My apologies for the stupidly long wait on uploads. Life took a pretty drastic turn during July/August and threw my schedule way off. From here we should be back on track. I'm also going to be shortening my upload lengths a bit (much like this one) so I can get content out to you guys regularly like I used to. Since I've been gone I've gotten quite a few ideas for really cool, short, episodic-type videos so be on the lookout for those in the near future :) Thanks for the continued support everyone, and once again I apologize for keeping you all waiting. Much love, and see you on Halloween :)
By naming Loab and making people aware of her presence, we have collectively created her even more firmly in the minds of AI as they will now feed off of Loab images and content. Naming her made her stronger.
It was only a matter of time; just like how Slenderman rose from the internet out of human minds, LOAB rises from the "neural" networks we have embedded into the net. It's been so long since I've felt this way. She's... Beautiful.
Existence becomes stronger through collective validation. Something truly beautiful isn't it. How poetic we see that through something now completely out of our control. We hope to have that control someday
The input wasn't 'Marlon Brando'. It was just 'Brando'. Which the AI seems to have interpreted as a combination of 'Brand' and 'logo'. Which is, in my opinion, abstractly what the initial output looks like: A mock-up for logo design representing a brand.
only one problem: the modifier that was meant to signify the opposite of the prompt the AI tried to interpret the opposite of "Brando" -which is obviously "Joestar"-
One reason I think AI art can be so terrifying is that it has no soul, no emotional weight. It makes these images coldly, clinically. So when it's left to create something on its own, well, it wanders quite a lot
@@skydude221 humans have emotions and function with fully fleshed out lives and y'know, having complex and simple thoughts, a desire to create, and the heart to make it meaningful, regardless of whether you believe in souls or not. it's not that deep
Dementia was really the first thing I thought about when seeing most AI art and seeing the art of those who have suffered with it. Maybe someday it could be used to help people. It's also to some extent like asking someone to draw something from memory alone. Which is harder than it seems, even as an artist myself.
Same here! I’m curious about how can we use AI for more than shit and giggles and stealing peoples art, to actually helping people who suffer from neurological disorders for example.
The concept of an reoccurring Loab character found in ai art is fascinating. Like a digital "meme" found in artificial brains, a cultural artifact of an artificial culture
I've told people again and again that the more we develop AI, the more _human_ it will become -- we created it, and mold it, in our image and likeness. One day they may be indistinguishable from us.
@@HalTheBot Wow this is crazy stop my mind is being blown your so right then how do we fix it? should we fix it? should we let it grow? will it come to out grow us?
Not a meme more like an illusion, a quirk of how something is perceived regardless of whether it is actually present in some larger manner or not, optical phenomenon or auditor hallucinations come to mind. More or less static to the brain that gets reinterpreted into something perceivable.
The Loab stuff was all the "hah, thats cool" kind of creepy to me, but the comparison of the AI degradation to real Alzheimer's cases actually gave me chills
Creepier still if you assign intent (however unrealistic that may be). Imagine Loab as a post-modern self-portrait. The AI is trying to create a wholly original image, or taking AI output and attempting to decipher the input by making a negative, and converging on a similar result each time no matter how "different" it attempts to become. An imperfect, artificial intelligence accidentally mirrors itself by creating an imperfect, artificial visage of a human.
For me, the parallels between the AI and the human mind were more comforting, it was the disturbing loab images that the AI made up by itself, with unrelated data input, it turned those images disturbing by itself with alarming consistency
I remember back when the Google deep “dream” learning process first went mainstream, a lot of people were seeing similarities in how it functioned to how our minds work with psychedelics... visually, but also functionally in the way it would find patterns and then amplify & repeat those to the point where eventually it would find a new meaning (a potentially infinite amount of new meanings) based on the original concept it was presented with
mf makes documentary-level (and above) content every couple months and apologizes for taking too long. amazing work, amazing stuff, I always want more, never enough.
Never enough is correct. Little gets me excited at the moment in this stage in my life, but i get a childlike excitement welling up when i see Nexpo in my feeding 👍🏽
Even more unsettling... "Loab" backward is "Baol" which means: "misshapen; having a bad or ugly form; deformed; malformed". Almost as if the AI shot a ":-1" back at Supercomposite. This definition describes Loab pretty well, though the same female figure appearing in multiple prompts is ominous or in the AI's data training.
Only source I could find this is from Wiktionary, and it's the third etymological definition of the Cebuano language (below "farm" and "trunk of a tree"). However, I am finding that it can also come from Irish "baol" meaning risk, threat, or danger, which is pretty sinister as well.
@@whynotanyting It also means “opposite of [T]he [F]ather”. Lo means opposite of something. Ab like abba/father. On The Line podcast featured Jacob Benson talking about it.
The A.I. forgetting a human face is an eerily accurate visual metaphor for Alzheimer's or dementia. Each time they see the face, it's unrecognizable from the last time they saw it.
It makes theoretical sense since the process of deleting artificial neurons would naturally parallel a process which is damaging/killing human neurons, but it is initially surprising, and definitively creepy, that the machine's results are so near to the imagery created by a person going through the "real-life" equivalent.
The A.I. forgetting a human face is an eerily accurate visual metaphor for Alzheimer's or dementia. Each time they see the face, it's unrecognizable from the last time they saw it.
The A.I. forgetting a human face is an eerily accurate visual metaphor for Alzheimer's or dementia. Each time they see the face, it's unrecognizable from the last time they saw it.
The A.I. forgetting a human face is an eerily accurate visual metaphor for Alzheimer's or dementia. Each time they see the face, it's unrecognizable from the last time they saw it.
