Your thoughts on Dalle2, Imagen, and other AI?
3 years ago
General
https://youtu.be/NYGdO5E_5oY
https://youtu.be/g9Z0pqsCUhY
https://youtu.be/yCBEumeXY4A
https://youtu.be/PdfFRlabohg
https://youtu.be/XjTENlTTWk8
Here are my thoughts...
In the course of my thirty-eight years of life, never have I ever been so unnerved by something and yet simultaneously entranced by it. I need this in my life to an extent that I cannot adequately convey but I'm also aware that this technology will, at some indeterminate point in the future, possibly eat my job or at least parts of it. There may be some people who prefer a human artist's work on principle or for aesthetic reasons but there will be many more, maybe even a majority, who will prefer the infinitely faster, cheaper AI options.
I believe there will be a grace period where Dalle, Imagen, and other such technologies serve more as tools and aids than direct competition. Being an aphantasiac, the idea of iterating dozens of options instantly, it's literally a dream come true. I have no mind's eye and I can't visualize my ideas which can sometimes make my creative process nightmarishly difficult. The idea of being able to just type something like "white cheetah gryphon standing on a crystal rock under a starry twilight sky" or "machine viscera" or "rainbow volumetric smoke swirling around a faintly glowing fluorite crystal" and immediately get dozens of iterations... I could probably sit here for an hour trying to adequately articulate how desperately I need that in my life and still fall short. Though, it's worth pointing out that these AI don't always hit the mark, lol. https://twitter.com/Chitwan_Saharia.....81129820512256
It tried. Bless its heart.
But eventually, that grace period may wane and we might see a point where digital artists become obsolete. The answer to this is probably, to put it bluntly, adapt or die. Perhaps our jobs will shift more towards cleaning up or curating AI imagery. I've actually already done some of that and I don't hate it. I find it kind of oddly cathartic. Or perhaps we'll dip out of digital art into real media. Many artists may just opt to find a new career path entirely. Whatever the case may be, this technology will rewrite the world, it will absolutely force people to adapt or die, and it will challenge us in ways we can't even predict.
In a personal bit of irony, I assumed "the robots" would come for my job last, not first. As founder and CEO of OpenAI, Sam Altman, said in his Dall-e 2 launch blogpost:
"It’s a reminder that predictions about AI are very difficult to make. A decade ago, the conventional wisdom was that AI would first impact physical labor, and then cognitive labor, and then maybe someday it could do creative work. It now looks like it’s going to go in the opposite order."
I'm prepared for the eventuality that I may have to find some other source of income. It does scare me a little but mostly, I'm excited. I'll still make art regardless, even if the AI "terk muh jerb." I do art because I love it and I wouldn't rather be doing anything else. It's my reason for living. It's the first thing I think about when I wake up and the last thing I think about when I go to bed. No AI can ever change that. I think AI will, if anything, enhance it. Rather, I hope so.
That is assuming these AI are ever released to the general public. As it stands, they are being withheld for safety concerns. Frankly, as desperately as I want it, I understand the hesitancy. Let's play some AI misuse mad libs!
• [Politician name] doing [terrible act] to [child, elderly person, dead body, animal, etc], images taken on a cellphone.
• Two male [racial minorities] doing [type of violence] to [police officer, priest, comatose victim, etc], security camera footage. There is blood splattered across the floor and walls. The two men are smiling and one is holding a [weapon description].
• [Celebrity name] doing [extremely perverted act] with [celebrity name, politician, teenager, etc].
• [Celebrity name] having sex with [celebrity name] on a hotel bed as seen from outside a window through dingy blinds, taken with a cell phone.
• [Input an image] Remove [person description] and replace with [person description].
• Ladies and gentlemen of the jury, I submit for evidence... [AI-generated images that are indiscernible from actual photos being shown to a room full of people who are completely uneducated on the topic].
These are just a few things I could think of off the top of my head. There are people a lot smarter and a lot more sinister than I could ever be who would come up with gods only know what. The reality is that a significant portion of the public would believe such images without question and if even a small fraction of them acted on it, we've got serious problems- I mean the kind of problems where innocent people get hurt, killed, or have their lives ruined. This isn't a very big leap. People have committed heinous acts based on way less.
I want this technology desperately. I want it more than I think I've ever wanted anything. But I don't blame Google or OpenAI in the slightest for being hesitant to hand the genpop such a powerful tool. That said, it's only a matter of time before this technology finds its way into public hands and if we're not prepared for it, we're gonna see some ugly shit.
