I won't pretend that this isn't a troubling development for digital artists, maybe even existentially so. I hope not.
One thing that makes me a little hopeful is that every image I've generated with DALL-E 2, even the best ones, would require non-trivial work to make them "good".
There's always something wrong, and you can't tell the model "the hat should be tilted about 5 about degrees", or "the hands should not look like ghoulish pretzels, thanks".
There's also this fundamental limitation that the model can give you a thing that fits some criteria, but it has no concept of the relationships between elements in a composition, or why things are the way they are. It's never exactly right.
It's like the model gets you the first 90%, and then you need a trained painter to get the second 90%.
But yeah, it will certainly devalue the craft, don't get me wrong. And anyone who is callously making comparisons to buggy whip manufacturers should consider how it would (excuse me, will) feel when AI code generators pivot to being more than a copilot, and suddenly the development team at your office is a lot smaller than it used to be, and maybe you aren't on it anymore.
If you spend a lifetime mastering some skill, and then it's just not valued anymore, it sucks, and you get pretty mad about it.
> It's like the model gets you the first 90%, and then you need a trained painter to get the second 90%.
Call me a doomer, but I think this makes the possible consequences even worse.
Remember the 80/20 rule.
A lot of modern product innovation is not really about improving quality - rather, its about introducing lower-quality versions of existing products which are significantly cheaper than the original but still "good enough".
Dalle2 and friends could fall into the same bucket. If they produce artwork that is objectively worse than a human-painted version would be, but still "good enough" for many mundane usecases - stock photos, concept art, etc - we might still see a wide adoption and displacement of human artists from those usecases - along with an overall drop in quality of artworks.
Ugh... I somehow hadn't yet even considered the part where we all have to tolerate almost every single image we see during the day being generated by some creepy AI model; but OF COURSE that's how this is going to play out :( :(. I mean, mant of the products I purchase on Amazon don't even spell check their product marketing images as it stands...
>but still "good enough" for many mundane usecases - stock photos, concept art, etc - we might still see a wide adoption and displacement of human artists
you're not afraid of DALL-E, you're afraid of an army of fiverr workers stealing your job. Stock photos and low quality art have already been commodified. Very few people go and commission bespoke stock art from the individual working artist, they get a subscription from one of the gazillion content stock photo factories for a few cents.
> If they produce artwork that is objectively worse than a human-painted version would be, but still "good enough" for many mundane usecases - stock photos, concept art, etc - we might still see a wide adoption and displacement of human artists from those usecases - along with an overall drop in quality of artworks.
If people are happy with "good enough" they generally don't hire a digital artist in the first place (the whole reason DALL-E can exist is because there's a lot of digital imagery on relevant subjects/objects available to it to train, and there's even more an internet search away) or if they do, they get one off Fiverr.
For mocking up quick concepts, that might be different, but that's a workflow improvement.
> If you spend a lifetime mastering some skill, and then it's just not valued anymore, it sucks, and you get pretty mad about it.
That is absolutely not what the OP is complaining about. They're not saying that because AI is good, they won't find work. They are complaining that in training AI for art generation, builders took works from living artists, without consent from them, and that in so doing allowed generators to make new art in the style of said artists.
The example given is that Stable Diffusion even tries to reproduce logos/signatures of living artists.
If I produced a rubbish search engine that bore a malformed "gigggle" logo using Google colors, how long do you thing I would survive before being sued out of existence by an army of Google lawyers?
But that's exactly what many AI generators are doing here.
Edit: the first version of this comment confused Stable Diffusion with OpenAI, and stated that OpenAI was owned by Google. OpenAI has a strong partnership with Microsoft. Stable Diffusion is not OpenAI. Sorry for the errors.
This is frankly what humans have always done, learning and taking inspiration from other artists. Now we have made a machine that can do the same thing.
In the case of exact reproductions, we have copyright and IP laws.
You're right, AI generated pictures should not be copyrighted, as is the case today. People should be free to mix and remix pictures via AI as much as they desire.
This is where I imagine things are going to get into trouble because how are you going to determine what is AI and what isn't? Especially when Stable Diffusion is directly classifying artist and cloning their signatures and watermarks. What about things that are started with AI and refined by human?
Exactly. The relevant law is with regard to the use of artwork on the part of the people who feed the index. Artists should have a say on whether their work gets included in a training set, if their works are not public domain.
It feels like a big stretch to consider an algorithm to be 'inspired'.
Where are the bits that correspond to 'inspiration'? Seems like that would answer a lot of big questions in philosophy.
I claim that claiming computer algorithms are inspired is a big stretch.
I claim humans can be inspired.
I don't claim to know how human inspiration happens, or if neurons have anything to do with it. (They may, but I make no claim). Not being able to describe the process by which human inspiration happens doesn't invalidate either of my claims.
If there is a satisfactory non-bit based explanation to how computer algorithms achieve inspiration, I would accept that to. We have the advantage with computers, that their activity is conveniently summarized by their programs which are represented in bits, so expecting an explanation in that form I think is reasonable.
The defense if the claim of human inspiration is (1) we have that word for the concept (2) we have thousands of years of thought, philosophy and literature giving support and definition to the concept.
Not only is your rant about Google misplaced considering Dall-E is OpenAI, not Google, but the thread is also not complaining about Dall-E. It's about Stable Diffusion (https://stability.ai/blog/stable-diffusion-announcement) which is explicitly trained on working artists. That's why it tries to reproduce the logo.