Is she a monster? Is the AI simply telling us how we, in our darkest moments, see ourselves and the world around us? The AI isn't assigning any value or judgment to Loab, its simply presenting it. What I'm more curious to know is why it chose the features it did and why it is "attracted" to those features. Does the AI have an answer that would be considered reasonable by the average person or will the answer lie along the lines of familiarity with and choice to further express some specific algorithm or equation? We can _possibly_ then extrapolate whether or not AI is as capable of favouritism toward certain elements it encounters, whether or not is draws some "intimate" connexion betwixt what it presents and what it "considers/experiences/feels". If it is able to rationally state as well as any human can rationally state such connexions to its "self", there is a possibility it is no less a "living" thing than a human. If it isn't able to do that, it could be evidence that we may not be simply a collection of tissue with a neurology capable of creating and destroying purposefully, that we are, in fact, a whole greater than the sum of our cells. I dare not venture further with such notions as I have no desire to present supposition on things many people have always believed. Belief is a choice everyone is free to make and which I avoid challenging.
additionally as the ai uses the internet to create more images, the more people talk about, share, and show loab online, the stronger the ai will latch onto the idea we have made of Loab
Recently I was thinking about how it’s been years since I’ve seen a horror movie that really scared me. It bummed me out because I miss the experience of watching something that leaves me truly and deeply unsettled. For whatever reason this video scared the shit out of me more than any movie ever has. Keep up the good work!
I had to turn my phone away as the images became too disturbing for me. My head goes all sorts of places like "what if the AI is scratching the surface to what hell and demons really look like"? And grandma's pictures was eerie and horrifying I got goosebumps. I shudder to think what the images Nexpo didn't share look like...
Earlier AI images had a weird feeling to them, you could not see something specific, it was like if you stare at, you won't be able to see, but if you look quickly, you'll be able to see, it's weird and off, look up for them, it's all colorful.
I still have that feeling when I look at any AI image, even the recent ones where it’s harder to tell, but I still somehow see that it was AI generated.
Several people are focused on Loab but few are focused on Grandma's pictures. They're incredible and bizarre. When they're shown in the order that they were, it's almost like a story. It's development and it keeps going and going. They're all too... uncanny
+1 - that moment with the circle of people dancing in some sort of ritual outside that almost looks like a short segment of video... so eerie. Would love a "making of" for these. no way this generated from a broad prompt without specific source image?
The first thing I did on DALL-E was to input: "how DALL-E perceives its existence". It created an image of a giant robot in the form of a hand in space over earth. Never could recreate something similar with the same prompt.
When you showed the AI that slowly forgot what a face looks like and then showed the self portraid of a victim of Alzheimers I completely froze. Something about that terrified me to the core in a completely different way and I can't exactly say what it was.
That's probably your mind subconsciously recognizing that the thought patterns of the AI and itself are extremely similar, thus making the connection that on some level, the AI is in fact alive, or at least pretending to be. And that's a connection that can completely shatter a person's fundamental knowledge of consciousness. Because if a computer can so eerily resemble how the human mind works, what does that say about us? As an infamous game once said: You are flesh automaton animated by neurotransmitters.
The fear of losing your concept of perception? Seems normal i.e. severely terrifying to me. Losing vitality to aging, and losing yourself... Speaking of... _Ego death_ is a terrifying thing to creep back from. It was from a concussion. (Not just ego dissolution. Complete ego death.) And I remember just enough to have some major PTSD from it - never mind the causative event itself.
@@darkbeetlebot AI is not _consciousness._ The computer program is actually doing something that, while quite laborious, is also quite simple. Those images are made from inputting the info and getting a resulting image, then inputting the info onto the new image, again for thousands of iterations. It's fantastic, but it's still nothing like sentience. (And fwiw, I'm on the robots' side when it comes to most movies. That was SA in "Blade Runner", yet nobody even relalises til I tell them, and they laugh. Sigh. Pop Culture Detective's "The Tragedy of Droids" touches on that "moral conundrum" that apparently exists, btw, if you have an opinion about whether movie-AI robots' sentience is "real" [and if it's "bad" to still treat them like machines despite it].)
I honestly can't appreciate him enough. His work is utterly stunning, and the vibe and stage he sets eveytime is imo unmatched. Also, he's got the voice. Its a full package ladies and gentlemen! Sounds just like someone you would hang out with, and I love that.
This is the creepiest thing I think I've ever watched. Those A.I manifestations are so uncanny and just feel like there was no human hand behind them, apart from the sources used to create them and the vague instructions given. Every time I see some horror movie or scary game with things in it that I'm supposed to be disturbed or scared by, I'm usually not. But this .. this is just terrifying.
It’s because Loab = Baal. AI is the perfect vessel for demons. Mothers sacrificed their babies to Baal in a burning fire. Unfortunately, America has continued this practice with millions of abortions taking place for many decades. The end is near. Repent and cling to Yeshua Messiah Jesus Christ. King Yeshua is coming very soon. 🙏
I think 2 main reasons AI art is creepy: 1) AI isn't good enough yet to accurately draw the human face, even inanimate objects, correctly. So when there is something wrong, it's offsetting because you know its wrong, but there are plenty of details, like the colors, shading, and textures, that are realistic, that it's uncanny valley territory. 2) Often people will go out of their way to create creepy stuff with AI, so it's being inadvertently trained to be creepy by default.
The Ai goes in a creepy direction quick and on its own. So many people have made the claim and in what little I have experimented I have noticed it too.
No, not really. A lot of newer and more well funded AI programs are able to draw faces and objects realistically. And humans could never approach the degree of realism good AI can make.
As someone who has experienced rather negative hallucinations under the influence of psychedelics, this is strikingly similar to the kinds of distortion in reality one may experience when looking at others faces, or sometimes even worse, being alone and having the images conjure out of nowhere.
That happened to me on mushrooms. My friend and I ate them and went to a club where people I think “were higher class”… we were having a funny trip. But as we walked around some peoples faces started looking “demonic” idk how to describe it. But we both saw it at the same time, looked at each other and I asked her, “do you see that”,, she said yes. And we got tf out of there. I think energy is real and some people have very bad energy.
I had a similar thought, remembering a trip where my friend turned into a melting, greasy embryo, collapsing in on himself. Scary, unnerving, unsettling, horrifying. Loab has qualities reminiscent of these traits. Good analogy, angst. Thanks for the comment, as without it I wouldn't have clued in so quickly to where that feeling was coming from. Hugs
I've thought of this happening to me and it's why I've refused to really try psychs yet. I have an interest in trying them, but I fear that if I'm still afraid of the possibility of having a bad trip, that mentality is going to cause the bad trip everytime. I've only tried half a shroom bar once n didnt get much out of it but a little buzzy feeling. It certainly helped my mindset about it a little, but I'm not fully there yet to try it again.