Anyway, those are my long-winded thoughts. How about you? What do you think?
https://youtu.be/g9Z0pqsCUhY
https://youtu.be/yCBEumeXY4A
https://youtu.be/PdfFRlabohg
https://youtu.be/XjTENlTTWk8
Here are my thoughts...
In the course of my thirty-eight years of life, never have I ever been so unnerved by something and yet simultaneously entranced by it. I need this in my life to an extent that I cannot adequately convey but I'm also aware that this technology will, at some indeterminate point in the future, possibly eat my job or at least parts of it. There may be some people who prefer a human artist's work on principle or for aesthetic reasons but there will be many more, maybe even a majority, who will prefer the infinitely faster, cheaper AI options.
I believe there will be a grace period where Dalle, Imagen, and other such technologies serve more as tools and aids than direct competition. Being an aphantasiac, the idea of iterating dozens of options instantly, it's literally a dream come true. I have no mind's eye and I can't visualize my ideas which can sometimes make my creative process nightmarishly difficult. The idea of being able to just type something like "white cheetah gryphon standing on a crystal rock under a starry twilight sky" or "machine viscera" or "rainbow volumetric smoke swirling around a faintly glowing fluorite crystal" and immediately get dozens of iterations... I could probably sit here for an hour trying to adequately articulate how desperately I need that in my life and still fall short. Though, it's worth pointing out that these AI don't always hit the mark, lol. https://twitter.com/Chitwan_Saharia.....81129820512256
It tried. Bless its heart.
But eventually, that grace period may wane and we might see a point where digital artists become obsolete. The answer to this is probably, to put it bluntly, adapt or die. Perhaps our jobs will shift more towards cleaning up or curating AI imagery. I've actually already done some of that and I don't hate it. I find it kind of oddly cathartic. Or perhaps we'll dip out of digital art into real media. Many artists may just opt to find a new career path entirely. Whatever the case may be, this technology will rewrite the world, it will absolutely force people to adapt or die, and it will challenge us in ways we can't even predict.
In a personal bit of irony, I assumed "the robots" would come for my job last, not first. As founder and CEO of OpenAI, Sam Altman, said in his Dall-e 2 launch blogpost:
"It’s a reminder that predictions about AI are very difficult to make. A decade ago, the conventional wisdom was that AI would first impact physical labor, and then cognitive labor, and then maybe someday it could do creative work. It now looks like it’s going to go in the opposite order."
I'm prepared for the eventuality that I may have to find some other source of income. It does scare me a little but mostly, I'm excited. I'll still make art regardless, even if the AI "terk muh jerb." I do art because I love it and I wouldn't rather be doing anything else. It's my reason for living. It's the first thing I think about when I wake up and the last thing I think about when I go to bed. No AI can ever change that. I think AI will, if anything, enhance it. Rather, I hope so.
That is assuming these AI are ever released to the general public. As it stands, they are being withheld for safety concerns. Frankly, as desperately as I want it, I understand the hesitancy. Let's play some AI misuse mad libs!
• [Politician name] doing [terrible act] to [child, elderly person, dead body, animal, etc], images taken on a cellphone.
• Two male [racial minorities] doing [type of violence] to [police officer, priest, comatose victim, etc], security camera footage. There is blood splattered across the floor and walls. The two men are smiling and one is holding a [weapon description].
• [Celebrity name] doing [extremely perverted act] with [celebrity name, politician, teenager, etc].
• [Celebrity name] having sex with [celebrity name] on a hotel bed as seen from outside a window through dingy blinds, taken with a cell phone.
• [Input an image] Remove [person description] and replace with [person description].
• Ladies and gentlemen of the jury, I submit for evidence... [AI-generated images that are indiscernible from actual photos being shown to a room full of people who are completely uneducated on the topic].
These are just a few things I could think of off the top of my head. There are people a lot smarter and a lot more sinister than I could ever be who would come up with gods only know what. The reality is that a significant portion of the public would believe such images without question and if even a small fraction of them acted on it, we've got serious problems- I mean the kind of problems where innocent people get hurt, killed, or have their lives ruined. This isn't a very big leap. People have committed heinous acts based on way less.
I want this technology desperately. I want it more than I think I've ever wanted anything. But I don't blame Google or OpenAI in the slightest for being hesitant to hand the genpop such a powerful tool. That said, it's only a matter of time before this technology finds its way into public hands and if we're not prepared for it, we're gonna see some ugly shit.
Anyway, those are my long-winded thoughts. How about you? What do you think?
