DALL-E Mini: A Meme Nightmare 

That makes a lot of sense

Or it's trying to tell us that Mark is a being beyond our underdstanding and that we should flee as far as possible.

Well, that aleviates my fear that people will use it to commit crimes.

also i think the creepyness fits more because darkiplier is already supposed to be creepy

Dall e mini does allow people like real people faces, politicians famous people etc, because it's first of all as I know not affiliated to Dall e, the original ai that is miles better but it's closed to be used by certain people and you guessed it, not allowed to make real people images and also because it's not that advanced so it cannot be used to try and incriminate people

I had this thought around the Vaporwave Astronaut, I would download a free jpeg of that! But, seriously people would buy prints of that to frame and hang on a wall. Me, I would buy.

Do you think this same conversation was taking place as photographs became more refined and commercially viable? I wonder if artist were saying "why bother painting a landscape when a photographer can do the same in a fraction of a second?"

Don't worry, most artist are selling out NFTs and not doing anything in the first place.

@@dudebruh8534 what are you talking about? Most artists hate NFTs.

Right now, machine learning(ML) is really based on correlation based training with hard/soft labelled data. With causality learning (ref. to causality ladder), models might be able to deduce logic iteratively within similar parameters of current ML practices and place things/objects/subjects into the right context accordingly. Multi modal models using text,audio and visual data have been researched, but the usecases for production purposes still need to be ironed out. But I have my doubts of ML models replacing human artists since these models are fed with other people's artwork, hence novelty would be greatly reduced.

True

Well its an AI generated video you can't expect perfection

Never forget that he even introduced himself to his ice cream like that “Hello Ice Cream, my name is Markiplier and welcome to... my mouth”- Date With Markiplier

What if he just doesnt want to say hello to us? Or that his name isnt markiplier?

@@Hithere-uz6wd I mean ur not wrong lol

1K LIKES

2K LIKES

Ikr it's so incredible

First real reply

The bots are here...

Çhrîśťofêř Pęžěť ×49 SHUT UP LMAO

@@Drifter4804 ikr bots...

ikr ugh, I used to love using that site but now it's dead bruh

What does kill a site mean? English is not my first language

Kill the people who created it?

@@juliatu2116 When a popular creator, like Markiplier for example, brings attention to a relatively small site to a huge amount of people, who then try to access it themselves, the site will get overwhelmed and crash. It's like an accidental DDoS attack

i was late by hours and its still working, slow, but working

i literally sorted the comments by newest to see if anyone said this already

@@RalseiTheRealOmg Guys Raleigh From Deltarne!1!

@@DaMartian hi yes it is me i am the real ralsei

I think it DOES pull from pretty much everywhere on the internet. I was mucking around with it and input a title from one of my videos, and it spat out an image pretty obviously taken from the thumbnail. I think it's fascinating seeing how far we can push the boundaries of art, and it can be hilarious at times, but yeah. I hope as this sort of thing grows in popularity we see more protections in place for artists.

Limited studies have been done about this, and some people have settled on the idea that, like a real person learning to draw, the AI sees references, and then draws their own original creation from what they learn - not by taking pieces of the pre-existing images. The more data we give it, the more original it becomes - like a real person might go from tracing an image, to redrawing it, to learning to draw that image in an entirely new way, if they have enough references.

They vet all the training content for nsfw and copyright, it’s the reason why it can’t generate anything nsfw

@@ponivi But if it vets out copyrighted things, how come it's still able to recreate spongebob and pikachu? it's clearly referencing copyrighted material. :/

@@kiddpenn because the vetters are human and also because it‘s not censoring pop culture icons or highly popular media.

@@𠔁 No

​@@𠔁 I hope mark deletes this comment

@@𠔁 I pressed the button twice

EXACTLY!