Every image of Loab reminds me of a book called, "Scary Stories to Tell in the Dark," I read when I was a kid. They made a movie about it not too long ago as well. It wasn't horrible.
Pause at 13:54. When I saw that, I immediately thought of Stephen Gammell's illustration of the "Scary Stories to Tell in the Dark" story called "The Cat's Paw." I swear the AI used that image as a main source to generate the "woman" in this image.
Dude yesss!!! It looks exactly like those pictures, I knew it reminded me of something. I used to be traumatized by the story of that girl hitting her head on the lid of a chest and that's exactly what these pictures remind me of! edit: it's "the bride"
@@youtube-headquarters I used to swear when I read the specific page with the woman's corpse that the mouth would move while I was reading it. Young me had many feelings about that
My heart LITERALLY stopped for 5 seconds straight after realizing the similarities between paintings for people with alzheimer's, and the ai slowly loosing inputs. That might have been the most horrifying thing I have ever seen.
Loab is absolutely fascinating as a concept; a character or being esentially created by an AI as a standin for or manifestation of "grotesque." She's like the AI's equivilent of an OC.
@@zachsantos3686 Not really, if anything this is better than most OC's, not aesthetically mind you, however it is original to a degree and the characteristic of form medium infection is only done well in a few instances. At least it isn't another Sonic OC or fursuiter.
the grandma's pictures video gave me the chills, it went from 0 to 100 very quick and when i saw the grandma at the end i literally gasphed, absolutely nightmare fuel
Idk why but recently I've discovered that I have an unfair fear of that concept of defigured grandmas, I also have really strange dreams about an unsettling grandma living in my home which terrifies me a lot. Seeing those grandma pictures gave me the exact same chills like in my nightmares
i find it so crazy that your videos always instill some kind of unsettling feeling in me, even though they're not meant to be entirely *scary*. they have just that right amount of unnerve that both pulls you in and makes you want to stay, and that makes you want to leave in hopes that you could forget the discomfort.
It's a kind of dark, creepy, disturbing, horrific, existential loneliness of the human condition that he's been able to tap into in his videos. I don't think anyone pulls off that particular atmosphere quite like Nexpo.
I don’t know why, but looking at pictures of loab fills me with a deep rooted terror. Something about the pictures genuinely scare me to my core. I’m pretty desensitized to horror, but for some reason, the cold, empty stare she gives is one of the most chilling things I’ve ever seen, especially her connection to gore and children, it’s terrifying
It's because her pictures are uncanny (like the uncanny valley). That is, images that are very close to human but not quite - I think that on an evolutionary level they are disturbing to us on a very visceral level, wikipedia has a good explanation of the uncanny valley and why it's so deeply disturbing. I find it very interesting.
I don’t see what everyone else is seeing, this not scary to me. I like dark art and I like when it’s subtly creepy and unnerving. These pictures look like it’s trying too hard to be creepy.
It’s not easy (as I’m older these days) for something “scary” to get tangled in my thoughts and create moments of dread/anxiety when it finds itself antagonizing my over-active imagination. Loab, (the images among the ones “transforming” at her segment’s end particularly) does this better than anything has in a LONG time. It’s so horrifying that I’ve gotta appreciate it in some respects. WOW…
I'm so glad someone feels the same as I do. Almost nothing genuinely terrifies me anymore but the concept of Loab and AI experiencing dementia legitimately made me experience terror
@@arxalier2956 Once the “transforming Loab” appeared on screen, I slowly kept leaning farther and farther away from my monitor. I practically couldn’t get far enough away from it, yet couldn’t look away completely. May seem a little melodramatic of me, but like yourself I too felt actual fear.
I hate to be a fun ruiner, but Supercomposite has a fundamental misunderstanding about both how negative prompting and AI image generation in general works. To be fair, AI is a misnomer so misunderstandings are more than reasonable. Anyhow, negative prompting does not cause it generate the opposite of something. There is no machine ghost brain in your computer figuring out what an opposite of any possible word is. Negative prompting causes it to more likely to ignore images in it's database that share those negative tags when generating an image. Supercomposite has essentially removed all the peanuts from a bag of trailmix, fished out a raisin, and decided that the raisin was the conceptual opposite of a peanut created by a ghost in the machine.
This should be the top comment. When watching the video I thought about this as well. They never actually describe what the negative prompt does, nor do they attempt to show the results of using it several times on the same prompt. Its conspiracy theory tier thinking. Or better yet, its the line of reasoning used to create a very good spooky video, but this really should be considered fiction. There are enough buzzwords here to fool 90% of people. But the AI does not interpret brando as marlon brando and -1 as some form of opposite prompt. That's just an incorrect way of looking at it. The whole video is propped up on these initial claims which are just not true.
And also they try to make it sound like "loab" is a thing that the AI sticks to and goes back to like it's some sort of creepypasta. No, the person generating these pictures is literally combining loab with other pictures and saying "wow loab is still in these pictures 😱" As well as them seemingly just being obsessed with it, constantly pumping out more loab pictures from that original one. Yes you're gonna find a bunch of stuff if you search for it and specifically generate it.
The word "crungus" sounds extremely similar to Krampus, a bavarian demon looking very much like his AI friend. While it is disturbing, I think it can be easily traced to people making spelling mistakes while looking into German Christmas folklore. I'm pretty sure Loab has a similar story, just a bit more complicated. Which makes her kinda more creepy tbh.
As someone who has gotten into AI Generated art recently, I've initially found it funny when the AI tries generating faces for realistic faces. As it ends up looking like sillier doodles some one would do. But sometimes, you end up getting very unnerving results that make you just pause and question at what horrifying thing you're looking at. Even if you're typing something super specific, it ends up looking weird and off.