FA+

Second of all, you're only seeing the successes. You're missing out on the far larger set of failures.
"AI," (which isn't even an AI) is a gimmick pushed by people who don't actually understand digital technology.
Amusingly, your second statement pretty much lines up with artists as a whole and the way they present online. Most people see only the successes and miss out on the far larger set of failures. So honestly, then, these models are closer to artists than some humans are. It's pretty cool shit.
"AI" is getting bandied about like a marketing buzzword right now, but these programs are NOT intelligent, they're just following instructions that get modified through an algorithm through whatever stuff they've showed it over the decades.
(In Google's case, probably through the results of those image-identifying captchas they stick everywhere, so technically we've all been working for Google on this for free, and therefore deserve a percentage of any profits they make from it! If you're reading this Google, call me. You probably have my number on record somewhere, obtained without my express permission along with a huge amount of other data about me!)
But yeah, really this is just "cloud" storage all over again. Marketing buzzwords. As impressive as these things are at processing text into images, they're NOT intelligent.
(It is pretty well modelled on the behaviour of the majority of humans though. "Just take in whatever we show you without questioning it, and do what we tell you to do, also without questioning it!")
When they come up with one that can actually think for itself and interpret imagery and meanings like a human artist can, then let me know and I'll be suitably impressed. Until then, I'll be over there watching my favourite artists as usual! :)
What we really need is the ability to add just a bit of help. Things I'm surprised AI hasn't given us more of like, "here is a static standing pose and reference for my character, give me some animated versions I could drop is as character assets in a video game." (Even nice tweens from key poses would be a major help for making animations)
Some other useful features would be, "Here is a color blocked version of this image, I'm adding point, and directional lights like this, automatically determine the normal maps of the image and emulate shading according to my style (soft, cell... ect)."
Or even, "Here are a ton of refs of my character, I've drawn a new line art of them. Please color block them in so I don't have to."
Mhm, I think for 2D this already exists as FPS enhancer. o:
AI has turned some historical video footage into fluid high quality vids.
It's the same between you vs AI generated content. People still gonna want to have something YOU made for them.
It's true that trades become obsolete with the advancement of technology, but when it's artistic content I see more often that the artists integrate the techonoly into their trade instead of being replaced by it. Right now a lot of artists thrive on painting over, building imagery with pieces of other imagery, digital tablets, automated tools, all helping make cool work faster and easier, but still is the name of the artist doing it behind its value.
I think all the technology will adapt around what we have more than replace it. I think it's other kinds of trades the ones that get fully replaced by tech. Like driving trucks, there's no cretive side to it, there's no art of being a driver... There's a lot of skill involved for sure, but without having artistic value to it, it WILL be replaced by automation, and rather soon.
On the topic of AI, it's a fad right now. It's a hot new technology and everyone is racing to make the biggest and best one. We don't know what the implications are yet for AI but when it become applicable, people will be ready to do the job. It could come in the form of a AI card for computers, similar to graphics cards.
So much this, yes, exactly. I feel like a lot of these "unimpressed" folk aren't seeing the writing on the wall. This technology is only just coming out of its nascent phases and I'm seeing it put out work that looks better than some of the artwork my intermediate students make. Hell, I've seen a few pieces that made me feel a little inadequate. XD The progress Dalle in particular has made in the span of a year is staggering. It went from looking like "baby's first attempt image bashing" to solidly competing with intermediate art students. It's staggering.
"I was theorizing there will be a lot of artists that work more like AI-mancers and use their art skills to paint over or direct ai-generated photo bashes."
I would LEAP at the opportunity. I've already tried this out with WOMBO and Artbreeder. It's not my preferred method of working but I do enjoy it and I'd sell my left foot to have a chance to try this out with Imagen or Dalle. I sometimes use WOMBO to generate color pallets and I've made a few icons in Artbreeder. It's actually kinda fun and weirdly cathartic.
"will our society do to make sure people who literally no longer have a use to society's means of production can still make a living and enjoy life?"
I don't have much hope there, sadly. I may be biased as I'm basically watching my partner die a slow, perfectly preventable death after she was denied access to life-saving medical care because she couldn't afford to pay 700$/mo for health insurance. It seems to me like we're living in a Social Darwinist oligarchy and there's very little hope of expecting any compassionate outreach on that front.
I never was good at utilising my potential or being well adapted… I want to stay relevant and change if I need to - but lack the sense for the right direction.
I sent a request to midjourney and am waiting for their permission to use their tool. Humiliating might be the wrong word, but feels odd to do this under the prospect of them taking something from us unasked.