Oh, that just drives the point home. Any flaw you can find in these models to build your hope on will just get corrected in the next iteration, only a couple months down the line.
The shadow was wrong in the man-tree-beach-sea image. I guess it might be artistically wrong, but I'm sceptical of what that even means in this context.
Considering the staggering speed that image generation is improving, that 10% gap will only continue to close.
Starting e.g. an art education right now seems likely to be extremely nerve-wracking as your talents may very well be woefully obsolete by the time you graduate; the exception perhaps being those top-0.1% talents that will feed the models of the future with new material.
So an art training will become more like pro sports training where “success” means to be top 100 or so in the world. Note it does not prevent one to do arts (or sports) as a hobby. People didn’t stop playing chess after Kasparov lost to Deep Blue.
Given how quickly AI image generation, and creativity generally, has progressed, I think it's perfectly plausible that within ten years we will be able to tell an AI "create a work of art that is unique, highly meaningful and that would be very difficult or impossible for most humans to create with their hands," and will get a work of art that is, in blinded assessments, competitive with the work of any master.
If that happens, I agree that the top 100 human artists in the world will likely have jobs, but they won't be successful in the sense that their work is uniquely valued by society. We pay to see the very most talented humans perform tasks that have been successfully automated, such as chess and lifting heavy objects, not because we need the service they provide but because we get an emotional kick out of seeing other humans perform way outside the normal range of human abilities.
I’m willing to bet money it will never happen. People said the same about self-driving cars, but the initial razzle-dazzle blinds people to the actual dullness of the algorithms, and obscures their limitations. AI can only recombine what has already been created. It has no ability to imbue art with meaning or to push the medium forward.
What you describe can only happen with general intelligence, not these fancy neural nets. If anything, they will become powerful tools to help artists augment their creativity.
I am a working digital artist, my personal feed is filled with takes like this. I understand the crisis that artists go through currently and I understand their emotions. I've personally made peace with it already, but it will take time for everyone and there will be a lot of concern, rage and bargaining in the process.
All of us artists woke up in the world where the thing we were learning to produce all of our lives became 100x cheaper. I've been kinda waiting for this to happen for truck drivers or translators and was honestly surprised that it got to the artists first. But, there's nothing to do now.
There will be a lot of takes trying different ways to deny this reality. Like if we agree that this thing is immoral or anti-cultural it will go away and our potential customers will forget about it. I think that it's a waste of time
The industry is not completely gone, but it will go though a massive transformation. The basic hierarchy of artists will probably stay the same - people who were good in the pre-ai era will still be good working with the ai. If you've learned your basic skills right - they will still work.
I've personally already changed my workflow to use ai art for the sketches, ideas and references. And then polish them to commercial level using my old experience and skills. It's funny how this thing is so new, but I've gotten so used to it that it's already hard for me to come back to my old set of tools. And I do recommend all other artists to do the same and to try to find their place in the new reality - it really seems like there's no other way.
I believe that it already did, honestly.
The output and range of Dalle-2 and Stable Diffusion is unmatched by any human artist. I am not only talking about quantity and range - the best illustrations generated by Stable Diffusion are honestly unbelievable and could totally get on the front pages on artstation on their own. The digital paintings, the 3d renders, landscapes, portraits, concept arts, character designs - I've seen too many world class examples of those generated by AI. The best human artists can usually be world-class at a single style or a genre, while Stable Diffusion is world class at hundreds of them.
There's of course no way to judge art objectively. And there are at the moment a lot of gaps that AI is not the best at. Still, if there was a single human artist that could do the things that stable diffusion can do - he would be the best artist that ever lived, no doubt.
I feel that the illustrations are still at least somewhat in the uncanny valley. I don't think that the quality is already there. What you see is only a subset. Many creations are really bad, creepy and unbalanced.
I think this will change up to the extent that the majority of people won't care.
Wow, interesting. I’m an ML researcher creating models for music generation and I can say with certainty we are pretty far from generating human level quality music.
Depends what your benchmark is. As the AI stuff is getting better, things like MIDI and Autotune lower the bar to where it's understandable some might see the convergence coming sooner than later.
The bar is the same as with human music - if people pick it on Spotify without realizing it’s AI generated with zero human involvement then it’s good enough.
I think the bigger point is that while it might not be better, it could be "good enough" and 1000x faster/cheaper at the same time.
Instead of hiring 100 artists, you could have this 1 model generating images, and then 10 artists touching up AI's results to take them from 90% to 99% on quality.
I personally couldn’t care less about 99% quality. I want to see genius level work - of the same impact as Monet, Matisse, Picasso, Dali - something unlike anything else in the dataset.
"We are suffering from the ruinous competition of a rival who apparently works under conditions so far superior to our own for the production of light that he is flooding the domestic market with it at an incredibly low price; for the moment he appears, our sales cease, all the consumers turn to him, and a branch of French industry whose ramifications are innumerable is all at once reduced to complete stagnation. This rival, which is none other than the sun, is waging war on us so mercilessly we suspect he is being stirred up against us by perfidious Albion (excellent diplomacy nowadays!), particularly because he has for that haughty island a respect that he does not show for us
.
We ask you to be so good as to pass a law requiring the closing of all windows, dormers, skylights, inside and outside shutters, curtains, casements, bull's-eyes, deadlights, and blinds — in short, all openings, holes, chinks, and fissures through which the light of the sun is wont to enter houses, to the detriment of the fair industries with which, we are proud to say, we have endowed the country, a country that cannot, without betraying ingratitude, abandon us today to so unequal a combat."