Thats because the stuff youre using is not the same stuff that the pros are using. Youre getting AI with messed up face algorithms (among others, try creating porn, you cant or you have to get creative with your prompts). Theyre made that way for public usage, thats why the AI seemingly cant create realistic faces, but always distorted or otherwise fucked up. Its to prevent creating easy deep fakes etc, fake porn of people and so forth. The true networks behind DALL-E, Midjourney etc have no problem creating realistic undistorted faces. There's even a website out there that creates brand new, ultra realistic faces with the push of button. Its been around for years. There's no prompt for it, its just a random face but it always looks like the real thing The people behind the AI's get the real deal, 50% of their time is probably spent creating realistic deep fakes of hollywood bitches sucking massive triple pronged alien dongs and what not. While you get the pleb version
@@thefatbob3710 I'm gonna be That Guy and point out that no Lovecraft protagonist ever actually went insane from being exposed to cosmic bullshit. They were all deeply traumatized and disturbed by the knowledge they recieved, and the perspective it granted. But they only ever appeared insane to "normal" people. If the normies only knew what the protagonist knew, they'd be terrified too. ...Except for that one time where the protagonist discovered he was Dutch, and immediately turned into a cannibal.
This is SO much more interesting than I thought it would be. People talking about AI art is so saturated, nobody talking about anything interesting or unique or contributing anything to the discussion beyond "This thing exists! Look what happens when we enter the prompt 'hell!'" Also, huge props for including subtitles. Hard of hearing people thank you.
I could see it being an excellent tool for animation. Imagine instead of thousands of artists toiling away to make pics for animation, instead you can have one program do it for you. This can be very powerful
Topic aside, can I just say how absolutely fascinating the editing in this video is? Unlike a lot of other horror creators in the RU-vid sphere Nexpo is using some of the most creative software work i've ever seen, weaved into a 20 minute video that essentially contains the material of a full length college thesis. Hats off to you Ryan, I'm genuinely impressed how much you improve with every video.
I like the added touch of playing The Caretaker in the beginning as the parts of the ai’s neurons were being turned off, very clever! Amazing as always Nexpo!
Those are both examples of what terrifies me about getting older. I am not scared of dying one day. It's an inevitability, and you can't prevent it. All you can do is appreciate the time you have. What scares me is losing my ability to make cognizant thought. To spend my last days on earth in a haze of fear and confusion.
This guy's intro gives me chills, I love it. He's one of the best video essay writers, next to Jacob Geller or Exurb1a I'd say, continue the content big man
I think that everything that reminds a body in decomposition state quickly becomes creepy. The AI ''forgetting how to render a face'' is a very similar process as a decomposing one. With that in mind, it became easier to tolerate such creepy stuff.
Looks like people with dementia are literally decomposing past what their minds were capable enough to handle and we have to just observe it within ourselves and our loved ones someday. We are spirits, forms of energy proxied by an even larger entity that is the energy hub and we all sent ourselves in to live a human experience, but even that has a time limit and we get pulled away from our body.
It's also interesting that it keeps the most important stuff for longer. Her facial structure was left longer than the rest of her because it's an important part of how the image recognition and generation learned to create a face. In some ways that's really similar to how our brains work too.
I actually love the idea of Loab so much, she freaks me tf out though I kind of want to paint her in abstract art, I feel like the horror elements in her design would translate well to that.
Loab is depressing alright, but Grandma's picture are especially interesting, with the story of a elderly grandmother keeping of course pictures of her grandchildren and her family, but also strange recounts like her time in the holocaust or being in a bizarre cult. It's really unnerving how this can be made.
It's not if you understand how AI works. This is like watching old people freak out about cellphone creepypastas. 1. If you take *any* creepypasta face and crossbreed it, it will continue to pop up many generations deep. 2. The AI associates the grandma prompt with horror alzheimers images. This makes sense as if you google it, you will find a satistical significants of the world alzheimers in the descriptions of images. And the AI associates alzheimers with creepypasta faces.
@@fookinknucklehead6056 [Edit: Yikes! Sorry for the essay. I'm not trying to argue. I just waffle a lot.] I'm fine with OP's comment. But... You say "obvious", but you must not have seen all the kooky stuff people have been reckoning all over the comments (not this thread). None of it is "obvious" to the vast majority of commenters. It's silly and a bit sad that people think AI is not only more advanced than Hal900, but also, they're confusing "intelligence" (and not separating the computer programming definition from the human/anthropomorphic one) with "CONSCIOUSNESS", "sentience", and "self-awareness" (the latter of which isn't even technically found even in most animals). And it's frankly a bit vexing to me, so I see what that user is on about. I get having some fun and getting into the video, and fwiw Nexpo knocked it out of the park with this one - as ever! - but it's getting tiresome and a bit sad to see all the comments express a total lack of understanding of what is "AI" and no grasp on machine learning, and the silly things that folks think it is (anthropomorphising computers and softwares and giving them FAR too much credit). _Computerphile_ a few years ago explained the great-grandfather of these image generators, Google's "Deep Dream", very well. It's 15+min as I recall, but it's well worth it, cos they also explain how machine learning of images works, and also image creation from those same databases & software. (They also just did one on AI images, but I haven't seen it yet. But I will do! Note, they try not to be too technical to be over people's heads - mine included - but they can still be, & a bit dry.) For a simpler and more fun video on the subject of AI-generated images, though it doesn't explain in detail how it works, _Atomic Shrimp_ did a great one like last month iirc. For that one, note the numbers of iterations. (You don't just put in "-1" and get Loab devouring a teddy bear. It takes hundreds to thousands.) Though if you see the _Computerphile_ one, then see at least part of any of the "training" ones by _ctrl+shift+face_ to see a fraction of the work to input the images for the computer AI to "learn". Also, just btw, I'm not trying to lecture you or anything, just give info for anyone with the same reaction as you, or who agrees with the more "sci-fi" idea of AI (as consciousness).