AI can be an option for commissioners (especially in 5-10 years), but i think artists always will be in demand <3
Here are 5 of my favorites. ^u^
https://images.nightcafe.studio/jobs/mGMRQCmIMJZuTMvtHjeq/mGMRQCmIMJZuTMvtHjeq_7.8125x.jpg?tr=w-1600,c-at_max
https://images.nightcafe.studio/jobs/SlOxf6zsRFBSddZLqg40/SlOxf6zsRFBSddZLqg40.jpg?tr=w-1600,c-at_max
https://images.nightcafe.studio/jobs/LOEvAbjpjGn2lTHUzS0j/LOEvAbjpjGn2lTHUzS0j.jpg?tr=w-1600,c-at_max
https://images.nightcafe.studio/jobs/LvG7EbOZwPCcPjQTOzem/LvG7EbOZwPCcPjQTOzem_2x.jpg?tr=w-1600,c-at_max
https://images.nightcafe.studio/jobs/vSNByGtF0QoUK31qIroe/vSNByGtF0QoUK31qIroe_2x.jpg?tr=w-1600,c-at_max
As for the prompts, most of the time I just type in a title I'd like to see produced, and see what Nightcafe does with it. For example, that first Image is called "Abandoned Outpost In A Forgotten Galaxy." Used a photo of a collapsing barn as a starting image and Nightcafe gave me back that amazing Borderlands Hut lol.
Give it a look if you want. It's free to make an account, they give you some starting credits, and as I said before, you get a free 5 credit top up just for checking in every day.
https://creator.nightcafe.studio/
I thought I was totally gonna blow my 16 credits on these, but they came out better than I thought they would lol. Results WILL vary, but I liked these ones.
Synthwave Apocalypse Ruins over a Dying Landscape
https://images.nightcafe.studio/jobs/vqxpFvOTtPxaDWXcZ8tZ/vqxpFvOTtPxaDWXcZ8tZ_4x.jpg?tr=w-1600,c-at_max
A Noble Phantom patrols the Halls of an Abandoned Hospital
https://images.nightcafe.studio/jobs/ZmmCZMNZe6coqePyPhLo/ZmmCZMNZe6coqePyPhLo_2x.jpg?tr=w-1600,c-at_max
I am so sorry you had to see that lol.
I could totally see that hanging in some grandma's kitchen. 😂
"What is it?"
"I dunno, my grandson gave it to me."
"Oh, how sweet!"
On the other hand, reputation is still going to be a key element - "made by Silverone", for example, may be a sign of pride for some commissioners, even if one of these programs can try to emulate your style. (Incidentally, I suspect that incorporating artists' designs into the input sets of these programs may become a point of copyright/trademark/legal contention in the future.) Artists who just want to make art as an end likely won't be as affected - or they may apply your concept of 'have an AI generate a base concept set and work from that'.
Physical art likely won't be affected until someone can create a painting robot (not a printer, something that splats paint with a brush or carves masterpieces out of marble). Even then, I wonder if art galleries will want procedurally generated pieces, or those designed and produced by someone they can talk to about the work.
What's really going to cause a stir in society will be the performance of high-level management duties by an AI. Imagine how big business owners would take it if their 'skillsets' could be replaced, with the multimillion-dollar paycheck cut, by a good learning algorithm.
It's really bad with words and letters. Like, to a hilarious extent. Someone typed in something like "Logo for a company run by monkeys that makes hats" and while the imagery was great, the wording was hilarious. It was spitting out stuff like "Mokey Mokes, Makey Makes, Monke Horts" and other assorted silliness.
There has also been at least one noticeable "break point" in every Dalle image I've seen where it's clear Dalle didn't know how to handle something. For example, it's oddly bad with fingers. Sometimes, areas will be blurry or "swimmy." I saw a picture of a bird standing on a rock and the rock had hints of the same downy feathers on the bird. Stuff like that.
It's very hit and miss. The hits are impressive and the misses are often kinda funny. Also worth pointing out that OpenAI HEAVILY limits what you're allowed to put out. Absolutely no gross, horror, or shock content, no real people, and no nudity.
https://www.reddit.com/r/dalle2/com.....eating_a_kiwi/
This disproves the assertion that a computer will never understand the joke, "time flies like an arrow, but fruit flies like a banana." It's able to interpret the correct meaning from context.