Just to avoid confusion, the page & excerpts you cite are from an 1800's French economist named Bastiat, not from the legendary 1980's neo-expressionist painter Basquiat.
Since this thread is about art, I figured people might read the name Basquiat in your comment and think that he wrote the quote you cite, which would in my mind give it greater relevance to this discussion than it has.
Digital art is not permanently interesting. My sympathies go out to people who feel like they will lose some of their business. However, a group of developers has shown what was always true about digital art: its computational art even if you draw it because it is encoded digitally. In other words, it was always going to be reproducible and remixable.
The problem with being shocked by Dall-E, in my view, is that it shows an ignorance about the historical development of art and its incredible diversity of practice + the final productions and forms of art. OpenAI have sort of Warholised digital art in a way and that's just very standard in art history. People went crazy when Warhol productised art but in reality this was an overreaction and plenty more stuff came after that which completely different in its orientation towards art (e.g. something like Hans Haacke). Dalle-E is a system for producing digital art in the way that Warhol's practice was a system for producing visual art as a commercial product.
Indeed, I'll get worried when DALL-E gets sick of everything it has seen, and comes up with something that's new, and maybe not even very good at first, but perseveres until it's recognized as important. All, while earning its living in some soul sucking day job.
I'd actually love to see DALL-E take on the greeting card industry. Now that would be fun.
I literally just made a greeting card with DALL-E. Generated the image I wanted and had the card printed and posted to the recepient for no more than buying a pre-exisiting card. Advantage here is that this image concept, which was very tailored to the person, didn't exist anywhere in any form.
I think you're wrong. Pretty much all human creations can be encoded digitally to a high enough fidelity that the discretization process no longer is distinguishable.
And 3D printers for paintings exist, that can replicate brush strokes and other techniques, and they will only get better over time.
So there's very little left in the human arsenal as the AI generation and AI painting techniques both improve.
Oke I see what you mean, but isn't that just a reproduction in physical form? They have Warholised painting with 3D printers by creating a systematic method to reproduce art using a new combination of craft. I like it but it's not a totality.
I can bring up a form of art that is not captured by this conceptual scheme. Carsten Holler's SOMA:
The exhibition Soma was installed at the Hamburger Bahnhof in Berlin in 2010. Its main element were 12 reindeer in two pens running the length of the former railway station. Half of the reindeer were fed the fly agaric mushrooms in their food, which are part of their customary diet in the wild, and turn their urine into a hallucinogen. The reindeer urine was collected by handlers and then stored in on-site refrigerators for use. The experiment was extended to canaries, which were housed in two hanging cage pieces, to mice, and to flies. A mushroom-shaped Elevator Bed was installed in the middle of the space, and visitors could spend the night on the premise for a fee.
Do you see what I mean by this digital art vis a vis the collapse of all art being an overreaction? Carsten Holler is a real artist, the art is good, and its not reproducible in the way digi-physical stuff is. It's experiential art and presented in a gallery. Conceptually, its about as far away from DALLE-E as possible but its still art and not "captured". I think art is not over and DALL-E is not poison for artists.
The kind of art you're talking about is "high art" and it's a luxury item. The majority of artists don't produce that type of art, and will instead lose out in this scenario.
A Ferrari is a luxury item. The exhibition was free but you could pay to stay the night. The guy also designs slides and the tone of his exhibition style is not "high art".
Ultimately you are talking about art as a product and that means you think all art is captured by Warhol's system of art. Which is what all digital artists overreacting to this also think. It's not the totality of what art is. I like DALL-E. Good work, but now what else is there beyond digital art? Or by pushing digital art to some more extreme outcome?
Technical artists are far different from performance or exhibitionist artists. They aren't even in the same realm.
Yeah, art is vast. It includes me saying "Kerflaffle!!" while I ride a unicycle and fart out a candle in a dimly lit room. But that doesn't usually pay bills and most artists don't do this type of art. It serves a different audience/consumer. It has a different market cap. Let's not confuse oranges and grapefruit.
Your link shows you can 3D print a "Rembrandt" if you hire a team of people to spend 18 months working on custom algorithms for it, and I doubt any art experts were as likely to actually be fooled by it as they were by a van Meergeren forgery.
Not sure the people selling paintings knocked out in an afternoon to tourists who visited their studio and chose to buy an original instead of the visually indistinguishable print for a fraction of the price have much to worry about...
Everything is originally bespoke before it is mass produced. Forgeries have gotten so much better than they used to be. I know antique dealers and they often rant about the job of recognizing reproductions having become significantly more difficult.
Like others have said, the first ones have issues, but late stage is when things get problematic.
I'm not sure what you mean by this. Reproducibility has virtually nothing to do with interest in my opinion. I was an admirer of plenty of artists before I'd ever seen their work in person. Perhaps I am missing the point.
I mean it in the sense that using digital technologies to produce art is not always going to be interesting. I don't mean that I will never find any digital art interesting though. Distilled to one point: DALL-E does not wholesale capture any rational definition of art reflecting actual art history.
I think DALL-E is a win-win by pushing digital art to some further extreme and forcing people who are open to art beyond the digital to try new things. Digital art had reached a point of maximum inertia before this imo.
Do you seriously want to go down the 'This is not art' path? That battle has been lost so many times for different forms of art that are much harder to 'get' for most people.