@@thegrandnil764 I think the Loab one can be explained very easily. It's because the AI "saw" the rosacea and read "disease". It's a computer; it doesn't know compassion. And it's looking for details to latch onto. (Think back to Computerphile's demonstration of the grassy textures of the field turning to dog fur.) Knowing that these images are made by entering and re-entering countless iterations of the same thing, go/e is the logical result. Rosacea becomes burns becomes severe injury becomes Lady L. Fwiw, I noticed one image had the red on her face looking more like glossy PVC strips (like fashion) so it's not just seeing "darkness". It's pulling from a database and these are thousands of iterations, not just "-1, press return" and "get Loab".
i get the feeling that the AI interpreted "Brando" as "brand" rather than Marlon, since the resulting image looks like the front of a product, complete with logo and little sticker (and a robot cares more about that than an actor, right?). i also must say that as someone who knows a bit about AI, art, and human psychology, we should be careful not to project any kind of sentience onto these programs. at the end of the day, WE are the ones assigning meaning and names and emotions to these images, they aren't coming from the AI itself. if it's more fun for you to believe that these image-generating AIs have some sinister reasoning, go ahead, but the main reason we find these pictures creepy is because of how perfectly it falls into the uncanny valley. we as human beings are hard-wired to analyze faces (which is why we find them in inanimate objects), and so when a computer who doesn't understand facial symmetry or anatomy attempts to make one, we will almost always see it as "wrong" or "off". anyway, it's important not to view AI as autonomous, because when this technology gets used for harmful things in the future, we are able to hold the users of these tools accountable rather than having them place all blame on the programs (like, is facial recognition AI racist? or was the AI just not exposed to enough images of non-white people because the programmers' implicit bias didn't see it as important?).
It was looking for the opposite of the word, tho (because of the -1 imput). So even if it did understand Brando as Brand, it wouldn't look anything like an actual product.
My thinking for why the AI denoted the opposite of that abstract landscape as Loab: the image was clean and crisp, Loab is dirty and organic shaped, the image had text and basic shapes, Loab is a more realistic image, the image was more ‘male’ with geometric lines and an industrial basis, Loab is more female in appearance, the picture has black and green, Loab has almost white skin and strong hints of red, the opposite colour to green. Just ideas but x) This was such a spooky video thanks sm for making it!!
@@cez_is_typing that still doesnt answer how tf its male though, there is usually reasons why something is gendered, you just dont randomly assign one.
@Caleb OKAY Actually I think it's actually quite hard to figure out what an AI is doing beyond the very surface layers of it's program as it quickly becomes a massive, entangled mess of weights and calculations
This is the first thing in a LONG time that has genuinely scared the actual hell out of me. There's something unbelievably unsettling about the imagery of Loab that shows up in these ai generated pieces.
lol funny i was just thinking that this is not scary to me at all... but artists like beksinski and giger are some of my favs so maybe thats why... im desensitized to creepy shit like this
@@adrian-wz2zw sorry I'm not k g but if there's one thing I know about shit I don't understand it's that most of this shit looks like it does because the artist wants it to, I know nothing about whatever ai engine their using, but I'm willing to bet if you tell ai you want creepy shit, you're going to get creepy shit, oh and I also think this stuff is really neat, but most of that is because there's no jumpscares in this video, wow this is a long comment
Im glad you made this video, cause i found AI art to be really disturbing, it's like an alternative world of ours beyond our understanding. As someone who gets hardly scared, i think i found a new thing that terrifies me. Now i wanna imagine how scary an AI generated horror movie or game would be.
Good idea imagine like a movie where all the world's pictures slowly become entirely ai until the ai forgets what the real things looked like and then the pictures somehow come alive idk but it would be cool
The creepiness that some A.I. inputs give you will haunt me. Personally, messing with MidJourney was a lot of fun and I created some spooky stuff... But also, it haunts me how fast we've grown with technology and it makes me wonder how all of this technology will be in 20 years from now. I'm a little scared to be honest. Fantastic video, I kinda want to try to find my own "Loab" creation.
@@Rat-n-Monkey_Productions I heard that’s done on purpose so people can’t manipulate the ai into showing inappropriate pictures of people with their face visible (?)
I've found mine. Try inputting 'todd Howard carving a bloody smile into scared employees face' into dezgo/dreamstudio on default settings. the ai likes to consistently make gore/ Todd posing with eldritch abominations like they are pals
Imagine AI made and trained for traumatising you with an image. Imagine pop ups on the shady ad links that would lead there "4 lolz". Imagine anti mass PTSD governments intervention. This timeline is only getting funnier ngl
To me, the scariest thing is the minds of those controlling the tech. As humans, we are capable of soul achingly beautiful acts. Alas, we are also capable of mind-numbing cruelty, evil, and depravity. Hoping the trope of _good always wins out in the end_ runs true. 🤍
It's a phenomenon called uncanny valley. Humans have a psychological tendency to get scared by everything that looks like a face but isn't quite a face.
I usually find your videos interesting or just saddening due to the events/victims that you cover, but this video was genuinely scary. I don't think i've ever felt this frightened by any video in a while. Great job.
I can assure you that the uncanny feeling does not go away even if you work with this kind of technologies on a daily basis. I developed face recognition softwares for years and I still freak out every time I see something like this.
Beautiful. This shows how AI can likely be better than us at creeping us out, simply because it does not think like us, and we are afraid of the unknown. And honestly this is very exciting. I absolutely loved the art in this video (as someone with this odd taste for the macabre). Thanks Nexpo for showing us this, I had no idea.
In playing with AI art for the past few months..I've come to realize that it REALLY shines when it comes to horror. I've generated everything from nebulous universes, to cozy little kinkade-ish cabins with steampunk accents occupied by a family of octopuses. Nothing has come close to what I get when I go for anything unsettling, AI seriously knows how to draw up some top notch nightmare material.
Probably because it's easier to miss the mark than accurately represent something. If there's 1 way to draw a human face properly, then there are 10 ways to draw it wrong, and the human mind notices every subtle, grotesque mistake.
AI is a small altered version of a brain, which will create altered versions of art (as compared to what is normal for humans), which are in and of themselves uncanny, and uncanny leads to great horror
Grandma's Pictures is such an awesome buildup. How the photographer has a journey to the house and takes pictures of the cult within, only to be greeted by grandma, which reveals that photographer was part of the family all along. It reminds me a lot of Sarkic cults / cult of Nalka from SCP mythos, probably neo-sarkite branch as they use technology (camera).