At least there will always be a market for traditional media. In a world of mass auto-produced everything, people would pay extra cash to own something physical and hand-crafted. Being an artist would be similar to knowing how to play a piano. Sure, everyone can just play an MP3, but it's a whole different situation when a living person is making the music. Quality versus quantity. Hand-made versus mass produced...
In a post-truth world, in order to fight deepfakes we'd need to appoint blockchains as reality/truth arbitrators. I'd suggest you pick a book about distributed computing and blockchain technology. We ought to learn how to fight fire (auto-generated 4K photo-realistic deepfake videos) with fire (advanced video hashing and time-stamping of source material).
Interesting times are ahead. It's entirely possible that the upcoming industrial revolution will create more jobs than it destroyed.
Well it looks like a 13 year old's attempt. It's young. It's still learning.
I'd suggest that you put yourself on Dall-E 2's waiting list and give it a try? (an save to folder, the iterations of your request.) See how it helps or not. They had a very interesting discussion of that on the Corridor Digital Podcast.
https://youtu.be/bqN_O1dPmNo?t=874
I think those working in animation and 3d Models still have some time., but for the most part I see this Ai being used for "concepting", and then refined by more skilled hands to produce finished artwork. There are little things, based on art director decisions, that probably don't really translate to clear language descriptions.
Ooh! Always down for some Ian Hubert. -enwatchenates-
Try something like it for yourself.
https://labs.openai.com/s/fhQvEO2q9aM28yG1pTJxAm37
Sadly, it doesn't seem to milk l like multi-species pictures...
The "network" it builds on is still 100% dependent on what goes in; it doesn't actually "create". Now while one can argue that any artist's brain might often be doing the exact same thing, taking elements from the past to create something new by combining them, I don't think it's the same, or even at the same level.
What I think might be happening is that low-threshold "wall art" might become a thing where this kind of "AI" will take jobs away, e.g. stuff you'd see on office walls as fillers or what not, but ultimately, those aren't the jobs artists tend to enjoy doing in the first place.
Anyway just my 2¢
Artists shouldn't fear automatic art generation any more than translators fear automatic machine translation. Did it hurt the profession? Yes for all casual use. But did it help the profession? Also yes, it's a great helping tool.
I think you nailed it with the "adapt or die" part. If you adapt, this will help you thrive.
Fake data is something we're not too fond of, and it generally shows in the end product as well. I know this isn't exactly on-topic with the journal but, it's my experience with AI and I view it as a powerful yet... interesting tool.
Here's some of my work, no fake data there!
https://cdn.discordapp.com/attachme.....69/unknown.png
https://cdn.discordapp.com/attachme.....76/unknown.png
https://cdn.discordapp.com/attachme.....60/unknown.png
And thank you! This is a really interesting use case for AI and I’m sure there’s tons of other weird niche hobbies or applications that it helps a lot with.
https://www.sciencedirect.com/topics/earth-and-planetary-sciences/ionized-gas#:~:text=Plasma%20is%20an%20ionized%20gas%20and%20also%20called%20the%20fourth,molecules%20of%20the%20bulk%20gas.
This should answer your question though.
I would absolutely welcome help from AI for my artwork.
I'm not entirely concerned about people getting their hands on it and doing shitty things because unfortunately they already do that with Deepfake tech which is improving rapidly.
IIRC, GPT-3 (also done by OpenAI) was supposed to be restricted as well, but people got their hands on it anyway.
Aaaa but AI is so fucking cool. I'm absolutely ready to adapt instead of die hahaha.
my personal prediction on the next phase of cell phone - it will begin talking to you as a person. In short order, it will understand you completely and become your 'best friend'. Society will then go full introvert, needing no other human contact, except, perhaps, for sex.
V.
Nope, they got that covered too! ;)
https://www.themarysue.com/wp-conte.....AD-480x480.jpg
(Probably NSFW link!)
and this: https://www.cnn.com/2018/12/28/heal.....ntl/index.html
Is it insanity, or the way of the future?
V.
*blushes* oops, did I say that out loud? :o
V.
Really, I have problems with these kinds of machines. The inventors cradle them like their "babies", I just see a machine. I broke a (early) conversation bot using normal English you'd find in a dictionary in normal full sentences, steered it down a bad logic pathway, until I had replies like, "William Shatner is a tomato."
There's a reason these AIs are being treated like precious "babies" at this time. They don't - DARE - let someone who sees it as a machine rather than magic give it inputs. It might break.