@@𠔁 nah this is just a bot advertising

I also think it’s because darkiplier has the same (or similar) looks throughout his appearances in videos while mark wears different clothes and has had dyed his hair

it could also be because we see some of the twisty/glitchy features as inherently 'Dark', so it looks like it fits more than the 'good' characters

D.o.n.t r.e.a.d m.y n.a m.e

another reason i think is the facial expressions are less varied

actually I was thinking bc darkiplier is a lot of fanart and art is just turning everything into symbols like how a tv can just be simplified to a rectangle in a rectangle and is recognizable as it is. So the program can recognize the eyes, nose, and mouth significantly easier than the jumble of stuff realistic faces have.

Mandela Catalogue moment *laughs in alternate*

*YOU SHOT YOUR MOM MARK!!!*

LET ME IN MARK

Uh... I don’t think that’s my neighbor

Ayo?

SJSJJASJ WHAT DID HE THINK WOULD HAPPEN??

THE GREEN, THE AI ASSOCIATED MARKIPLIER NAKED WITH JACK

AAAAAA SEPTIPLIER!!!

Cole nooooooo

wait what part

@@memewin7783 Yes

@@memewin7783 i agree

Same. Rewatching this was sour in the current context.

it is not the enemy; it is the solution

I remember when AI was just a gaming terminology. I want those days back now

@@orang1921 yeah, really solves my problem of having to pay artists

←Notice the name! 😍 Limited time offer, so hurry!🥰

Next thing you know, Lixian starts using it for that series

making his job easier, thank you artificial intelligence

True

​@@𠔁 No 😍

@Don't Read My Profile Photo Ok i wont then

yeah, i think i read on their info blog that that's what it is! nice job figuring it out :0 /gen

Technically DALL-E Mini is called "Craiyon" now, because it was too close to OpenAI's "DALLE-2" (which is in closed beta, but yields much better results than Craiyon).

Well, duh

oh shoot I just commented something saying basically the same thing you did. I swear I didn't try to copy this comment lol.

@@fluxx6181 all good! a lot of ppl watch mark so a few of us are gonna bound to have the same idea

yeah this is exactly what i was thinking!

i'm sorry, the what one? i've never seen anything like that... i dont see a video about this on markipliers video page.... ..... memento mori. ;3

They lowkey look like some indie punk album covers to me

@@zantanzuken Unus Annus is Ethan and Markplier

@@tarajipowell9227 yes, he knows. he’s being sarcastic. (momento mori everyone)

@@Ark00 o h… memento mori

thats because it isnt allowed to make real faces due to the threat of deepfakes

dall.e can't legally recreate faces of real people

But also DALL-E mini legally cannot create ais of real people's faces. So "characters" are fine, but celebrities and real people are not. ( I say characters in quotes because some characters have real peoples faces attached to them, so it's basically a real person that's a character. Which is weird since they can't make ais if real humans)

Give it 140 years. It will be possible. 🤖

@@Sorrowdusk id hopenot, wed run out of jobs yknow (as an artist)

Most people want art to appreciate the artists effort, style and thought behind it. No matter how advanced the ai version is, it still can’t compare to the value of something that was created by hand

@@Sorrowdusk sooner...speed of advancements is tied to speed of information and we are getting close to having a great jump.

@@bigmanbazzaaa that most = a small minority of people . To earn money , an entire industry must need artists , but since AI will take over that part , there will be no jobs. It will soon become a hobby.

I love you for commenting this

Lmao

I understood immediately and this is excellent

Bruh just as i was about to mention mandela lmfao xD

ITS THE ALTERNATES!!!!

Dear God ಠ_ಠ

For fun I looked up Darkiplier with a knife. It was incredibly recognizable as well.

2017 fan theory moment

Living in a fake world is fun

if your dreams look like that you're probably suffering from detail loss.

@@freezingcathedral imagine what DMT looks like

But, Darkiplier isn't a demon

←Notice the name! 😍 Limited time offer, so hurry!🥰

I think for a legitimate reason is that it has fan art to draw from more than real images. Fanart is not usually messy, it is cleaner lines, less gradient, easier to process. Of course, I don't actually know that for a fact, this is more of a theory.

