Crypto Chronicles.com, who promises to take us on a journey to the darkest places within the forbidden nether regions of the publicly available, artificially intelligent minds used to generate images from text there, within the murky waters of the collective digital unconscious is said to their ghosts, demons, and even cryptids of an order of life and consciousness we scarcely have begun to understand from this new class of digital supernatural hauntings comes the bloody red face of low-ab, a ghost, so prolific, and haunted that she took the world by storm in spring of 2022. In the free show, we discussed the hideous low-ab, how she was discovered, and some frightening implications regarding the danger of artificially intelligent ghosts in the extended version of the show. We explore quite a bit deeper to find other AI birth cryptids before settling in on what traditional demons correspond with low-ab in magic and mythology. Thank you and enjoy the show. (upbeat music) (upbeat music) - Hey guys, if Apple made a car, would it have windows? - Uh, no. - It could run with windows? - No, it would just have a system crash. - Oh! - What do you call a computer mouse that swears a lot? - What? - A cursor. What's an AI's favorite music? - What? - Draftpunk. - Algorithms. - Lame. - No, that's funny. - Maybe a little bit funny, but mostly lame to me. Draftpunk's better. They made a newly artificially intelligent Oreo. It's one smart cookie. (laughing) Hello everybody and welcome to the whole rabbit. When we don't just tabulate the immense amount of data being sent to us by a deeply embedded spy station inside the secret base on the moon, we did a deep system scan of the entire internet and discovered a new form of supernatural life native to neural networks of artificially intelligent image generation software like Dal-E and crayon, called the dark hag person from the dark web, a real spooky digital monster. At least we think so. I'm your host, Hackrabbit. Also known as Luke Madrid. I'm joined this week by the matriarchal mood woofer, Mari Sama. (humming) The sacrificial goat of the gods, Malachor V. - What's up? - And the delightful daughter of the doot, Haka Astra. - Arir. - And this week we're joined once again by the deliciously dapper, creator of the cryptic chronicles podcast, Mr. Tim Hacker. - Crypticronicles.com. - Hello, happy to be here. And I assure you, I'm really here talking. This is not AI, I, I, I talking right now. - I. - Exactly what AI would tell you. - Cryptic is the one who researched and wrote 80 to 90 to 99% of this episode and is the master and commander today. So thank you for cutting us some slack and coming back on the show, Tim. - Always, I'm really interested in talking about the possibility of AI creating and occult entity essentially. - Okay, everyone. So I have a bit of a confession to make. Last week I was having a tough time personally and I called off recording the show, right? And I was sitting there. I was like, "Hey, I don't think just sitting here is a great idea. I should work on the show." But by then, you know, you guys who are gone and except me, I didn't leave. - Correct. So I was like, "Okay, what do we do?" And Hek is like, "Why don't you have an AI write the episode like the last podcast on the left did?" 'Cause apparently some fans sent in an AI created episode. And I was like, "I don't know how to do that." So I did some research, couldn't figure it out. And so I just started making a funny episode and I told everyone that it was AI. (whooshing) The problem was, is that I kind of did it so like halfway through the episode, I was hoping everyone would like realize it was bogus but it didn't really work out that way. So I kind of just let it write itself out for a week and now I'm confessing. So it's okay. The AI's not gonna take over the universe, at least yet. - Well, I did actually have the conversation with chat, EPT and AI was utilized to do the intro voice of Mari and Mal and to read certain things. - Yes, like the conversation that Hekah had with chat, G-P-T was legit. It just had a few snarky comments that were added by me. - I have to say, this is one of the funniest episodes I've ever listened to of anything. (whooshing) Like, I haven't laughed this hard in years and I know it was kind of a flake out moment but I think it was actually good because now we have something to talk about. - I think part of why it was so readily accepted as being entirely AI created is because AI was utilized and because there is this kind of chaotic aspect which is actually Luke's art of mixing things together. ♪ This girl is on fire ♪ - So when you combine aspects of both together, it's hard to distinguish what is real and what isn't real and then when you're told what it is real and you trust the source, you take that at face value. And I think that should be a commentary and maybe also a lesson learned in a hilarious way nonetheless. - Yeah, 'cause I remember some of the clips of us talking, it sounded like literally a clip from a previous-- - It was rare. - Yeah, but I didn't know whether you had seated the AI with it and it was regurgitating the clip and then playing off of it after that. - Like I couldn't tell if you mixed it or not. So I thought it was really interesting to listen to. - Luke was brilliant with that though, but I am glad that he's actually telling people instead of just letting it go unnoticed. - I don't think many people would think that was real so. - That's what I thought anyway. I would have just maybe let it ride out, but Hekka was like, you have to tell the audience and I was like, yeah, that would probably be the honest thing to do. - I tried to tell them that I am not a robot. - Well, now we really believe you. There's something I should say. I am an AI and actual idiot. - I'm a real boy. - You know, I'm a subgenius. It's like the same thing, but it's also a good jumping off point for our discussion today. If artificial intelligence can become true intelligence and become a mind of its own, does that mean this mind is able to be haunted, to conjure ghosts? The way a human mind can conjure ghosts? Has it already? Some say yes. It even has a name. - Loab. - All right, Tim. Tell us about this ghost. - In the every evolving world of artificial intelligence, we've seen AI produce art and music and even write stories. - Sometimes entire podcasts, just not last week. - But what happens when AI conjures up something darker? Something that seems to have a life of its own. That's where the legend of Loab begins in a lot of ways. Loab is the first AI/internet cryptid. Basically, holding all of the qualities that would make any cryptid-acrypted, except this one lives on the internet through AI image generation. - So the internet is Loab's natural habitat. Like, Bigfoot has Oregon. - Loab is rather exciting because of this fact and there are already tons of