...I have no doubt all this is being trained as cheaply as possible, by people who don't really know what they're doing. I mean, for example, the first trainer of Google's self-driving car was a programmer who HATED driving (which is why he made it) and hadn't left the office in months. That's where it's core "personality" came from - apparently it like to make wild swerving turns without warning, that made all the passengers scream, thinking they were going to crash and die. They fixed it by... hard-coding it NOT to do that. ... but like a dog trained not to bark, secretly, it still wants to.
Gods know how many fixes like that are lurking beneath the surface of the new AIs.
Still to see any of them that cram in so much detail in so photorealistic a way as you do...
If I need to commission some artings, I'm still thinking of you first!
Timing is everything. This is my prompt post for this week... only three pages but I think it will make you smile.
https://www.furaffinity.net/view/47381467/
Mind you, I have written an entire novel where AI was the bad guy (comedy/drama/parody) - he called himself God.
V.
It's all in who and what you are willing to trust.
V.
and such (Walter Benjamin). Artists use any media to make their art, I see it as another extension that we can use to express ourselves. I really would like to try this new tech so much, it sounds really fun. And people make shit with or without it lol.
For one thing, while it may give you an image based on a certain set of specifications, you give those same specifications to a hundred different artists, and you'll get a hundred different interpretations, and it will be up to the professional artist to prove that their particular interpretation is worth paying for. There is a certain similarity of style to all the examples, and you can be sure that like every other cool visual trick, if it gets used, it will get overused, and it won't be long before get so tired of looking at yet another generic AI generated image that they'll leap at the chance for something different.
There is also the issue of status. In our fandom, there is a certain cachet associated with having a real artist bringing your vision to life, with some artists having more cachet than others. Artchod McPopufur who *never* does commissions, paints your blue foxgirl, and people will oooo and ahhh. ArtBot 5000 generated your blue foxgirl? Yawn......
So I'm going to say that there's no need to pack up your art tools. The machine hasn't beaten us yet.
We don't have the basic minimum income, health care, or human rights that gave people the luxury of attitudes like that in Star Trek. Just a capitalist value of the end product and hatred of the fact someone had to be paid to make it.
It's certainly very powerful. The icon I'm using was just a fursuit head photo; removed the face with content-aware; gave it a new face using a bovine pelvic bone and an image of a cat. Barely touched it up, and then used AI to animate it. It looks bloody 3d.
It may not produce as nice results as an artist, but it's vastly more efficient - and the industry is fixated on efficiency. Concept artists wouldn't photobash or use textures nearly as frequently if that weren't the case; so if I had to bet, I'd say they'll be the first to feel the heat. When AI starts churning out lewd furries, this fandom is fucked. Many 2d artists here will probably adapt by switching to 3d (so you're probably well ahead of the curve), or things like building fursuits (physical labor...)
Someone used an AI program to generate a music video using the lyrics as image prompts.
https://youtu.be/0fDJXmqdN-A
With those new AI developments, the huge difference is that they've made a system that is much more independent and can understand simple text cues. Is it useful for me though? Not a whole lot. I like fine-grained control over the details of the image content, so such independent image maker taking its own decisions is pretty much opposite of how I want things to be.
Additionally, I've spent more time than it's healthy on thisfursonadoesntexist.com and I've seen my fair share of eldritch horror that AI can create, alongside with actually cool examples. In fact, I've borrowed some of these better ones for my own 3D sculpting studies. And this is pretty baffling to me, how do these new systems avoid the gruesome-hilarious fails that are so common with simpler AI. Can some curation be involved? I wouldn't be surprised, especially considering their highly restricted access.
And here we get to another point. If these systems get locked behind a paywall and/or with heavy restriction of possible content, there will always be work left for human artists. Especially that a lot in art is based on exclusivity and limited availability. If I want a commission from Silverone, then I'd go for a commission from Silverone, and computer-generated fakes won't cut it, however technically great they may be. They may be entertaining for a while but ultimately not the "real thing". Rather a toy, fun to play with but whose outcomes aren't to be valued too highly, precisely because they can be churned out by the thousand.
This way, the biggest cuts that AI can make in the customer base can be expected in the low-price crowd, folks who don't really care from where they get their desired imagery or what the details are. Those willing to pay and knowing exactly what they want can be expected to stick around their usual suppliers of art. And if the AI's paywall is too high, then even the low-price sector won't be so affected.
And then, we get to the can of worms called the furry fandom. Its favorite contents are notoriously corporate-unfriendly and this fact may well provide it with a lot of resilience against AI automatization. Even with thisfursonadoesntexist, the input sample was supposedly kept SFW but certain AI outcomes I've seen show that NSFW managed to sneak in regardless. So yeah, I kind of want to see a furry pr0n-generating AI, just for the sheer unbridled absurd hilarity of what would ensue...