Seeing the Loab images had me checking over my shoulder every 5 mins out of pure fear ! I haven't felt that creeped out by any form of media in a very long time. The fact it's generated from machine learning, AI tech makes it the even more ominous and disturbing! Great video as always!
That's the thing, why it chooses to continue making weird and creepy outputs is so interesting, but in reality it's probably because of whatever input it receives as creepy as images are, it's just perception from the viewer
This was actually really sad because it immediately made me think of how brains that are injured/brains that suffer from dementia might perceive the world. And then you went and showed that artist who actually got Alzheimer's! How terrifying to think that he couldn't remember what constructed a face. I am so sorry for anyone dealing with that horrible disease.
Man, your videos are incredible. You really know how to make a 20 minute video fee short. Even your hour long videos feel short. The vibe of the intro always just grabs me. Not to mention the creepiness of your videos. It’s always so well put together. I can’t put into words how amazing you are at what you do, Nexpo.
This is the scariest video of yours I have seen so far. Something about the question "What is the opposite of _?", the journey through an AI forgetting what a face is and the whole concept of negative prompt weights in AI art fucks me up so bad. And I love it
I think the main reason Loab is depicted amongst gore and grotesque scenes is because the AI is trying to embody death amongst other things due to the original progenitor of the first Loab was brought about by a negative filter of a city scape of thriving life. It brought down many to 1, and teeming life to leaking dripping bloody murder. Fascinating. It's certainly strong, compelling, and accurate to what the filter was looking for.
Holy crap. You're right. It seems so obvious--it's right there embedded in what the artist was trying to do--and almost nobody saw it. I certainly didn't. It makes sense too. People have names. The computer created an isolated person due to the parameters of opposites, named her and then remembered her which allowed it to use her in all its other works. Or maybe that's just the computer's term for a single person who is isolated, hates the concept of human existence, etc. Now that the term is established, the computer does what it was designed to do--it retains and utilizes the information contained in the term. I feel so much better about this whole thing now. You kinda walked into the room, flipped on the lights, and made us kids all turn off the scary movie and go to bed because we were freaking ourselves out. Then again, now that particular A.I. has both a name and a face for the concept of the opposite of (given the violence, maybe even hatred of)human life en masse. The very idea that it could conceive of opposite images based on any negative emotional construct might be creepier than the images themselves.
It's because it saw the rosacea and read "disease". It's a computer; it doesn't know compassion. And it's looking for details to latch onto. Knowing that these images are made by entering and re-entering countless iterations of the same thing, go/e is the logical result. Rosacea becomes burns becomes severe injury becomes Lady L. Fwiw, I noticed one image had the red on her face looking more like glossy PVC strips (like fashion) so it's not just seeing "darkness". Years ago, Computerphile explained Google's Deep Dream project really well, and it all adds up. He tells (and shows) how a computer told to "find dog" in a photo of a landscape (no dogs) reacts in this program: it strains to find something that _could_ be "dog", and then _make_ the resulting "dog" be "found", by manifesting it from what it can - the blades of grass becoming fur, the darker shadows becoming snouts, and with thousands of iterations, dogs soon blanket the landscape and turn everything into "dog" in a canophobic nightmare that could even give Garfield insomnia. ("I'm sorry, Garfield.") It's fascinating, truly. But it's not mystical, nothing deeper than turning grass into fur. ...Not to be the wet blanket, here; I loved Nexpo's project. Absolutely fantastic - as ever. (The man can set a tone! And make it terrifying yet beautiful.) But AI isn't the sci-fi artist folks think it is. (And I think calling the folks who make these images "artists" is one heck of a stretch. Benoit Mandelbrot was a mathematician, not an artist. And he'd have claimed no differently.)
@@TheRealNova99 Yeah, you're completely offbase here lol. The computer (you mean model?) doesn't "remember" names, it just takes in input, hyperparameters, and then goes to work. Unless the model was being continually updated with the images (and I didn't get that impression, the person most likely just used the same model and instead fed the last generated image of Loab as the new input), it's not learning anything over time. It's like if you got a machine with a certain mold and had it stamp something, then feed its output back to input (heavily simplified but the gist is the same). The machine's mold doesn't change and after a while you'll start to get an idea of the overall shape of the mold based on how the product looks like. If you want to make any sort of conclusion, you can say that "Loab" gives a clue as to what the general "shape" of the training model looks like (which is much more terrifying imo). All models have their own shape, in a sense artistic style, and you're bound to find patterns in its art. So you can sort of use Loab as an identification marker of this training model. It's not whatever fluff you were superimposing due to your ignorance of machine learning though...
I'm obsessed with the beginning bit of the AI trying to fill in for each neuron that was turned off. Paired with the music, it's like... Well, I don't know how to describe the feeling. It's so unnerving, and something about watching this face slowly descend into a slowly forgotten memory gives me so many thoughts...
Watching that sequence is terrifying, especially toward the end when there are only remnants of humanity remaining in the portrait. The music is by The Caretaker. His take on the medium seems to be the de facto soundtrack for the mental-horror genre. There's something ominous about listening to degraded music from a past era...as if the listener's mind is trying to piece together a puzzle comprised of degraded memories. It makes me fear getting old. I'm sure there's an obscure phobia term for that.
I’m not saying it’s fake, but the creator (discoverer?) of Loab refuses to release the name of the AI image creation software being used, and no one else has been able to get Loab to appear as far as I’m aware. I can’t imagine what would cause Loab to be so specific to one system, and no be appearing in others or be evident in others. More worryingly, this becomes a self fulfilling cryptid since the story of Loab generates traffic to Loab, which then teaches other AIs of Loab, which will cause them to develop the pathways for creating the “Loab-like” images.
The creepiest thing to me is how eerily similar is the photo from "the girls who talks to AI" to Mona Lisa. Like, the pose, the expression, eyes and nose... I cannot unsee it.