Also, since you like to emphasize the 'digital' part of these works, don't forget we're probably not much further than one Master-level project for some mechanical engineering students from having AI-generated physical paintings.
Ah, I see what you mean, as in digital creation is just another in a long line of technologies used for art creation, and at some point will just be mundane? I agree completely with that.
Yeah and it'll evolve unexpectedly but not in a way that most artists can't adapt to or develop a practice with. I just don't agree with the idea that DALL-E is art poison. That's overly simplistic and reactionary and not realistic about the diversity of form, materials, composition in artworks.
A lot of art created now, especially for film/ media, already goes through a similar process that AI just automates and accelerates. Several big concept artists I have met have lamented on the fact that many artists in the industry basically churn compositions out through mashing references they gather on google images or asset stores, not to mention just straight photobashing, which has made a lot of the concepts produced fairly derivative and homogenized since everything is just referencing what google images serves or assets that everyone else uses.
Even fine artists have likely had their practice influenced by the whims of what the algorithms on major social media outlets are willing to favor in order to get engagement. It kind of feels like many creatives have been incentivized into becoming slaves to these processes, which has in turn made them seem replaceable with AI.
This sounds similar to complaints I've heard about music for film/TV. Often stock music is used as a placeholder during post-production, and then composers are told by producers to just rework the stock music instead of being given time to make something original. This leads to most shows having very "samey" soundtracks.
strong concur, evaluating the creative processes requires acknowledging influences, and if pop culture referrence go beyond inspiration and become a core influence in style then the art aspect becomes industrializable, and in this case replaceable by AI.
Transfer learning to emulate artistic style [0] has been around for the better part of a decade and has had zero impact on artists’ livelihoods.
People consume art because they enjoy admiring the human talent that creates it, celebrating that some individuals are capable of extraordinary feats the vast majority of people are incapable of. It’s the same reason people watch sports—they enjoy admiring the top echelon of human physical ability. Very few people would watch Olympic Games performed by realistic androids.
I do agree that tools like this could eliminate mediocre graphic designers, or anyone else creating visual products that are so mundane that their viewers never bother to consider the artist. Corporate Memphis [1] designers’ days are numbered.
The danger of tools like this is eliminating those margins. This type of automation will eliminate junior graphic designer roles and add new requirements for the experienced graphic designers. Both of these affect the job market by making it harder to get started and stay. There will always be the high end and boutique jobs. The picassos and rembrandts will continue to find work.
The starving artist who sells out to create ads or create content for commercial entities will find those opportunities dried up
A single agency using this tool effectively could "in theory" produce 500 times the artistic output from a single artist. Vastly shrinking the market for a decent paying career.
Yes. The journey to master artist begins with finding a place to crank out a lot of apprentice/journeyman level art. You get a little bit better at some aspect of the craft with every piece you make; eventually this makes a very visible difference from your early work. Regular critique from your peers and mentors helps a lot too.
Rembrandt started out as an apprentice, with about four years starting around 13 before he opened his own studio. Picasso was trained by his art professor father from the age of seven in what sounds like a very traditionally hardcore fashion.
Problem is that the entry-level stuff is going to be done by the AI and fewer people are going to have the chance to go through the steps of becoming a master artist.
Not only that, but your average person is not able to asses the quality of an image. Just like the average person could not discern your code (or it’s quality)
Why would that be a problem? They can either pay an expert (if they want someone to assess the "quality", whatever that may be) or just be happy with their pretty picture.
It won't get rid of the bad ones, it'll hollow out the middle just like similar processes have everywhere else. There will be a few superstar concept artists making ridiculous amounts of money, a lump of often barely-competent copypasters/LLM-prompters doing it for $5, and no more space at all for people to have a career to comfortably grow and train their skills.
> Why not? a cyclist is a "realistic android" to a runner. It would just be different, more like F1.
But the human element is the main thing. This is the crucial part. Take us out of the picture and there's nothing very meaningful left for other humans to see.
Speaking as a professional digital artist, I am not very concerned by this. Images created by AI without comparable human experience (and therefore meaning), is not the same as art, although it can of course look like it. People who are more than casual appreciators of images follow and admire artists, and their translation of human meaning into an artwork.
I can see this eating into the commercial artwork market, which is bad news for the artists that rely on that for income, but it's not an existential threat to the meaning of art.
It's odd as a design manager, I feel like people focus on the wrong things to be alarmed about with DALL-E, and aren't alarmed enough about others.
* feels like there will be a subtle shift from (corporate) designers as creators to designers as technical sheepherders and retouchers. senior designers who have a real nose for art direction and understanding the fundementals will be fine. interns and juniors may get hit with a widening canyon of blindly using these tools without understanding why the "base" styles/logos look good. the act of creation is a learning experience that this removes.
* I feel like phi-y in a different comment has the most accurate read on the situation, given how this will not replace the superstar artists who've already gone through the years of training and sweat-shop work to refine their craft, but the beginner artists who "need" the sweatshop years to get a real-world sense of their craft and the money from the more easily reproducible works.
* there is a difference from using digital art as a tool (say, an apple pencil) to digital art as a resource-crutch (photobashing) when crunched for time or resources. I have mixed feelings about photo bashing as an art but those who rely on the resource-crutch sign may get hit hardest versus the ones that already have the raw illustration skills and can adapt/art-direct on the spot.
* copyright laws are gonna be interesting when people start trying to "base" results on logos like disney and coca-cola and get lawyers sent after them if it's too obvious. My fear is this creates yet another power imbalance between those who have the money to sic lawyers on others versus not.