To attempt summing up this incoherent train of thought: I don't think these systems will blow digital artists out of the water (after all, photography didn't remove visual art as such, and digital art didn't remove the traditional). Some culling may be involved but I'd expect it to happen in the low-price areas the most. Otherwise, such tools might actually come in handy as instant generators of desired photobash and (pseudo-)reference. Not much different from digging through Google image search, only this time much more smart and efficient.
And should the AI relieve me of the necessity to have a job - all the better. Depending on the scenario, I'll finally get the necessary motivation to go and live in the forest, or will have all the time in the world to focus on art and other fun things. Nothing for me to worry about.
Almost certainly. Most of these AIs are already locked behind paywalls or they have aspects locked behind paywalls. Dalle doesn't but it's only accessible to those with a big enough Twitter/YT following and eventually, they've made it clear they intend to paywall it.
Personally, I'm not afraid of it so much, I just really want it. I'd use it to generate textures for 3D assets, to ideate and iterate, to create collage pieces to work into paintings, to make my own image-refs, to interface with clients who don't have refs or can't accurately explain what they want.... the list goes on. Alas, the only AIs I have access to are kinda garbage. WOMBO can make pretty colors. Nightcafe is hit and miss and is half-behind a paywall. Disco Diffusion looks like a PITA and it's also half-behind a paywall. ehck
"I don't think these systems will blow digital artists out of the water (after all, photography didn't remove visual art as such, and digital art didn't remove the traditional)."
Maybe not yet but someday? Probably. Look at Dalle 1 versus Dalle 2 and how far it came in only the span of a year. Now imagine where it will be in five or ten years.
"And should the AI relieve me of the necessity to have a job - all the better. Depending on the scenario, I'll finally get the necessary motivation to go and live in the forest, or will have all the time in the world to focus on art and other fun things. Nothing for me to worry about."
Man, I wish I was in that position. Art is my job. I've been doing art for so long, I don't know if I can do anything else.
The new iterations of this are quite impressive. I tend to see them as tools, and I'm eager to get my hands on them. I'm assuming that in a year or two there will be something pretty close commonly available. I can't wait to see how I can tweak, push and abuse this.
As far as this technology being a danger to artists or their livelihood, it may well be used for some things, but it's just as likely to be used productively by artists themselves.
Politically and economically if it DOES end up having negative effects on the livelihood of artists it will just be yet another symptom of the ongoing failure of capitalism. Since at least the 70s wages haven't kept up, and money keeps being sucked upwards to fewer and fewer insanely wealthy men. Productivity and profits have risen, but our quality of life hasn't. People should be sharing in this. We should be getting paid more and working less as this happens, and we haven't been.
I might even consider paying something, if it isn't too much.
BTW, I started screwing around with DALL-E mini last night and had a blast with it. I should be posting some of the results this Wednesday.
https://d.furaffinity.net/art/sikfo....._edit_comp.png (More in my scraps)
I'm finding a particular delight in the anarchic imaginary potential of this technology at least particularly because of its common availability as a tool. I'm always especially interested in probing the boundaries of such things.
You've poo-pooed DALL-E mini, but I think you actually might find some inspiration there, even if it doesn't have the depth, breadth and resolution of some of its more capable cousins.
I'll bet that as an inspirational tool, at least because of how it has the capability of surprising you, and even as a source of potential novel textures and such, you might find it interesting.
In the weeks to come (Usually posted Wednesdays) you'll find some of my further experiments, some of which you might find amusing.
It's no where near good enough to replace the mind's eye I lack. Dalle and Imagen come pretty damn close but they aren't freely available yet. There's also the matter of the Dalle-mini's servers chronically throwing down that "sorry, too busy" message.
I'm glad you're able to find some fun it it. I'm not saying it's bad. It's just not going to be really useful for me. What I really need is Dalle, Imagen, or a mind's eye like everyone else has. :\
I've also found the inability to generate satisfactory or even interesting results from some prompts frustrating. I discard a lot of results as being unsatisfactory or uninteresting.
I've found the "Sorry too busy" thing to be only a small inconvenience. My prompts usually go right through, and when delayed, they only take one to a few tries in a pretty short period of time to get through. For a free tool of its quality and novelty, it's a minor inconvenience.
Our purposes are at least sometimes different. One of the things I enjoy is looking for the broken results, and probing the capabilities and limits of these sorts of systems. I'm as often pleasantly surprised as disappointed.