@@dontyodelsohard2456 an ai image generator theory

@Observer that doesn't make any sense, bro.

I think that’s right and the same for antisepticeye they have a clear look that hasn’t changed over time

Yep, the more consistency the thing you describe has in it's looks the easier time the ai will have in making a recognizable image, it's also why all the markiplier prompts kinda use a wide array of different looks but with the handful of defining features always present, it has a wider range of looks to generate and with this model that just means it'll be a little less specific, hence the weird hybrid images.

Yeah, I also feel like the fact that most of the AI's references for the characters being fanart (as there's probably more fanart of Dark & Anti out there than uploaded screenshots of videos) helps a lot bc it seems to do that easier than it does photorealistic stuff, at least in this version.

yeah, I imagine that the AI is drawing from very simple and clean 2d fanart instead of complex 3d photorealistic pictures, which makes it easier to handle.

4:03 they killed it

Literally

who is darkiplier

WAIT is he the guy that tricks you in adwm because I hate that guy

@@yellobanana6456 it's the evil alter ego of markiplier

But also that legally they can't create ais of real humans but since dark is a character, it's allowed. If someone tries to make art of Mark for example, it'll mess up the face intentionally to prevent people doing doing really shitty things with the generations.

I know I just wanna try it!

It sucks because i really wanted to try it but now i can't because so many people are overloading this small project.

To be fair, it's been overloaded for several days already. This video added fuel to the fire, but it's been been giving the too much traffic message for a while.

Just keep hitting run, itll eventually work. And probably set their servers on fire.

@@beaveresque9404 Yeah, I think vinny fans are also to blame.

LOL 🤣

Another entity confirmed

Meat canyon:try me bitçh

Me: hmm… adjust here… and got it. I have a new refrence to draw from.

@@endstudios9345 k

by stealing

@@jstar3382it’s always been stealing from the start but it’s way better at it now

i can't believe its just been one year, holy cow dalle 3 can even spell

​@@jstar3382yeah, there's a huge amount of labor theft going on. The fact that without knowledge, consent, or compensation the work of countless people has been stolen

​@@jstar3382every artist since the dawn of time has stolen in this way. The greats learn by looking at art that already exists and trying to recreate it then they mix the things they have learned.

The art part has already happened using other tools, but this one has an insane ability to match almost exactly what you expect.

The new version Dall-E 2 is already there. They currently have rules against generating images of real people but it could probably make realistic versions of all of these

@@ModernEphemera I can imagine why; it'd lead to a whole new genre of deepfakes and then some serious legal trouble

i think its cool took i just wish the site would wok

The Dall-E AI has an applied filter where it will not allow the AI to generate real-life faces. It is fully capable of doing such, but the company behind the Ai has filtered this ability so people cannot create realistic photos of celebrities that could cause some bad stuff to happen to that celebrity. These photos can get *extremely* detailed, and with some touch up could easily pass as a real photo.

I think a horror game using this would be really neat. Like monsters are randomly generated by AI so you never really encounter the same thing twice

Yeah I was thinking the same thing. I think this could open up a new range of ideas, especially for indie horror game devs

I actually have used some ai generated art in my horror game. It wasn't for anything as cool as your ideas though 😅 I just couldn't figure out what to use as a portrait for a slime monster

There's some horror games (sorry, can't remember the names _or_ who made them) where the _backgrounds_ are made using a similar AI-art-engine-thingy. Helped add to the surreality of what was happening in-game.

It’s already used (I think) by jerryterry’s song ‘Kiss Me (Kill Me)’ for its art cover. You could take that as an example

But also the suits seem so much clearer than the face so I reckon it sees 'black suit' and takes stock images. That combined with the fanart.

@@ghosty8193 not just stock inages, there's probably like portraits of ppl in suits from other sources like senior portraits or business exec photos.

@@ghosty8193 Oh yeah of course! It’s not exclusively using fanart but I think the AI can read objects in drawings and paintings much easier and then takes reference from many sources of the things it recognizes, like the clothes, hair, things they’re holding etc.