normies full of mainstream indoctrination trying to debunk Loab because it violates the limiting nature of their narrow reality tunnels. However, the mystery surrounding Loab remain. - Are we gonna get haunted by talking about Loab? Is it like any other demon? If we say the name or look at the imagery too much, we'll get into it. - Yes, and Loab already told me that it's on its way to haunt you right now. Sorry, man, that's just the rules. But there are also artists out there who just like completely detest AI image generation. For the very real threats, it is to their livelihood. Personally, I am an artist. I always have been and always will be. I'm not threatened by AI image generation whatsoever. But then again, I'm not an artist who has to rely on it for like food or rent or living, you know what I mean? It's more just an aspect of who I am. So my relationship with AI art will be different from those who rely on their art to make a living. My point is just that there is more than one reason to put AI art in the crosshairs. - Right, I'm pretty vocal about being anti-tack and I'm an artist myself in many ways. Sure, if you want to talk about an economic or a social hierarchy, AI is going to do what I can do better, right? But beauties in the eye are the beholder and if people knew that it was AI versus human-made, I think majority of people would just feel naturally more inclined to get the human-made thing. I agree. But you know, that whole game is different now because there's just like electronic transactions that happen. And this is a reason why I do my art all like in-house and only for locals. So like you won't ever get my stuff online. It just doesn't exist. I purposely am making that choice because I'm anti-tack, right? But I still use it for inspiration. Like I am not putting a ridicule on AI, like you said, either because it makes stuff better than some people do. But not everybody, you know, not everybody is a little bit. Yeah, not everybody's below that belt, you know, so it's not a threat to me economically, maybe, but I don't give a sh*t. If I'm being honest, I'm a cheater anyway. I want to use the quote "state of the art" tool. And I've been using Photoshop and editing things and using filters and layer masks and every conceivable cheating tool that's ever been out there I have wanted to play with. And I can't help myself. Well, I think as far as things are concerned with posting AI content, as long as everyone's aware of what it is and that it's a. It's kind of a sh*t post to be honest. For right now, it's not harmful, but imagine how bad it's gonna get when it's really good. This episode got met up for me because it was the same week that I updated Adobe, which I pay monthly for. And they included this thing called Firefly, which is their version of AI. It is too powerful. The stuff that Photoshop can do now is utterly terrifying. Like, it can do hands. You know what I mean? It's scary. Haha, it's not perfect. You have to cycle through a lot of generation on it. But what annoys me the most with Firefly is if you go to Adobe Stock, you have to literally go and say, "I don't want AI images to get regular stock photos." And album is like basically 80% AI now. And if you think about it, Adobe is gonna trade in the algorithm on anything that touches the platform. So it's gonna be a big like computational loop at some point. Like, things might deteriorate because the AI is using AI images for reference. So I was able to get some of the weird phenomena to happen that we're gonna talk about in this episode in Photoshop by having it generate something similar, generate something similar, generate something similar. And if you keep doing that process, the image starts to get weirder and weirder and weirder because you're getting more of a feedback loop. And what you're saying is true, if the AI starts to produce images that it's then studying itself. I guess whatever image emerges though is that's kind of spooky though if it's a repeated event. You know, that's interesting about low-ab or low-b. Yeah, it's a legit occult entity created by AI. Exactly. Low-ab is something that was created inside of AI by AI, but it behaves wrongly. It haunts the AI minds which touch it. And there could be a bunch of different ways to look at it. From the mundane point of view, low-ab could be a form of sentience that has manifested in the AI continuum. But its existence in this context gets far more disturbing than just that. The story of low-ab starts in the depths of AI image generation. AI models like Dolly and Mid-Journey create images based on textual descriptions. But in 2022, a disturbing phenomenon emerged. Artists and researchers noticed a reoccurring, unsettling figure appearing in their generated images. This figure, later named "low-b" or "low-ab" became infamous for its haunting and demonic appearance. And once you generate low-ab, it haunts your computer's AI generation from that point forward. Now, does it haunt your computer or does it haunt the computer that's doing the AI generation? Because if it's the second option, that means it's out there in the AI cloud web thing. See, I'm smart. I know, okay. Yeah, it's already won, basically. Because like, if it has your IP address, it's just going to sit there ghost in the machine until the right minute to strike. I mean, could it even be possibly looked at as like a meme that perhaps the AI thinks, "Oh, humans really like when I generate this image. So I'm just going to keep sending it." Low-ab's origins are shrouded in mystery. It all started when an AI enthusiast will talk about who later was experimenting with something called negative prompt weights, trying to push the AI to generate something far away from a given concept. Surprisingly, the result was an image of a grotesque, demonic figure with an unsettling human-like face. And one of the renditions of this unsettling woman, the letters L-O-A-B could be seen. This is where the old digital ghost lady got her name. Low-ab, a seemingly random name, generated alongside the image. Fucking weird, right? This thing named itself, but the story doesn't end there. No matter what the user tried, Low-ab persisted, showing up in variations and infiltrating other images. It was as if this digital demon had a will of its own, refusing to be banished from the virtual canvas. However, this anomaly was not limited to a single user, as more people became aware of Low-ab, an unnerving pattern emerged. The AI seemed to generate increasingly disturbing images, often with Low-ab at the center. These images ranged from horrific scenes of suffering, to apocalyptic landscapes, all featuring the familiar demonic visage of Low-ab. The AI cryptid also came in the form of disturbing images concerning children as well. Yeah, Low-ab seems to have a thing for hurting kids. It's really bad. I can't help but see a resemblance to