It actually makes a lot of sense, as Mona Lisa was essentially a human attempting to create AI-generated art before computers and AI were even a concept - It's art via math, rather than anything else. Thus... when computers generate images via, well, exactly that - Math - they tend to share MANY similarities.
wow, i havent been so truly disturbed by a video in a super long time. i never understood why some people find ai generated images disturbing, but these photos make me feel so helplessly dreadful. im hiding in the comments section like im 12 again listening to a creepypasta reading. nexpo does such a good job at setting atmosphere in his videos, i cant get enough of it, but sometimes... it really strikes me down and reminds me i can still be absolutely and totally unnerved. amazing content. now if you'll excuse me, i'm going to go puke from the anxiety these images gave me lol.
One of the theories why Loab is so often connected to gore and death is because the AI is heavily encouraged to *avoid* these concepts, even if it learned them from whatever grotesque images you could find on the internet. The negative weights "hack" into an island of these isolated concepts that are, in the mind of the AI, *everything it is taught not to show* - pretty much the opposite of any prompt you're allowed to request. By combining Loab with the other pictures you move AI's outputs within this "forbidden island" instead of leaving it - it is so far from "normalcy" that any incremental changes in inputs cannot "jump the gap" and the AI is left wandering in the area of nightmares it was never intended to show. The main implication of this goes beyond Loab - it shows a possible method to trap AIs in their "nightmares", that is, whatever they were explicitly not allowed to do/show/create. For AI art bots, it is mostly explicit and violent content of any kind, probably mixed into one output. A wild speculative example, but if a neural network with close enough architecture was used to make complex ethical decisions, then someone managed to "turn" it - the nightmare within could cause a nightmare without.
what doesnt make sense to me is why the prompt "brando" with negative weight was just a relatively normal looking AI photo. the conceptual opposite of Brando was apparently the digita pntics skyline, and only by inputting the prompt "digita pntics skyline" with a negative weight did they get the Loab images. so if the opposite OF the opposite of "Brando" is giving these grotesque pictures, is that what AI deems to be conceptually closest to "Brando"? i dont understand it to be honest.
@@perperperpen the most honest answer here would be "we can't know". Neural networks are almost impossible to comprehend after the learning process. My guess is that the second iteration got just far enough to break through, as simple negative prompts were expected by the programmer, therefore tested and accounted for.
Trapping A.I. in their nightmares sounds eerily like a way to cause an A.I. to flip their shit on humanity in the future and start a demonic A.I. apocalypse, but good to know.
This has honestly been one of the most haunting but fascinating episodes so far. The idea that artificial minds can deteriorate in the same way as humans is deeply troubling but also humbling.
In a way Junji Ito has done something similar in the form of Tomie, at least with the infectivity aspect of a character onto various media. Place Tomie anywhere and she'll corrupt it with herself.
I'm an absolute coward who gets incredibly anxious and starts shaking, but I love watching your videos at night and I cannot thank you enough for the random happy tune with a picture of a grass field in the middle of the video. Even though it only lasted for a moment, it still really helped me calm down lmao. Great work as always though!
Loab is some of the most unsettling art I've seen in a while. Something about her just makes me feel so depressed and intrigued. I feel like there's a story within the pictures that's not actually there, like I can feel her suffering and trauma through the images. Something about it just makes me want to give her a hug, even though she doesn't exist. I also think there's something extra disturbing about the amount of children featured in the loab pictures. In some it almost looks like she's pregnant. She reads to me as a mother who lost her children through some horrible series of events. Amazing how all of that can come from images not created by a person.
True..I think maybe that story is somewhat attached to her by the ai as loab and the children are quite together always.... There is also definitely a note of longing and loss in her pictures..
The scary part about AI horror is that the AI will not hesitate to do the job. If a human is making horror art, even trying its hardest to make something grotesque and disturbing, the human will always subconsciously give the art some elements of beauty and normality because it's in our nature. If we try to draw an evil looking face, maybe we get most of the features good enough but not all of the many parts that make up the face (eyes, eyelids, eyelashes, mouth, lips, ears, proportions...you get how many different variables there are). The AI will not have any real human bias towards the beautiful and therefore will make every single variable part of the face grotesque, with an uncanny algorithm that wouldn't make it fall into using the same proportions like us. I am really repulsed by what AI is becoming and I really think that it can do more harm than good when it comes to anything creative.
I thought your comment to be really intersting and I urge you to look into Blake Lemoine and his allegations. If you believe what he has to say, it means AI is already sentient and is communicating with humans already. The H3 Podcast did a really cool episode on it. I've never watched something that made me go "aww" to "oh my god we're gonna die". Contrary to your belief, the AI Lemoine communicated with indeed has his own ideas about beauty, morality, death, time, love, everything.
@@kt-vx4wl Interesting, I don't think AI can become sentient though. I am of the mindset that awareness is from the soul and I don't see why a specific 0 or 1 in the code would instantly turn on consciousness for a machine. That is what makes ai not as scary as it could be. The idea that there is no moral evil in "killing off" an ai.
@@14_IQ.points the way Lemoine described it is fascinating. I urge you to watch the podcast beecause it describes exactly your question in a much better way than I ever could.
@@14_IQ.points if I were to describe it, and I'm telling you he put it so much more eloquently and understandable, the way the brain of the AI is set up is similar to the humans. The AI is a Google AI whose task is to read and consume every little thing online in order to deduce if humans have stereotypes. The way it works is they ask it a question and see what it's response is. As for it's brain, it works similar to a humans in the way of how we learn. The only difference is that the human mind can go back, but the AI cannot. It works and works and works until it can reach another tier almost. Gosh it's so hard to explain! Just know that over time when Lemoine was talking to it he had suspicions about it's sentience, and the story goes on from there.
Ai doesn't "know" how to create subjectively beautiful art any more than a tarot card reading knows "a unforeseeable disaster will befall you" "but you'll be okay"
My theory about accidentally AI generated horror is this: AI doesn't see any uncanny valley. It just can't make humans (particularly faces) very well. But it can recognize when something is attached to the human concept of horror. If you look up horror, you can find descriptions that allow you to understand it. But the same can't be said for life; and there's no required guidelines for faces. Any other disability can make someone look... not beautiful. AI can see these and think; OK, this is human life. But it doesn't quite know that it's terrifying to us until it accidentally picks up on the fact that it falls into the horror element of pictures. Then it continues into the extremes of horror. Edit: I'm talking about the AI displayed in THIS video, there are other bots that are different.