That's also what I thought!

Its so impressive how some even try to get the Anaglyph effect Dark usually has and have varying defrees of success, specially when it comes with red

What I’m interested in seeing is if humans do become obsolete/wiped out and AI/robots “take over”, would they actually “think” humans looked like this or would they have enough information to see us properly

How long will it be before AI can create original styles with simple inputs? This technology can very easily put all artists out of work.

​@@thelouster5815 it clearly cannot lmao

​@@thelouster5815it literally cannot do that right now.

Exactly. The dark characters are largely driven by fanart, which is simpler than real-life photos and therefore more easily replicable.

I think the ai is just not allowed to make real human faces, or so the top comment says

Nice try, Lixian in disguise

@@michaelduncan8531 Yes. I am Lixian. Totally… totally Lixian…

Nice try Lixian...

@@TheDankShrimp Thanks, I am totally Lixian. Completely.

@@SmuggyOcelot Yeah, caught you good this time didn't I?

Something interesting to think about is that, if the AI makes such good art that everyone uses it, it will likely end up raising the quality of all art. The baseline to start off with, IE any old shmuk typing some words and getting a picture, is about where professional non-AI art would be. But, those who put their thoughts into words the best, think of creative things to make into art, or things like that would be able produce the best art. Artists might learn new languages so they can specify even further to get a truly perfect image, or they might learn to or be given the tools to tinker with and modify the mechanisms of the AI directly, which would give artists each their own unique style based on how they built their AI. And if AI programs that can really produce art that good do become widespread enough to overtake all other art forms, then it is a guarantee that things like that would emerge, because anyone clever enough to use the tools they are given the best will be recognized and gather a following, like artists of today do, and the owners of the AIs would seek to provide as many customization tools as possible to make art that utilizes their AI be better than competitors’ AIs. Assuming Communism doesn’t take over the world by then, there would be competition in the field, because it’s literally powerful enough to take over the entire medium of art, so the best AI maker would be rich. So in the end, creativity would still be a part of art, because it’s not like, once we get great pictures (by today’s standards), we’re just going to sit down and say “that’s it, we’re not going anywhere further with this.” Think of CGI; having ways to create fake things in motion picture didn’t kill movies, even as CGI became good enough to look exactly like and then even better than reality. It killed some aspects of movie making, but replaced them with even better aspects. Films that looked like reality, with full color, sound, and creative special effects used to be the peak. But now we can have realistic movies about going to space, fighting dragons; hell, any action scene from a marvel movie would blow anyone’s mind before CGI was invented, but now thats just the new normal. Creativity is an integral part of art, and it can never be taken away. Edit: Holy shit I just wrote an essay there. TL,DR: Art will be fine.

@@stickmanonastick6089 i agree but i believe this is too short term in thought. first humans are the artists and now they're moved back a step to being to ones guiding a machine to create great art. but what comes next? an AI is made to pilot the art generator and create brilliant art without any need for human input whatsoever. why wouldn't this happen? so far, anywhere that AI can be put into has had AI put in without concern for if it should be so why would this be different?

@@Asodym “They’re moved back a step to guiding a machine” Couldn’t the same be said about digital art? It’s not like the software is doing the art, but you are no longer mixing paints or finding the right brush, all that is done by a machine, and you just select the right color or brush or whatever it is you want from it. So, somebody before the digital age could say that same sentence about digital art. As for an AI so good that it needs no human input, then a genre of AI only art would emerge, and if it advances enough, it could replace mainstream art. But creative artists, whether they are good or not, whether they are recognized or never show a piece to anyone, artists will still want to make art for the fun of it, and thus there will have to be a way for artists to create art. How many people today make art that nobody lies eyes upon? Those people would still keep going in this theoretical AI age because it’s not like, just because someone, or in this case, something, can make art better than them that they will just stop. Art isn’t only defined by its beauty in others eyes. Plus, it’s all subjective anyway; some people might like AI art, other might get tired of it and begin appreciating human art. In the end, there will still have to be creative art, because there will always be creative people to make art or defy norms, the creation of art cannot be stopped, and I doubt that a single art form could be so powerful so as to wipe all others into irrelevance. Like pixel art; just because modern computers can produce images that look like real life, doesn’t mean that the retro style, pixelated images disappeared. As seen most often in video games, the birthplace of pixel art, going retro is now another style that some choose to do, and people appreciate, even if the newer, more realistic graphics are “better.” Jeez, why can’t I write this fast at school? I’m impressed you even read my first response, it was so long.