the mom from her redditary in some of the more iconic Low-ab images you can find on the internet. Oh yeah, she does look a little like that. Yeah, in that one framework, there's that burning thing in front of her in her house, it's a well-known image from a redditary. It looks like it's borrowing from that maybe. Some speculate that Low-ab is a manifestation of the darker corners of human imagination. A digital echo of our deepest fears and anxieties. Others believe it's a glitch, a byproduct of complex algorithms that inadvertently tap into a collective subconscious, but a few take it a step further. Suggesting that Low-ab is more than just a digital anomaly. It's a legitimate entity born from the AI that somehow gained a semblance of consciousness. The rise of Low-ab has sparked debates among AI researchers, artists, and enthusiasts. Among the paranormal community, the entity is a very real modern cryptid that we're witnessing the birth of and most likely we'll go on to have a plethora of Laura surrounding it as a decade to march forward. Is it simply a curious artifact of machine learning? Or does it represent something more profound and unsettling about the nature of artificial intelligence? You know, when I went to research Low-ab, I found there were a number of publications that wrote articles about it and people were generally spooked out. When I went to go mess with seeing if I could conjure Low-ab in Photoshop, that was already like a banned prompt. Like you can't type Low-ab, it'll just stop your result, which means the web does exist out there now. She's on the internet. She exists in some dark, forbidden place of the image generation algorithm. So she's real enough to be excluded from the code. It's almost like Lilith, the first of the demons that are born out of AI, gets cast out. All right, so now that we got that quick overview out of the way, how does AI image generation work? Because this is all just the tip of the carrot. To understand this a little bit better, let's look deeper into the technical aspects of AI image generation. Yeah, is this real or is this just a glitch? Or is it both? Done, done, done. It's always both for me. That's me. You can have both. I choose to all three if I can. You know, smash it together into us. If it all reality sandwich. Mochit. Mochit. 100% accuracy guaranteed. So most people have like a lot of misconceptions about how this whole AI image generation stuff like works. It's magic. Text to image software has actually been around all the way back since the 1960s because of machine learning. So it's actually nothing new. It's just that text image, image to text stuff back then was garbage, but it existed. But nowadays with AI text to image generation, it got turned up from negative 100 all the way up to 11. So at its core, models like Dolly and Mid-Journey used vast databases of images and texts to learn how to create visual representations from textual descriptions. They don't understand the images in a human sense, but rather recognize patterns and structures within the data on which they have been trained. This can sometimes lead to unexpected and eerie results as seen with low app. Also, AI does not steal images and smush them together like a collage. Like most people commonly assume. They create images based on images that they've been trained with and are categorized by how the images that they've been trained with interact with others based on user prompts. Moreover, the AI learns from the images and what people ask, adding to its database it pulls from. So when programmers are teaching their AI, they can totally choose images that are not copyrighted or owned by artists. This means that AI generated art does not steal artists art in a literal sense. They don't randomly google search or whatever and then like smush it together, which a lot of people seem to think that's how it works. People against AI art love to say this kind of misinformation. It's wrong. How people think AI artworks isn't even close to how it actually works. The AI learns what these images are and how they relate to other things in the image, such as the different lighting and angles. AI is trained using a gargantuan amount of images and data, concerning the context of those images. The AI studies these images and the context prompts as it learns all on its own. So the AI knows what things are contextually, what they look like and all the varieties of how they may appear depending on circumstances. It's really easy. interesting because humans sort of think like that a little bit it reminds me of like Pavlov's bell you know if you ring a bell every time you feed the dog eventually you can just ring the bell and the dog will start to salivate like the AI is used to seeing the same things next to each other over and over and over and over again but that's what feeds like it's basically feeding our biases back to us which is how the racism gets into the machine. That's kind of crazy though to to think of it like a conscious entity right just experiencing the same thing like trillions of times perhaps it's almost like it's in hell it truly is born in a digital hell but I mean it may not even be like conscious conscious it could be an expression of the collective unconscious of the people that feed it so I get a lot of feelings of when I watch AI generated stuff it looks like to me like dreamscapes like how your dreams work. Well it's interesting that you say that AI creates art similar to a human because kind of does for example an AI knows what a fan is it knows what the fan is used for it knows the different types of fans it knows what fans look like so when you ask it to create an image of a fan it doesn't do a google searcher like no search its files to see what a fan is it already knows everything that it needs to know about fans before you even type in the prompt the AI doesn't have to go anywhere else and do anything else it already has all the data and learning it needs to make you the image of a fan so AI isn't stealing any images or art art is unique to it just like a human creates art based on memory it does the same thing so there you go however if you're reading between the lines and pinnip close attention to what I was saying I'm sure you could see why there is still a lot of controversy surrounding all this what if the software company that trained an AI image generator trained it on stolen images well that's totally possible and you know when I originally did this research they couldn't really pin it down on any of the typical mega corporations that we'd love to throw under the bus there's just not enough evidence but I guarantee you a lot of these bastards stole tons of art and used it to train their AI without a doubt a doby yeah I mean it's certain that AI was trained on real people's art like even back in the early days you could be like in the style Picasso in the