* I'm mixed on DALL-E because on one hand and to be perfectly blunt, I feel like the most easily reproduced work is ugly/same-y as sin, and lacks all originality compared with the more creative/taboo-art circles who are the first ones to push stylistic and thematic boundaries. I'm not worried about DALL-E replacing styles that are inherently harder to reproduce.
On the other hand ... a sense of grace should be extended to artists who are affected. Being slowly replaced/treated as obsolete is a hard thing to experience, much like how natural age is rough on people's emotional states to the point where it's a type of trauma. It's messy, and it's hard, and I acknowledge there's going to be casualties.
Agree with most of what you're saying, but a bit perplexed by this
>interns and juniors may get hit with a widening canyon of blindly using these tools without understanding why the "base" styles/logos look good. the act of creation is a learning experience that this removes
Surely with digital art more than most professions the understanding of why the end product works is pretty orthogonal to the actual technical implementation details anyway? You don't understand composition or style or taste or brand identity or proportions by creating textures and shapes from scratch or knowing your way around a tool. To that extent, spending their days filtering mediocre compositions not created by them and fixing their flaws might give them a much keener sense of what isn't quite working than trying to judge mediocre compositions they've put a lot of effort into themselves objectively.
I can see bigger issues with the thought that might be comparatively dull work for a lot of people.
Hire a human and "pick any two: fast, cheap, or good".
Or hire the computer and you can have all three.
You're telling me I don't appreciate the genuine artifact? Well, tell me how realist painters survived film photography, and how film photographers survived digital photography. I'd ask but I don't know any anymore.
Subjectivity defines what is good. This thing makes art as reliably as 9 bucks gets you fries, a coke, and a burger. Talk about disruptive technology.
As a hobbyist artist my main concern is less the automation will replace artist as much as that 99% of that automation will come from the 3 companies with deep enough coffers to train and run the most complex AIs, and 2 of those companies will be FAANGs. Ultimately it will be a future where instead of paying an artist X amount for decent art with the artist getting a ~7% margin, it will be paying 90% of X to a monopolistic tech company for almost decent art with the company's CEO getting a 70% margin
AI generated art seems like a good way to open up solo game dev to people who can't draw. Just run the AI over and over again until you get artwork close to what you want.
There are already tools out there for people who want to automate the programming part of game dev. I don't see how that's any different in principle from an AI that generates art. This seems like a better solution to the problem of making a game without being any good at art than buying assets since your assets should be unique.
I sympathize with all artists who are concerned about these new tools. Once I thought about being a full time artist, but life got in the way.
However, visual art creates a cultural feedback loop that is constantly evolving, it is a crucial mirror that shows us stark naked. Just read the history of art.
People should not forget that there is no way this technology can make such a vital form of collective expression relevant without artists participating in it.
Of course it will disrupt a lot of things, and nobody enjoys that, but new ways of meaningfully involving humans in it will continue to emerge. Otherwise it would just cease to be relevant.
The linked reaction is probably slightly too alarmist but I am also still surprised that this was the field that some manager decided to devote millions of dollars of research and development funds to automate.
I've read in advice for technical founders that programmers tend to want to jump and automate the wrong problems, and this feels a bit like one of those cases in retrospect.
I think the tweet author is missing the meat and focusing on the gravy - they themselves indicate that the AI just replicates existing styles, so it will only ever produce existing art styles with the signature AI blurry weirdness on the details.
A main hallmark of breakthrough artists is inventing new styles and making them their own, a feat that AI has yet to perform very well (unless you include the aforementioned blurry telltales of generation and scrambled letters). Otherwise artists are just borrowing ideas from each other anyway so AI removes the need to learn someone else's style to replicate it.
In my opinion it will hurt the bottom tier of artists, but the ones who create new styles along with their own works will continue to thrive.
Let's someone creates an AI where you describe your program and it writes it using only existing code snippets. It might put 90% or more of software engineers out of work. Saying "but software engineers who create truly innovative systems, design patterns, languages and approaches will continue to thrive" is very limiting.
I fully grasp the concern here. It's not unlike what happened during industrialization, but on an intellectual level. I don't know what the solution is. Basic income has been floated, but I'm having a hard time evaluating all the arguments for and against.
Automated trucking, rail and ocean transportation sound ideal for the cargo industry, except for the millions of truck drivers. You put those people out of a job and you have a societal problem that needs to be solved. I'm not against progress in any way, but it comes with problems in need of a solution
Technology is supposed to put us out of jobs! But I'd argue it's been doing a pretty terrible job of it in the last few decades - we're all still working as much as we ever have been.
Personally I'd love to have a tool that meant I only needed to spend half as long working to be just as productive.
But it requires more than just technology to make that result feasible for the wider population.
Like I suppose many of us, I had a conversation with an artist friend about this very topic. They are concerned about it. Some of the generated results are impressive and surely can sufficiently replace some subset of the current human output.
But! Seeing DALL-E artwork from a bird’s eye view, I am more and more convinced that it’s just another tool, not a replacement: It is amazing good at producing unexpected results, and just like GPT3, it does so by regurgitating what is already out there, remixing it in somewhat interesting ways. This is an especially useful tool to quickly produce mood boards, or communicate an artistic intent. However it is not a replacement for the kind of thinking required to produce good artwork.