@@stickmanonastick6089 i just hope you're right.

@@stickmanonastick6089 but it’s scary to think about with how a lot of artists do commissions and earn money off of their own art work for a living, but now there is a ai that could do it faster and free, their customers are going to go down and money could get scarce.

I'll be sure to never test it out.

@@elmonextbot Good idea.

Someone please link me the website to make these images! I’ve been trying to find it for so long now, but have not been successful

@@markherrera7640 just look up dall-e mini and click on the first link

@@cameronordaz2320 Please?

I literally just thought that.. AI progressing so horrifyingly fast-

yeah AI has come a long way since... fr

@@_vannah_uwu3433 not.. really "horrifyingly" people tend not to understand that AI is only as good as its dataset, the early days of image generators were basically tech demos, using a very small dataset, just enough to get a "coherent" image. Its moving as fast as technology usually moves, and has slowed down quite a lot since the big accessible boom.

Also "this is not a happy place"

imagine the mandela catalogue generated from this, holy shit

It’s not trying to form a face, it’s specifically coded to not be able to accurately replicate human faces to prevent it from being used to make deepfakes. It may want to make a face, but it can only make faces that are recognizably nonhuman

maybe soon

funny

another thing is that it simply isnt allowed to make human faces lol (good observation though! not trying to imply youre dumb seeing as that info wasnt included in the video)

@@kaIeidoskuII This is Dall-E mini. It can do faces.

@@nikkiofthevalley ohh my bad then

I don't

Hi Random Markiplier Clips.

Hi!

@@FunnyMarkiplierClips You make awesome clips

They're so fun!!! I can't wait to see them in newer Mark and Lixian compilations lol

the unus annus one where it put both Annus and Unnus as spaghettis (the very middle image) is what got me

i was gonna comment something along these lines

Hmmm. There is literally a photo of Mark in nearly every single frame of every video he’s uploaded and more. Even if it was 1 frame a second, I’d still wager it beats out the amount of fan art. But that only matters if it can extrapolate from videos of course. Top comment right now says it’s because the AI is not allowed to accurately recreate faces of people, but characters are allowed - don’t know whether that is true, but it sounds plausible.

It's because it cannot fully legally show real people's faces, but since Darkiplier isn't a real person, it allows it, even though it's still Markiplier. This can work with any live-action character too. Typing "Captain Marvel" would look better than typing "Brie Larson"

The main reason it fucks up faces is because DALL E isn’t allowed to make human faces

they didnt allow dall-e to make real faces to try and prevent people from making like deepfakes and stuff

←Notice the name! 😍 Limited time offer, so hurry!🥰

Excuse me?

Same. It’s very similar to when he was doing stuff on artbreeder.

Ugh so many bots

The bond Markiplier and Lixian share inspired me to encourage my girlfriend to ask for things or mention details for editing. As her videographer and editor, it's fun to try new things and learn along the way. I'm certainly not at Lixians level of editing, but is definitely someone to look up to. :)

@@STICKOMEDIA it’s a bot

@@xaviorshideout oh lol

@@xaviorshideout so confused now, there's missing comments or something. what/who's a bot?

@@i1DERProductions you sure act like a bot.

I think so, too!

Interesting

^^^^This! And also likely that a lot of depictions of characters like Dark -- both canon and fan art -- have him facing forward; the images that the AI used as reference are more or less the same, which is why the result comes out so clear

I think part of why stuff based off of drawings looks better is also because a wobbly drawing doesn't look as uncanny as a wobbly realistic picture

Also, the colors in depictions of those characters tend to be more muted, which means less variables to have to figure out.