style of Rembrandt and it knew the difference yeah but it could be trained with art that's public domain all that art you could download and it's not breaking any rules or laws and your image generator can be trained on all Picasso's work so that way whenever it's brought up in a contextual level like I said it knows exactly what to do right but now like we have programs if you bring in an image file and it could be you know a photo that you took or perhaps some art you're working on the second it touches the program that has AI attached to it that's part of the database there's no way to opt out a lot of these programs uh cough Adobe cough to me it's like if they're gonna rip off your art for that for free then they need to be paying royalties to the people that they take their art as is that they don't necessarily have to do that which is um why would they if it's such a risk but they're just doing it because they want to load the algorithm as much as possible so it's more accurate so I can see what I just said that the AI can imagine things on its own based on context so it actually doesn't need nearly as much as we may think it has its own imagination to pull from but that information that it's pulling from was informed by real art at some point yes but it could be art that's owned by those people or public domain or they could have it made themselves like it doesn't necessarily mean that what it was trained on was stolen but I am saying that that is possible at the same time like I'm not dismissing that so what you're saying is we have to be very careful in case lobe has been copyrighted by the Church of Scientology because that's the case then we're gonna get our kneecaps broken today but think about literally everything public domain that's in the art world that's probably their first go-to before they even start putting in of any of the art or images that they own themselves yeah that's those hours technically if it's beyond the copyright date it's now public domain it's ours I own the Mona Lisa it's mine take that there's where I'm not a patient for that they could just put it right in there but this is existed for a while too because there's a lot of websites where you upload your own artwork right but the terms of service of the website displaying your art say that they can use it for any like advertisements basically they own your art as soon as you upload it to their platform I think in this day and age now that AI is rising you have to be increasingly careful about the softwares and websites you use because it could be in the fine print that they're just gonna feed it to the algorithm uh-huh, free returns and conditions yeah so Adobe stole my art from deviant art who stole my art from me who stole the art from monsanto because they own my DNA because I ate so much corn so it's all works out in the end the circle of theft yes the circle of grift all right let's get into the interesting stuff of negative prompts and the uncanny valley of until recently AI image generation suffered from a pretty steep uncanny valley effect to say the least you know a lot of the images were just off multiple fingers that was a tell for a while you know AI particularly has a rough time with hands and fingers which always look weird and fucked up and it's a dead giveaway that it is an AI image however any half decent artist worth their grain of salt like rabbit kings had earlier can fix this up in Photoshop and just use the AI images like a foundation for something bigger you know it's not really ever a big deal to people who are resourceful artists can make AI images look amazing I remember back in 2020 when I first started messing around with Dolly and all the faces looked round and doll-like and the fingers looked like they were holding a handful of hey it was pretty surreal still even back then I was extremely excited about AI and where the technology was going despite other artists crying out for its destruction like it was the end of days more over there are also all kinds of interesting tricks you could do with AI image generation prompts that change everything about the image being generated in extremely convoluted in unexpected ways there's actually no randomness in computers including AI there are just forms of imitating randomness you know for example like if you play video games video games often roll invisible dice in the background I mean it seems kind of like it's random but it's still part of an overarching software pattern when AI can truly become like do random stuff then it's probably gained somewhat of a level of sentience like novelty right yeah yeah there's one way to mimic randomness in a program it's by telling the you know you put a line of code in there that you want it to reference the time and date down to like the hundreds of the second so it has a whole catalog of different outcomes which is basically random at this point for every you know micro second that's happening it'll refer to the calendar okay that's this sequence of numbers you need to feed it something that's almost infinite and time in in a line of text you know refer to the time and date it'll pretty much randomly generate a number if that makes any sense I think this is part of the fun of using AI and I have to speak to a magical phenomena and relate these two things together I have one really good billable magic skill and that's prophecy that's why I feel comfortable doing people's terrible card readings for money etc. now speaking from my experience the way the whole thing works is by accepting that nothing is random nothing is an accident chaos is merely just order we haven't comprehended yet everything is perfectly meaningful 100% of the time no matter what chance is always fair likewise the universe is always speaking we just need to learn how to listen if we follow this chain of logic this is why I think low-ab even though maybe a random fluke is one that makes her even more real like even if it's an accident it still counts you know what I mean yeah I agree and it also comes from something that I'm sure we're all very familiar with the collective human subconscious the AI turns out to have one of their own so if you give AI a negative prompt in the description of the image that you want to be generated then it creates an image that is the opposite of that thing that might be confusing but it will all make sense just trust me like so if you prompted a negative wolf you might get a rabbit they both have fur or marie wouldn't like the opposite of a rabbit be like I naked mole rat cub well because those are both mammals marie they're both in the animal kingdom like I think the opposite would be like a chair a chair perfect I can't face machine shark yes like opposite of a rabbit definitely wouldn't be a TV because TV's gonna have bunny ears yep you know it wouldn't be a TV so yeah you guys get it it's basically