There is a response to this tweet where someone says "stop gatekeeping art" which while I understand the gist of what they mean, seems like a weird thing to say to console someone who is concerned they are about to be automated out of a job.
I guess the context is "as an art seller, I am concerned", not so much "as an artist". Which is understandable, of course, as much as anyone would be concerned if their job is in jeopardy.
But framing it as a matter of art does sound like gatekeeping. It'd be "that cannot be art, because x" when all they mean to say is "please don't stop paying me for this".
I don't believe they are framing it as if it is not art. The problem they are raising and that it is an AI trained on other peoples' art down to copying their signature making it appear as if it were a genuine piece from that artist and of course yes, they are concerned that they are watching their career die in real time because of it.
>t does not seem that they have any safeguards in place to protect artists
From what? Until strong AI comes about, good human artists will always be able to compete on originality and relatability, and even afterwards, human art could be valued for its "authentic flavor", or something like that. Until then, among humans, only the hacks need fear these systems.
I think my heuristic is Magic: the Gathering card artwork.
Can DALL-E make set defining breathtaking high res promo art, with fantastic composition, style and proportion? Maybe, but probably not as well as ex. Raymond Swanland right now.
Can DALL-E crank out the 50x lands and commons needed for the set? Almost certainly.
Superstar artists will be fine but I fear the bottom will fall out of the craft.
A friend and I experimented with using it to generate images for DnD items. It’s a quick way to get a visual for something that doesn’t exist, and therefore, does not need to be totally accurate.
I think the concern is fully justified. Those still being skeptical of AI art fail to understand exponential progress. Usage of various algorithms is about to explode, feeding back billions of signals to correct the model.
A second or third generation model may really understand relations between objects, depth, lighting, etc.
It's both amazing and scary. A little behind the curve, but expect similar developments in music generation.
Given the speed of all of this, the socio-economic discussion on AI should really get prioritized, as very little cognitive work is going to be safe.
I get the sense that the same thing is about to happen to many forms of creative work as happened to hand crafting during the industrial revolution, and if so the fallout may not be pretty.
People still buy hand crafted items of course, but they tend to be relegated to high-end or niche markets. The same would apply here. The larger number of people employed churning out relatively uncreative material would be displaced just like weavers and seamstresses were by textile mills.
“A girl in a red dress throwing a Molotov cocktail in a field of flowers in the style of Banksy”
“An obese politician smoking a cigar being held up by a crowd of starving people wearing American flag pants in the style of Norman Rockwell”
“A web page sign up form in the style of Uber”
Content violation or ??. Heck, any cartoon in the New Yorker would probably be a content violation.
The hard part, I think, is art making an “on the edge” statement. If they allowed this, it could easily - and would - be taken too far (depending on what your version of too far is).
I just don’t think human art and all the subtleties behind it can be algorithmically replaced. They weren’t algorithmically created.
Corporate soda ads… that’ll probably get replaced (but will still need layout and mixing)
UI design and video game art (not concept art in game art) seems too precise. And meaningful art seems to have too much in baked human judgments.
I am as afraid of this as I am of Github copilot. Meaning, not much.
DALL-E is super cool, but like AI becoming conscious, I am not holding my breath.
I'm watching an artist play with DALL-E for the past few days; it's proving to be a useful tool for inspiring concepts and leading her to her own visualizations and execution of things that she had not gotten to "gel" previously.
I can recall a little bit of the fuss at the beginning of digital art, "is that even real art?" I don't think this fuss has as much to stand on. It's a tool for making art; whether it's really art or not is a judgement only the individual viewer can make.
That's a good point. In a way DALL-E just shifts the burden of creativity on to how someone can come up with and idea and translate that into written language.
I’m still not sure where the line is. Human artists steal style all the time. Maybe they eventually evolve into their own unique style. But any great artist could definitely imitate many other great artists’ style.
I’m not sure where the line is for machines. Copying the signature is funny, but also fixable. Is copying the style fundamentally wrong? If yes, is it wrong when humans do as well?
I think it’s telling that the primary ethical concerns around AI development (at least at OpenAI, if not others) are that AI tools not be used to enable propagandists, or to generate violent or sexual imagery.
I see comparatively few ethical concerns about the economic fallout professional artists and illustrators will experience as a result.
I'm pretty unimpressed with the stuff coming out of Dall-E etc (I know, controversial opinion on HN, but bear with me). It's technically impressive, but imho, it still hasn't closed an essential gap need to make AI truly compelling, and still feels like just a modern ELIZA: simultaneously impressive given the technology and a parlor trick.
That said, I'm also not too impressed with many contemporary "artists", especially the type this person is afraid of being replaced with AI.
What I see here is that artists that have turned themselves into art making robots are vulnerable to being replaced by robots that are making art.
As a counter example I present the work of Ben Kovach [0], a generative artist, who makes art with robots essentially. All of his work is generated from computer programs but there are many (and certainly not all) pieces that capture a genuine sense of art that I have not seen in either the slew of human artists churning out robotic work to satisfy customer demand or current generation AI generated art.
Kovach's work reminds me of the work of Factory Record's designer Peter Saville, probably most famous for the cover of Joy Division's Unknown Pleasures [1]. In a sense Saville is pushing form the opposite direction of Kovach, trying to use human skills to create art that feels somehow robotic and sterile (whereas Kovach does the opposite, create something that feels more human than your typical machine generated art).
I don't see current generation a threat to art, but rather the trend towards the reduction of art towards being a reproducible, automatic process. That is the art that can be replaced with AI was already art created by an ad hoc "artifical" artist.