Mark can see the future

^^^^ What is with all these replies? It's like a bot civilization being formed.

@@𠔁 stop scaring people with this shit

@@𠔁 lmao amogus script

@@katherinesmallbean3594 bots reply to top comment so they are seen more easily than a regular comment

@@katherinesmallbean3594 yea sometimes big youtubers have a time where there's multiple bots under every comment. Bots copying other comments, etc. It's so annoying

new junji ito story? 💀 /j

Always keep a recorded version of the behind-the-scenes process as proof

That's kinda already happening. It's disappointing to know as artist that one day I could lose my job or get accused or using ai instead of putting hours of work in. It's definitely scary.

Are you AI!?

@@destrojer124 yes. You are next.

What? That doesn't make much sense

Without the clarification, it would be better

You ruined the joke, congrats you lame-o

That got me for a second... Nice one man 😂

Curiously, I too did this. But it gave me a bunch of stuff that vaguely looked like video cameras sooooo... 😬

PFFFT

100% just wrote the same thing. How much power are we going to give computers over human beings and where will the line be drawn with A.I? I know artist gives me the purpose i have in my life besides the droning days of all the stuff humans have to go through. Sleep, eat, work, go home repeat, maybe have a house, save money, possibly have kids, die. Anyone cause do the basics. Art is what i feel like drives me as an individual. I drew my picture thats of my husband and i. And so i do make side money tryin to do art. And it's already hard as heck now days because everyone expects stuff for extremely cheap when materials are not. It's making physical medium not worth anything and instant gratification more popular. It will cause a new lost market for people to do it. It concerns me as well.

you know how people have a fondness for vinyls and books despite mp3 and e-books existing? don't worry. humans love human things. an ai can't replicate the human soul and, besides, we can create for ourselves and ourselves alone, worst comes to worst

I was thinking the exact same thing. What if I create art and it’s not as valuable? someone can just throw their commission into an ai image generator instead of paying an artist

@@thatweirdmoldycarrot4262 there's a lot of people who would threaten an artist with that to try get them to lower their price as well

That's how people felt when computers were new in the 80s. It's just that again. As artists we have to evolve with the times.

I think that's only for DALL-E 2, not for this one.

I mean I put in "man eating orange" and the stuff it generated was equally as horrible

DALL-E 2 is not the same as DALL-E MINI. DALL-E is only by invitation and users agree to avoiding using specific keywords. DALL-E Mini is a 'clone' created by a completely different team, and its open to everyone. I'm sure they can also want people not do to said things but nothing is stopping people. and as we know how the internet works, if they specifically ask people not do something they will do it.

I saw it create a 3d homer simpson in a suit And it actually looked like fan-made

Ikr

how would it be dangerous?

@@ashtons.8501 People could probably use it to make pictures of real people doing unsavory things for blackmail/slander/defamation purposes etc

@@3000spiders oh so it would be like the problem with 'deepfakes'? ig that makes sense. i still wish it could be available to everyone though

And Midjourney is shit compared to Dalle 2 or Google's Imagen. Things are advancing so damn quick

@@amazingsoyuz873 I think some of the creation that Midjourney is capable of, Dall E 2 and Imagen lack- (difficult to say definitively, but I’ve been using Midjourney for a while now, and it has its own ‘personality’ so to speak, with certain types of visualizations that it excels at (and others it does not.) I can see these tools ultimately having their own specializations, and people going to a different AI for different types of image generation.

As of now

Use this time wisely while you can!

This will change the moment someone powerful enough deems it profitable to do so

And I think there was even an official court (Supreme Court maybe?)that only humans can hold copyright, so there's have to be a case to overturn the decision, which isn't impossible, but makes it more difficult hopefully

There's still a question about whether creative works made by human-directed AI can be copyrighted or not.