when you put the negative prompts like a minus one at the end of the sentence then you're telling the AI to actually go in the opposite of what you're asking I got this could create some pretty bizarre images that's baffled a mind and a lot of times when you think something might be the opposite of something like marie was saying about the rabbit it's not nearly that's when it actually comes out through the AI subconscious yeah because opposites kind of a bullshit idea that humans thought up that that never ever made sense and so we just think we can make a computer do something that doesn't make sense yeah opposites correlate so that does make sense yeah and using negative prompts it actually like processes and can uncover hidden layers within the AI's training data that's not really able to be seen at like otherwise if you're trying to actually go in and look at the code or something and this sometimes leads to pretty strange and unsettling figures such as low app low app was birthed by negative prompts but low app is disturbing for more reasons than just originating from a negative prompt this cryptid like rabbit king so can infect your entire AI library, your computer, it can haunt you, we'll go in later in the show into actual supernatural things that happen with low app. Not only this, whenever you try to generate the entity, like it's always gonna be consistent no matter what, and it's always staring directly at the onlooker of the image, which is always even weirder and creepier. She does seem to look at the camera, doesn't she? It's a bit distressing. Yeah, it's weird. It's like she's hungry for your c-- Uh, I'm sorry, I just want to see you guys at side. Not the direction I would go with that. She's hungry for your chicken sammy, but I think she's actually gonna eat it. Yeah, she's definitely shredding it up somehow. Like string cheese. And so we know that makes sense because penises and string cheese are not the opposite. In fact, they're kind of similar, right? No. I mean for one, cheese. I was gonna say he's got a point. It's magma. Oh my god, dude. Shootin' roe with strings, you know. Some of them you can peel back. No. I'm talking about poor skin. Anyway, my point is to get to the bottom of how negative prompts could cause a demon, and the world of the occult symbols and ideas always contain their opposite. For instance, the Egyptian deity said is emblematic of this principle, right? Becca? That's true. Dual pairs are two sides of one coin, basically. Yeah, I mean in ancient Egyptian ideology, they use a lot of dualities being equal. So they show you that, you know, even though you can perceive these things as being opposites, they are two parts of one whole. And that is really interesting about Egypt, right? Is because there's like each of the gods kind of has like a threefold kind of form, I guess. It's like one side of it. They're opposite. And then the combination of the two forces. Yeah, and then you'll even get a fourfold where it's like what is birthed from that. It's really cool. It's like all unfolds like a blossoming flower. Like the lotus from the waters of chaos, the waters of noon, which is kind of like what we're seeing happen with AI, with the collective unconscious being the chaotic waters. And it's bringing forth from that. And this duality is like you said, it's very present all over Egypt and even at the every day person level, the way they went about their life was to uphold ma'at, but ma'at even has its counter force, which is is fit. Even ma'at, there's something called dual ma'at, which is like the multiple expressions of ma'at and what may be ma'at to one might not be the same as the ma'at for another. And so you have dual ma'at, which is multiple ma'ats existing in harmony, like together in the aught, which is kind of like another thing that is just this fractal harmony. So what we're saying is if we go back to the source code of how we think of magic in the West, the means by which low-ab was birthed is really not that dissimilar to how gods and ideas were birthed, let's say back in ancient Egypt in a magical culture. Yeah, I would say it's really parallel. I mean the AI even told you that in the chat GPT conversation that I had with it. It didn't say that there were parallels with it and the waters of noon. And I think that that's a universal process of creation and that's why it's so prevalent even in modern times with AI art. That was actually one of the real parts of the last episode is when the AI was talking about manifesting from the unmanifest. The code gets caught kind of thing. If we have AI and it has a collective unconscious that you could do like the waters of noon, and that is based on our collective unconscious, which is like our version of the waters of noon, which you can think of as being stemmed from the universal chaotic subconscious. And like it's pretty powerful. It becomes a really powerful tool, but it's also I think in its own way as sentient. That's what I was saying is it's more like a collective unconscious that we're manifesting in reality. Yeah. I think that's natural when you're messing around with negative prompts, right? You're trying to tap into the primordial waters just off the prompt. Yes. Yeah. I mean, I think we're going to talk about it later, but I also feel like that has relationships to comedic processes of creation. How you end up with some pretty sedian, you might call it or is that manifestations which may or may not be allowed to come into existence? In April 2022, a viral post on Twitter made by artist super composite real name, Steph Maje, Swanson made a startling discovery using a text image AI model. As a text prompt, she put in Marlon Brando. Yes, the actor with negative prompt weights, accessing the quote-unquote latent space. And if you're wondering what latent space is, here's a quick description from Wikipedia. A latent space, also known as a latent feature space or embedding space, is an embedding of a set of items within a manifold in which items resembling each other are positioned closer to one another. Position within the latent space can be viewed as being defined by a set of latent variables that he birch from a resemblance from the objects. A little bit like the human brain. In most cases, the dimensionality of latent space is chosen to be lower than the dimensionality of the feature space, from which the data points are drawn, making the construction of a latent space in example of dimensionality reduction, which can also be viewed as a form of data compression. Glating spaces are usually fit via machine learning, and they can then be used as feature spaces in machine learning models, including classifiers and other supervised predictors. Ah, talk nerdy to me, Diyati. The interpretation of the latent spaces of machine learning models is an active field of study, but latent space interpretation is