Ha, yes, I feel many artists have been hacking away at styles that look good pretty much whatever you throw at them. There's a few tricks, and one you get it, you can produce stuff that looks great quickly and easily. It's not surprising these can be imitated by AI.
Not saying AI/ML won't be damaging and cause negative effects, just pointing out that other uncompensated prior art has been used to create transformative works before we ever had DALL-E.
Yeah sampling and remixing are actually extremely similar to one aspect of DALL-E's overall system. Sure its more technically impressive but from a HCI point of view you are really just sampling via text based input all be it with much more complexity and control over the synthesis parameters.
It'll be just another tool in an artist's toolkit. The jump in quality among artists since tablets (pens to draw on a computer screen) and stock photography became widely available 15-20 years ago, has been astronomical.
Not to mention the 3D graphics of the past 2 decades, oh ma gawd! It's artist paradise compared to what people had 50 years ago.
Either way I support AI art and AI in other fields. Just because artists are mad it's gonna take their jobs does not seem like a legitimate reason to halt human progress. It's just inevitable the way things are going.
This. People thought that photography would make artists obsolete, then Picasso came round (simplifying here, ofc). Who knows what crazy creative things humans will do in the presence of those tools?
Just because some trend feels inevitable doesn’t mean we shouldn’t oppose it. We shouldn’t be beholden to the march of technology like it’s some force of nature that we are powerless to resist.
Can you give examples of when technology was successfully rolled back for social reasons (not because it was found to have dramatic side effects, like asbestos eg) and that didn't coincide with a massive drop in HDI (like the fall of the Roman empire)?
I wouldn't call that a successful roll-back unfortunately, they've been somewhat contained at best, but they're still being produced and more powerful than ever.
Those discussions aside, what I meant by social reasons was people wanting to see some tech go away because it's automating jobs.
Yes, many of us will turn into cowards when automation starts to touch our work, but that would not prove this sentiment incorrect - only that we're cowards.
Dude. What the hell kind of anti-life philosophy are you subscribing to that calls "being unhappy about people trying to automate an entire field of human behavior" being a "coward". Geez.
Because automation is generally good, but making an exemption for specific cases of automation that personally inconvenience you is rooted is cowardice/selfishness. Similar to NIMBYism.
Yeah, and we should really do something about those abominations like Excel and Python that let ordinary people create programs without hiring us to do it the right way.
There is a big difference between "this is a tool that makes some things easier" and "this is definitely endangering the skill I spent a lifetime learning".
There is no big difference. The steam engine put 90% of the world out of agricultural work. That is a big difference. Now you can buy strawberries all year round and drink coconut water in Alaska.
Nobody will care where their art comes from any more than you care about how your food's field was plowed; and all their lives will be better for it.
There is a huge difference in how you will feel when the tool that might put you out of work shows up.
Keep your comment somewhere. Come back to it when you look at a new tool that promises to merrily trample on your entire field’s income and provide an endless source of “usable, I guess” substitutes. Let it provide solace as you stare into a future with no room for the craft you’ve spent a lifetime honing.
I don't see this as a likely outcome for programming assistants.
Software development is heavily labor-constrained, if copilot can make everyone a 10x developer, we'll get slightly less than 10x the features-per-year on an industry-wide basis after contributors shuffle around.
The effect will be most pronounced in application development, where a team of 1-5 is about ideal for a coherent app made with taste, and that team could produce the output of 10-50 developers. Not such a bad thing.
Unfortunately this is unlikely to be true for visiual art, I don't predict that making artists ten times as productive will meet a latent demand for ten times as much art. Could be wrong, but my sense is that about as much art is purchased as people want to buy.
What progress? We’re bruteforcing solutions without any way to learn from then. ML eliminates serendipity. I’m not strictly anti-ML. Horses for courses and all that, but I’ve got to admit I get weird, “The humans stopped learned and the computers started,” sci-fi vibes sometimes.
This is a statement that is pretty quickly disproven if you actually pay attention to the generated art. Lately I've been seeing TikTok videos where people are using DALL-E to create "new aesthetics" - "vampwave", "neon apocalypse", etc.
Whilst automation in textiles, agriculture, etc. brought measurable benefits to society, this will have few as it stands. I can think of a dozen things that are worth automating before this. This leaves me suspecting that projects like DALL-E are gunning for Shutterstock's $500 million annual profit and that's it.
Elsewhere in this thread, people celebrate AI for some reasons. I'm not so impressed by the tech when the goal is just profit without some sort of ethical upside.
Every time I've been to an art museum (and this is a handful, not just SFMOMA), there has been a person or two in a room with some, well, art hanging on the wall. They had notebook of the appropriate paper medium and art supplies and were copying the existing work.
Copying of existing works is part of how an art student learns. That this one happens to be a math model at its core is an interesting philosophical problem. The other people's work to serve as your training set is exactly what art students do - and the works they copied and the works that they have yet to produce are not royalty encumbered.
Nor should a mathematical model. It happens that the developers working on this problem have gotten it so that it can do its learning and creation many times faster than an art student in a gallery... but it still can't get hands and faces right.
> The other people's work to serve as your training set is exactly what art students do
“Training” an art student and training an AI model are vastly different, and your equating the two is, frankly, nonsensical and absurd.
An art student isn’t a trivial weighted model capable only of mapping stolen text prompts to stolen image representations of them.