If the AI gets your character though? You are in a world of happiness if you have the setup for it XD. (My SL character is basically "Cute anime boy with messy brown hair and brown eyes", I can fill my entire MyDocuments folder with pictures of any adventure I want with him and then my mind tries to process all that information when I go to sleep - I did have to give up on him wearing yellow T-Shirts though).
Alas, no access, and the competitors are lightyears behind.
I do think competitors will be fast on their heals - and might beat them. Their data set isn't all that unique, but just grabbing the entire internet and dropping it into a dataset. Open AI also seems keen on avoiding all the hot button issues in AI, so I suspect they either leave out or cut training data with more risque elements - which ignores a huge amount of art on the web.
The biggest thing is having the GPU compute power available to train them. On the flip side, I get the impression, that, if you're skilled with AI, you can actually add your own art to the training data without requiring a refrigerator sized computer filled with 3090 TIs. If so, that's super cool. Open AI does seem to be able to modify or create variations of pictures, too. I've also seen game developers using it to generate tiling textures on demand, and it would be super cool if it could learn to create 3D models.
The content explosion. Endless content for creative play for EVERYONE :D. It does look as exciting as the internet, though, and I'm jealous they're hoarding it all for themselves so far.
I get that a lot. I can't see imagery in my mind but I can still conceptualize and ideate. Like, when I daydream, it's like there's no TV screen and I'm experiencing the events through some internal sense of proprioception. Like, I can't see my character standing at the steps of the West American Capitol building, intimidated, and nervous as soldiers ascend the stairs around him into the vast city-castle but the information is there in some way I can't really explain. Maybe it's more like metadata. The TV screen in my head is not working but I'm still aware of the information in some way that's really difficult to describe.
I wish I had a mind's eye. It feels like it must be a superpower that everyone else has except for us. I recently found out there is a whole suite of other related senses I also lack. I don't hear music in my head, can't relive experiences from a first person perspective, and I can't taste foods just by imagination.
>The content explosion. Endless content for creative play for EVERYONE :D. It does look as exciting as the internet, though, and I'm jealous they're hoarding it all for themselves so far.
Same here. >;\ I want it. Actually, to take that further, I feel like it's kind of selfish of them. Their AI is standing on the shoulders of every artist, photographer, and content creator who uploads their work to the internet yet we're not allowed to make use of it? Hm...
On one hand, as someone with aphantasia, it's literally a dream come true. AI-generated imagery is the closest I will ever come to having a mind's eye excepting some medical advancement possibly over the horizon. AI will revolutionize industries and rewrite the face of the world.
On the other hand, someday, this will put artists out of work. For now, I think it will be an amazing tool but make no mistake, sometime within the next decade or so, this technology will come for my job. And no, this isn't like the invention of the camera. AI just straight up does the literal actual work of an artist, orders of magnitude faster, and in many cases, better.
I can't speak for any other artist but a lot of my identity was tied up in being an artist. It made me feel a little special. I had something I could be proud of. Now, my entire body of work can be replicated by a machine, easily, far more quickly, and probably higher quality too.
But all that said, I'm tentatively optimistic. I feel like AI will be an amazing creative tool. I want to believe it will push the bounds of creativity further than we can currently imagine. Until we blow ourselves up or climate change turns the whole planet into Mad Max.
Maybe I'm not quite so optimistic after all. ;P
The AI then tries again. You still "write" the story, because there is an image you have in your head, but when it happens, the image is awesome, but not quite the same as what you wanted. It's like both the artist and the ultimate user ends up "discovering" that imaginary space for the first time. It's magic and fun. Like a collaborator that lets you change anything you want.
I think for indie devs, this is a god-send. The sheer amount of talent it takes to make an indie game makes that career path impossible. The time sink to produce and release your own custom video game on steam might jump from years of work on the side to... months or weeks. It becomes in reach for so many people and I'm excited to see what happens when the massive barrier of content creation that is indie game development is no longer a thing.
But yar, it's hitting everything and at this point, I can't help but imagine that the whole thing is going to accelerate. I will soon be obsolete.
https://huggingface.co/spaces/stabi.....able-diffusion
It's got a bit of a wait time, but it can be fun. I've had a number of times where totally G-Rated descriptions got me unsafe content warnings. Which is hilarious. Either the model is overly protective or it really likes to make kinky stuff and then doesn't want to share XD.
https://discord.com/invite/novelai
Go jump on their page while it's still around go play a bit and see what's going on, it's absolutely bonkers O_O