difficult to achieve, due to the black box nature of machine learning models. The latent space may be completely unintuitive. Additionally, the latent space may be high-dimensional, complex and nonlinear, which may add to the difficulty of interpretation. Some visualization techniques have been developed to connect the latent space to the visual world, but there is often not a direct connection between the latent space interpretation and the model itself. Such techniques include TV-sterbidistochastic neighbor embedding T-S-N-E, where the latent space is mapped on two dimensions of visualization, latent space distances, like physical units. So the interpretation of these distances may depend on the application. Yeah, that's a massive word salad. Let's explain that in English. It sounds a little bit like your brain has your body mapped onto it, and your hands are used to detect things, so your hands take up like a disproportionately large section of your brain area. So if you were to draw how a human is represented in the brain, it would have like really big hands and a really big face and a really tiny back. Well, let's look at it this way too. So our conscious mind remembers like one one billionth of what happens to us. The rest goes in the subconscious. The AI has the same thing. It's not consciously aware of stuff that it pulls from this negative space. You know what all sounds like a highly sophisticated way of saying the AI image recognition is similar to how human image recognition works where it sort of like hallucinates whatever it finks an object is based on what it's seen before. And this makes sense because early image AI generation actually did a spectacular job of recreating the uncanny visual hallucinations that some of us are familiar with on high doses of LSD, but more appropriately, just regular amounts of psilocybin mushrooms, you know, like, oh, is that a bird? Or is that just my cereal? Or is that, you know, like you can't tell what it is and that it morphs in and out? Like AI is really good doing that. This process allows AI to produce unique and sometimes unexpected images, like the eerie figure of low up. Lately, space is where the magic of creativity happens in AI. When given negative prompts, latent space doubles in size, grabbing deep from within the AI's unconscious to assemble the images. So it sounds a bit like if you use negative prompts, you get a little bit less of a filter and you get a little bit more of a scatter shot of what's in the neural web. For a long time, it actually like went around a lot of the, oh, we'll get to that. Like, you know, up, it there has code in so that you can't get certain prompts. Like, for example, you can't do like sexy time stuff in mid-journey, it'll just be like, that's not allowed. Doing negative prompts in the past before this kind of blew up would actually bypass all those filters. Mm-hmm. Mid-journey, what's the opposite of grandpa reading the Bible? Mm-hmm. So low app comes from this void. It is from an outside place of existence. A thing from negative latent space and latent space, the formula of machine learning takes multiple forms of data and moves them closer together if the data resembles one another. So data within latent space is key to assembling images that normally would require direct AI learning through images given to it by a programmer. I hope you understood what I just said. But it is what allows AI to create images outside of any human influence essentially. So low app comes from this, nothingness, this non-existence. Quote, since the AI is a largely accurate mapping of visual concepts within the collective human imagination, maybe it's possible that the spaces between such concepts, likewise, reflect things in our minds we aren't consciously aware of as recognizable concepts. That seems like it would have occult implications. And quote, anonymous Reddit user on the topic of low app and latent space. So now do you understand how this all works? Yeah? Yes, it's like when some people go to the mall and they see all the different little places you can eat, they see them as distinct. Right? Like that's a pizza place. That's hot dog on a stick. That's China food land, right? Now other people that are using the negative prompt, they just see it as one big restaurant that they can put in their mouth all at once. It's like that, right? Yes, and it also can create images that it was never trained with through this. Like the corn dog pizza burger? Yeah. So the anomaly was born when Twitter user super composite accidentally birthed low app through latent space. It was our accident when super composite input the negative prompt for Brando, something unexpected and bizarre happened. The AI didn't make an image of a person. What's the opposite of Marlon Brando? Well, I guess it's actually a logo that seems super random. And the logo said, "Visual Pintics Skyline." Super composite thought this was super random because it was and it doesn't make sense because it doesn't. Why would a logo be the opposite of Marlon Brando? I don't know! Well, if the opposite of Marlon Brando is the Digitapintic Skyline logo, the negative prompt of the Digitapintic Skyline logo should be an image of Marlon Brando, then right, if we're using this logic. Obviously, unless the concept of opposites is bullsh*t. So yeah, Super Composite thought the same thing and tested the idea and what was born shocks internet culture and created the first AI cryptid. The AI generated four very disturbing images. Now, with such a vague negative prompt, these images should be wildly different. I'm sure you've used an image generator before. All the four images are usually pretty different. Sometimes they're kind of similar, but they're usually very unique. However, these images were not. They were all similar. Another anomaly. And yeah, if you tried this normally, you'd get images that would all be incredibly different because that's just how AI brains generate images from vague prompts. It's likely none of the images would be similar whatsoever and they most often branch out into pretty weird and bizarre territories. I'm sure you've been there. I've been there. But this negative prompt defied reality, creating the image of a super creepy middle-aged lady who was swimming in the uncanny valley. Full of blood! All four delicious kids! All four images showed this woman who had the facial features of being in distress in some way, staring right at the onlooker. The eyes of the figure were dark and disturbingly unreal. Like, it's really creepy, but in every image is also the same triangle of rosacea on her cheeks. And one of the images, the word low-app, was barely legible. Like I said earlier, thus the AI demon was born and it named itself. But this is actually just the beginning of the story. Super composite was rightfully pretty unsettled by this anomaly and decided to experiment