> It happens that the developers working on this problem have gotten it so that it can do its learning and creation many times faster than an art student in a gallery
It hasn’t learned anything.
It correlates stolen textual descriptions with stolen images, and then regurgitates mash-ups of the same.
This type of AI model cannot produce anything other than purely derivative work stolen from others.
When it comes to playing around in blender - my designs are obviously derivative of others - do I need to credit those artists? Even the ones that I don't remember more than a "I saw this print at a comic art show once..."
How original does my own work have to be before it isn't a mashup of stolen images that I half remember?
> If I was to dabble in sci-fi art and made something that fit in the art style of Steward Cowley … do I need to credit the art?
Probably, yes.
> When it comes to playing around in blender - my designs are obviously derivative of others - do I need to credit those artists?
Again, probably, but nobody is likely to care if you’re not actually selling your work.
> Even the ones that I don't remember more than a "I saw this print at a comic art show once..."
Then that’s not the prompt you should be starting with if your goal is to produce an original work.
> How original does my own work have to be before it isn't a mashup of stolen images that I half remember?
How original does it have to be before it’s not plagiarism?
Now, remove your ability for individual creativity, such that you cannot come up with an original idea. All you can do is plagiarize.
That’s the difference, here. This isn’t an AI trained to have creative thought, a genuine understanding of what it’s making, and original ideas. It’s an AI trained to regurgitate mashups of plagiarized works based on weighted correlation between the prompt and the (also plagiarized) descriptions of the works it’s regurgitating.
What do you think the artists "trained on"? Didn't they ever see anyone else's art? Don't you think it influenced their art?
If it's a 1:1 copy, I agree. If it's a "that looks vaguely like the style that xyz likes to use", I disagree.
And I assume you'd run into plenty of situations where multiple people would discover that it's their unique style that is being imitated. Kind of like that story about a hipster threatening to sue a magazine for using his image in an article only to find out that he's a hipster and dresses and styles himself like countless other people, so much so that he himself wasn't able to tell himself apart from another guy.
Your entire point hinges on a false assumption; “training” a human artist (or programmer) is the same as training an AI model.
It is not.
The AI model can only regurgitate stolen mash-ups of other people’s work.
Everything it produces is trivially derivative of the work it has consumed.
Where it succeeds, it succeeds because it successfully correlated stolen human-written descriptions to stolen human-produced images.
Where it fails, it does so because it cannot understand what it’s regurgitating, and it regurgitates the wrong stolen images for the given prompt.
AI models are incapable of producing anything but purely derivative stolen works, and the (often unwillingly) contributors to their training dataset should be entitled to copyright protections that extend to the derivative works.
That’s true whether we’re discussing dall-e or GitHub copilot.
> The AI model can only regurgitate stolen mash-ups of other people’s work.
We all stand on the shoulders of giants. If you're very dismissive, I think it's easy to say the same about most artists. They're not genre-redefining, they carve out their niche that works (read: sells) for them.
The art that was trained on was all Creative Commons apparently, so perhaps artists should understand licenses first before giving their work a permissive license.
All CC licenses, other than the “CC0 public domain dedication”, require attribution, but CC themselves have taken what I’d consider to be an ass-backwards position on the matter:
> At CC, we believe that, as a matter of copyright law, the use of works to train AI should be considered non-infringing by default, assuming that access to the copyright works was lawful at the point of input.
It's not backwards. It's the same as if a human artist studied it.
Backwards would be thinking that any creative work you make is a derivative work of many creative works that you've seen in the past which you aren't copying from.
Where is this said? I'm looking at the collection of artists this site is happily offering to rip off (https://twitter.com/arvalis/status/1558632898336501761/photo...) and there's a lot of people with long careers making copyrighted work. I really can't imagine finding a ton of CC work by Bernie Wrightson or Wayne Barlowe or Brom or Junji Ito, for instance.
I think a more 1:1 comparison would be self-driving delivery vehicles - which we haven't figured out yet.
and
I imagine a primary motivation for investing in art-generating AI is cost savings.
disclaimer: By default, I tend to be less sympathetic to artists here. It's due to their communities' collective silence about decades of ever-ratcheting (and typically purchased) copyright laws.
Artists will have to find people who pay for the genuine thing. We can all buy cheap mass-produced tableware, accessories, and so on, yet many thrived on Etsy.
People are still gonna draw with pencils lol. That's not going anywhere. No time like the present to pick one up if you have a reason to need to render images imo.
one area where human artists excel: learning. if i ask dall-e to draw something that was made after it was released then it wouldn't be able to do it without a very detailed description
One thing that makes me a little hopeful is that every image I've generated with DALL-E 2, even the best ones, would require non-trivial work to make them "good".
There's always something wrong, and you can't tell the model "the hat should be tilted about 5 about degrees", or "the hands should not look like ghoulish pretzels, thanks".
There's also this fundamental limitation that the model can give you a thing that fits some criteria, but it has no concept of the relationships between elements in a composition, or why things are the way they are. It's never exactly right.
It's like the model gets you the first 90%, and then you need a trained painter to get the second 90%.
But yeah, it will certainly devalue the craft, don't get me wrong. And anyone who is callously making comparisons to buggy whip manufacturers should consider how it would (excuse me, will) feel when AI code generators pivot to being more than a copilot, and suddenly the development team at your office is a lot smaller than it used to be, and maybe you aren't on it anymore.
If you spend a lifetime mastering some skill, and then it's just not valued anymore, it sucks, and you get pretty mad about it.