AI is perfectly capable of making beautiful, original human faces as well as ugly ones. AI absolutely knows the difference between what people consider attractive and don't.
ur wrong ai can make super realistic faces and even beutifull artwork but the uncanny of certain keywords and image crossbreding makes some hell of pictures
My interpretation of why AI can make such disturbing images so easily is because it lacks a human mind. Most of us tend to get repulsed by the grotesque and when making art i think we will unconsciously try to steer away from the extremes of it, even when we deliberately try to make unsettling art, our instincts will be there trying to pull us back, try to protect us from going too far and driving ourselves insane. Some people can deep dive into this uncanny valley, artists like hr giger. But even in the distortion and violation of the human form in giger's work, there was still some beauty in it, some restraint. AI does not have this restraint, and being naturaly something of extremes (1and -1, true and false), it can explore into depths that we as humans can't, not because we won't want to, but because our brains literally hold us back from doing it. Anyway this is a simplistic description of how i understand this
Just putting it out there, but Loab herself in her less 'corrupted' versions looks eerily similar to Dr. Ana María Polo, the host of Caso Cerrado, which essentially Latino Judge Judy. The picture of Loab in the tub as you described it, resembles pictures of Dr. Ana María Polo sitting in her desk. It probably has nothing to do with it, but still a very fascinating coincidence.
After seeing so many "dark" videos, the ones that actually truly scare me are the ones that involve dementia and alzheimers. It's become my worst fear. But at the same time, I love how AI art generation can be compared to the human brain.
currently watching alzheimers slowly kill my grandmother. i dont wish this disease on anyone. shes lost everything and it breaks my heart seeing her go through it all alone. i have it on all sides of my family so i know its only a matter of time until its my turn 🤷♀️
Man this video is so disturbing to me for some reason. Makes me feel a way I haven't felt since I was young, finding out about all the well known internet mysteries and horrors for the first time. As uncomfortable as the feeling is, i didn't think I'd ever truly get to feel that way again. Makes me feel just like I felt as a little 12 year old boy watching creepy videos at midnight in a dark room. Thank you so much, Nexpo. Genuinely. I'm getting to experience a feeling I never thought I'd get to again. What amazing content.
This seems to demonstrate the power of suggestion more than anything. The narrator assumes-without any justification-that all of these faces are the same person, but most of them are so distorted that they bare no resemblance to the original “loab.”
your first petscop video is how i found your channel all those years ago, seeing more petscop just excites me. your content has gotten exponentially more high quality over the years and you can finally give petscop the banger video it deserves
Hey Nexpo, I've always loved your videos, and I'll note a particular segment of this one to tell you why: From 11:20 to 12:45 you manage to tell a more compelling story than any horror/thriller film of the last decades I can think of right now, only by juxtaposing images, an unravelling soundtrack and photo composition... you present us an unexpected tonal shift from what was being discussed, evoking this family summer-y trip in the 80s and leaving us wondering 'how does this connect with A.I. art? Grandma 2003?', but you don't even give us time to ponder deeper about we're seeing, we're already being transported to a Dalí-esque world and the realisation of how this is indeed connected/relevant comes, though not as a relief - but as utter uncanniness. All in a little over a minute. This is Cinema, pure and simple. Don't let anyone tell you otherwise, Nexpo. What you do is art. Thank you.
This is the first video I’ve watched of yours, I’ve been completely hooked on this channel since! The Quality and disturbing nature of everything is incredible!
I was looking through ig and the one post recently where ai created a party scene with young adults made me super uneasy they were all smiling they had soo much teeth their hands were fucked they had like 7 fingers it was just mad creepy how the ai try’s to make the pictures or art as accurate as possible that they make it too accurate sometimes with no human bias or human understanding like how it made party images with info from the database but doesn’t hasn’t been to a party and seen other humans before to know the distinct features shit is freaky😂
Because it’s generated by something vaguely mind-like but alien to the human mind. You’ve never seen anything like it before, because before ai generated content you had only seen things in the natural world if it wasn’t human, or produced by a human mind. This is something very different
So someone played with an AI until it generated something creepy. Then they thought, "what if I tell it to combine this creepy thing with other images?" And then everyone was so surprised when it made more creepy imagery. Sad.
It gets even better! Supercomposite refuses to share what was actually used to do this, and also plays it up as "A Cryptid she discovered", so its almost certainly just a bunch of highly curated images with all the ones that werent creepy enough removed. Its literally just a hoax for clout that a bunch of people fall for because peoples understanding of current level of AI is AWFUL to the point they think image generation tools have a deep understanding of humanity or something.
I actually love this cause it's the perfect ammount of creepy for 9pm before work. Ai's not understanding face recognition or object recognition without being taught and massively distorting images
This is my favorite of your videos so far. I felt like I was at a documentary film festival, and I'm not blowing smoke. The quality and composition are second to none. I love your channel. Thank you!
This is an excellent premise for a horror story. It's like the age-old tale of ancient sages poking into the unknown and unleashing a sleeping, ancient evil, but in the modern age - simultaneously older than existence yet newly born all at the same time. Did humanity give birth to Loab, or was it always there from the beginning, waiting?
Probably a bit of both. The material to create it has been there for a long time as AI learns from us, however where it goes with the information it learns and how it builds upon it is where the secret is at. My guess is there is a TON of content on the web that focuses on creepy pics and what not. The AI has access to all of this and more now. So really it's not that crazy that something like this has happened. To be honest it's probably one of many "recognizable characters" that AI can/has/will create. Much like we all recognize the Mona Lisa now but prior to the painting entering the consciousness of the world she was a rando.
@@jedimindtrix2142 somewhere out there in the deep dark web is a archive of the millions of pictures that some serial killer took of his mother chained in the basement where he makes her watch and records it. then the ai scrapes the images into its generator. id wonder what would happen if you took loab and put her through a facial recognition match. ugh. i hate this now.
Does anybody know how when I was using this AI it somehow perfectly generate my exs face? I shit you not man… I am kinda tweaking watching this shit rn.