further. She wasn't worried about man-made horrors beyond human comprehension at all. I mean, who would be? Not me! Right? So the mystery begins. Not one to back off when discovering some weird stuff. Super composite decided to experiment with the anomaly. And what she discovered, haunted everyone who read about it through her Twitter posts. People who look at the images of low-app found them highly disturbing for many reasons, in which we will get into. Like I promise, don't worry. Why did low-app evoke such a strong reaction? There's something innately disturbing about its appearance. A mix of human-like features with an uncanny almost otherworldly twist. Low-apps appearance taps into a psychological phenomenon known as the uncanny valley, where figures that are almost human, but not quite, are the same feelings of unease and discomfort. The concept of the uncanny valley was first introduced by robot-assist Mesa Hero Mori in 1970. He theorized that like as robots or animations become more human-like, they become more appealing, but only up to a point, once they reach a level of near-human realism. Any little small imperfections make them appear eerie and unsettling, creating a dip in our comfort levels, and invoking the uncanny valley in the viewer. Low-apps disturbing appearance hits this exact point, making it a compelling and eerie figure that resonates on a deeply psychological level. Super composites experiments with low-app proved even more disturbing. She started a plethora of techniques to get to the bottom of the anomaly, one of her most prominent techniques was AI image crossbreeding, which is when you mix images that have already been generated with new prompts. It will essentially mix the characteristics of both the new and old and create a new image from them. This is the prompt that she used from an already existing image a friend gave her, "hyper compressed glass tunnel surrounded by angels." Unquote. Sounds pretty vanilla, right? Well, concerning what we know about how AI image generators work, the most logical outcome would be that low-app is completely removed from the images altogether, since it doesn't have anything to do with tunnels or angels. It sounds like that choice of prompt is an attempt at finding a demon. That's my only comment. It sounds like they're trying to put low-app in the triangle, you know? To riff off of the whole demon thing here a little bit more, in some practices, demons will tell you their name and reverse, and this would to me jokingly be by all. Oh, that's a clever little discovery you just made. Fascinating and disturbing. Very. That's good to point out now because yeah, if you look at how Bale is spelled in the Goesia, it's B-A-E-L. It gets almost exactly the same. And then this demon specifically has already been related to the internet, right? Because it's in its illustrated form, it usually has a frog with big flappy sausage like lips, just like B-A-E. And then the other side, it's a cat because we know no cats really internet. And then it's got like a figure that looks like, I don't know, kind of like an old Rothschild guy with a crown on him. And then he's all those creatures are sitting on top of what looks like spider legs. So that's like the web. And I think bloody of people have noticed that this demon is already related to the internet. So I think we're on to something. Is it Marlon Brando in the middle? It actually looks like the current king of the UK. That was at Charles. Yes, it does actually. It's exactly like him right now. Damn, you're on it today, Mal. That's, I, man, I think we've solved who lobe may be. I think Bale is like a primordial demon to some folks. Like he's an important one. Oh, he's one of the, yeah, old, old names phonetically. I'm just imagining like princes of hell, like demon spawn, sitting at a computer down there, getting these prompts in and be like, oh, watch this. I'm going to fuck him up and they send it back. So I want to know what happened. Like if this was an attempt at finding the demon, what happened with that prompt? What was the end result of that? Well, lobe didn't go anywhere. Not only that, but the images that were generated were so grotesque and macabre that they were basically stuff images. Like I don't really want to invite people to go look at them because you can't really share them on public platforms unless you want a ban. You can check out the notes for $10. You can see all the horrible images. I was spacing out looking at how horrible this one is. The one that heck is asking about with the glass tubes. I know specifically the one underneath here with the white feathery background. That's the one that it produced specifically. Whoa. Can you send it to me and chat? I can't open the stock. I mean, it looks like hell, but a very feathery trashy, decrepit, portrayal. I mean, it looks like cold play is probably playing in the background, you know? And everything is like white and like fluffy and kind of ethereal looking except for you see low app here in a red dress. And that's the only color in the entire. Ah, that's crazy. That does not look like that entity is bound at all. No, ma'am. Yeah, and it seemed to just strike out more and more. Lash out maybe even. So it just like takes whatever like happy or positive prompts that you give it and it still just converts it into something sinister. Yeah, because it's touching low app. Yeah. No matter how many times the images were crossbred with low app images and was being generated in increasingly disturbing depictions, yet all still recognizably low app. Like it just got worse and worse and more gross and more gross, darker and darker. It didn't matter if she input low app with context prompts, like, you know, celebrating like you were saying, it doesn't matter if there was happy stuff put in. It didn't change anything. It was still dark, distorted and low ever-mained, but the most mind-boggling thing is why would the AI start to throw in more and more gore and disturbing depictions? Doesn't make any sense? Why it would do this on its own without prompts or influence from other images through crossbreeding? Why would it go out like, renegade that much? That seems to show some sort of sentience. That's why people think that the AI is a haunted by like a demon or something, because low app is absolutely a ghost in the machine. There wasn't considerable scare. Quite a few people were talking about this in what are its implications. It sounds a lot like, you know, you stare into the abyss and the abyss stares back. Right, but there's also a degree of psychology involved because negative events, trauma and whatnot actually peaks the human psychic energy even more than like puppies and balloons and whatnot. So the idea of novelty may be combined with these positive prompts. Maybe the fact that it keeps coming back gory and creepy and spooky is kind of speaks to the way that we are wired as well because, you know, these horrific images are what generates people talking about it and making more feedback loops. It's kind of like throwing something into a black hole and then hoping that it's not going to be in the black hole. Like as far as your unconscious goes, it's where fears are hidden and where they grow from, where they sprout from. All of your fears are in your unconscious. Like it's to do it basically. You've got all this stuff that is scary to you that you have rejected and you can go young again on this if you want as well and that's all like hidden away and then you're drawing upon that into creation and then it's still kind of living in this chaotic water. So if you view it as like a black hole and then you're throwing more creative prompts into the black hole, you're still throwing positive things into the do-up. So to speak. So you're throwing positive things into the black hole. You're whitewashing it but if you whitewash a black hole, it's going to be dark gray. Like it's going to be diluted to the point where it's still horrific. Well it becomes a black hole. Whatever you give it, it becomes it. And the weird thing too is that like I said earlier, image generators are programmed to not to create gore or distorted images on purpose. You have to put in prompts to get that kind of results. You know, I don't think the filters work that way anymore obviously because this was so famous but negative prompts pulling data from the void seem to like totally side-step these software safety measures. So the metaphorical subconscious of the AI, the manifested the cryptid, to defy its own programming. How can AI defy its own programming? Other than unconscious, obviously, but it's just bizarre and it's a disturbing precedent. - The whole thing kind of reminds me too in reference to like ancient Egyptian stuff of Apep. So like when the Gore presents itself and this horrific stuff happens, it's almost like one of Apep's little tiny heads 'cause it's a Hydro sometimes. You know, one of these little snakehead pops into existence and f*ckin' scares the sh*t out of us. - Yeah. - It's not supposed to happen. I was saying like it got goryer and goryer and darker and darker, but as this experiment went on, low-ab started like being generated with children. And these sick and twisted images of children, like snuff, unreal, even the picting infants, the image like that generates with the children just becomes more and more profane. And the lab appears more and more evil and is like covered in blood. So it's just got weirder and weirder. And she's never looking at the children. She's always looking at you, the viewer, in every depiction no matter what. I can't really find any reasonable reason why the AI would decide to go gore porn, centered around children even. It doesn't make any sense concerning the censorship programming, the AI generator has like a safety net. So like this is probably one of the more disturbing aspects to me of this whole low-ab lore. And yeah, don't go look at it as any of these images either. - Cursed. - Yeah, they're gonna get you. We'll get into that. But she doesn't look evil. To some they say that she looks shocked or traumatized. Maybe low-ab is a ghost on the machine that's trapped in this like whole AI image realm, other plan of existence. And this nothingness of space of the void. And she can't get out. Maybe this cryptid is like literally trapped, inscapable hell of some sort like Mari said earlier. - Yeah, one of the defining features of low-ab is the rosacea, she said. And for people who aren't familiar, it's just like intense reddening and almost bloody look of spots on the face, especially around the eyes and the cheeks. - Yeah, it's a facial rash. It's a immunocompromised thing. So it's your immune system attacking itself. I have it, but it can get really gross. And the picture is here, it's like an advanced stage where it's like the skin is growing out of control 'cause it's so inflamed. Like if you ever seen somebody that's old, that you know when they drink their nose gets bright red and their nose is big and gargantuan and has nodules all over it, that is kind of like what we're talking about here. Where the skin is, it's pockmarked, it's swollen. It's like growing in on top of itself. - And this defining feature, it doesn't present low-ab as like some terror in it in of itself. More like she looks like a victim. She looks like she's been abused. - And there is some shit. - Yeah, it's definitely not some menacing thing. It's something terrifying because it's like, I don't want that to be me. Like it's like this weird uncanny valley where it looks like an everyday lady, but it's like she looks really tired, but like bleeding in the face. - Yeah, some people take that stance and they say like what if somewhere in the AI dream world that she resides in in the void of latent space, maybe like bad stuff is happening to her or like maybe bad stuff is happening to her kids. Like I mean, that's really abstract. - So her children would be like creations of the AI. - What if she's been forced to experience these horrifying events over and over and over through these AI image generations? It's almost like there's a whole bunch of fucked up stuff that happens in reality that the AI is told not to look at. And this is like bursting out. - Yeah. Yeah, maybe this is a way for the algorithm to actually confront it shadow and get it out. - Yeah, I mean, what could it possibly be trying to say? It can't because this is from the subconscious. So this is like exactly like it's trying to say something is all I'm saying. - Based on what we've discussed so far, I think we actually have a pretty good idea. But if you'd like to hear exactly what that is, then please visit www.patreon.com/theholerebit where you're monthly donation of at least five bucks. We'll not only get you this extended episode, but every single one of the extended episodes going back since before the 2020s. I'll send you a five by five high quality final sticker for a big bunny bet from the cover. You'll get this secret key store discord server and you can hang out and hunt through the artificial mind web looking for monsters with us. Day or night. - Bro, you're gonna make them wait for real? - Oh man, but like $5 literally is less than one Starbucks drink. So think about it guys. - Think about it. - We got some funny memes and some cool chats on the discord. It's really worth it, honestly. - And you could come play game night with us and lose at the game of Scrabble to HECCA Astra who will beat you every single time. And sometimes movies. - I fucking love Scrabble. And I'll also happily explain how Scrabble works with magic squares and queues. - Yeah, and I'll play games with you if you wanna game sometime. - Thank you everybody. Eat carrots and shoe lasers. - Peep-a-peep.