AI-art isn't art: DALL-E and other AI artists offer only the imitation of art | Hacker News Reader

AI-art isn't art: DALL-E and other AI artists offer only the imitation of art

erikhoel.substack.com

76 pointshardmaru143 comments

Amazingly no mention of Andy Warhol, who made a factory to produce art, via silkscreens and assistants. Sometimes he added embellishments of paint on top, other times the work completely called into question the value of authenticity in an age of mechanical reproduction.

The introduction of the camera caused similar reverberations. And actually, if you go back to the impressionists, and through cubism to abstract expressionism — it was all a revolt against “image” being “art”. The surface, the paint, the whole canvas — that started to take on more importance. Art became less and less about the sheer skill of reproducing a scene photographically. After all, a camera could now do that so well.

Warhol was interesting because he called into question things like, is business art? Is art business? Can art be mass produced? Is that a statement on society, etc.

Honestly, people get too precious about What Is Art. And what people in galleries and museums and art school consider art is always way different than what the average person calls art and hangs on their bedroom walls.

Personally, I’ve always went with a definition my dad told me. Maybe it was Picasso originally: “Art is what makes the modern world felt.”

The thing about Warhol's "Factory" was that it wasn't really a new thing. Most pieces of art even from the Renaissance that we think are made "by an artist" are in fact made by a collaboration of an artist with their apprentices, much like how most science is done by a primary investigator and their students/postdocs.

There is a story that Verrocchio hung up his pallet after seeing how good one of the pupils in his workshop was at adding a cherub detail to one of his paintings. The pupil’s name was Leonardo. The story is likely apocryphal but to your point, the ‘factory’ model isn’t new to art.

A fun take on this - https://moforgeries.org/

This confuses conceptual art with the process of creating a silkscreen (or spot painting, etc).

Duchamp's readymades also come to mind, as an earlier example asking the what is art question.

And those really were highlighting the 'looking at looking' mentioned in this piece. Either it's worthless, or it's branding, or it's about having a chance to see what this artist has seen. In art markets, the answer is branding... And that sort of thing - as both Duchamp and nfts have shown us - is pretty independent of the quality of the art itself.

> Art became less and less about the sheer skill of reproducing a scene photographically. After all, a camera could now do that so well.

That art was ever mainly about this reeks like a modernist* lie to justify their dismal works. Look at https://en.wikipedia.org/wiki/Wanderer_above_the_Sea_of_Fog - is photorealism what made it good or famous? Is any random 100% photorealistic picture taken with a phone on a hiking trip therefore strictly better?

*Or whatever the art movement is called where three vertical lines or a banana taped to a wall are considered grand achievements.

I feel as if I just read an article from an 18th century cobbler complaining about how only humans will be able to make shoes because of their innate understanding of the human experience of making and wearing shoes.

1. If 'art' is expression, then it should be judged by the 'quality' of the expression, not the tools used to make it.

2. Most 'commercial art' is made to a spec generated by a customer for a specific need, not for 'expression'. Future DALLE variants will do it faster and cheaper (arguable even better) so that will be what is used regardless of the philosophical implications. Artists are not going to be needed to same extent as today (we already saw this with clip art, public font databases, and gig sites). The economic impact will be the same regardless of your interpretation of 'art'.

3. I am sure There will still be famous and acknowledged 'artists' known for their ground breaking work regardless of what tools they use. As an analogy, There are still companies that make shoes by hand, they are just an expensive luxury.

I am guessing we will next see article from a translator about how automated translations system lack the innate 'sensitivity' to the human condition to be used in professional translation (it's true to a certain extent - for now).

For me, the primary concern isn't about Art as an economic activity. It's about what happens to human society when communication becomes divorced from human meaning.

The original purpose of art (and language) was to communicate the thoughts & views of one human mind to another. It's a fundamental need, the essence of being human, the cornerstone of culture and society.

DALL-E and GPT-3 are about to unleash a tidal wave of content that no longer performs that function, simply because _there is no human mind at its source_. What happens to us then, when we are constantly receiving an overwhelming babble of voices that appear on their surface to be meaningful, but lack any true meaning at their core? What happens to our voices, our relationships, our society?

Someone still wrote the prompts, perhaps iteratively until they got results which pleased them enough to save and share with others. Other tools allow modifying the results by coarsely sketching where you want semantic changes. You could have a dialogue with the AI criticizing whatever you did or didn’t like about the current draft, and what changes you wanted to explore.

If someone made a comic book with these kinds of tools, I’d score it as a win for enabling personal creative expression.

When writing was invented, I suppose one could have feared the loss of natural human bilateral aspect of communication. Writing detaches the writer from what is natively a direct interaction between people, and allows us to “speak into a shell,” so to speak.

100% agree. It's going to shatter communication into a thousand shards. Maybe hipsters will save us - they might bring back unmediated face to face communication as one of their anachronistic hobbies.

Isn't that Reddit?

> 1. If 'art' is expression, then it should be judged by the 'quality' of the expression, not the tools used to make it.

What about the amount of effort it took the artist to make the expression? Doesn't that count?

Gatekeeping what is and is not art is an exercise exclusively for fools and assholes.

What does it matter how many steps lay between the fingers of the artist and the work of art? Is a photo less art than a painting? Is a photo from a wildlife camera not art? If I paint a mouse's feet and let it run around on a canvass is that art? Can art be unintentional?

AI art is about to lay utter waste to traditional visual art forms. This will be so much more destructive than what the Internet did to music.

See? Assholes. Whining that an explosion in creativity and accessibility ruined music.

Not just what is or isn't art, but what is or isn't meaning.

There is a distinction between “fine” art, and what I’ll call “working” art people with artistic ability produce for money. Think graphic designers/illustrators/etc

Fine art is going to be okay. Authenticity has always been a big thing there. New social critique is going to be something AI trained on last year’s data will struggle with. Ditto for pushing boundaries or establishing new styles.

Working art though… it does look doomed. Who’s going to pay an illustrator for something DALL-E does for pennies? You may still need someone with good taste to pick the images, but as the author points out, that’s not the same as being the artist.

Yeah I am not sure how this will play out for professional illustration...

You can trace a constantly evolving visual style in web design back to a root cause: "The need to look modern and legitimate."

It's hard to remember, but there was once a time where good stock photography was a way to look modern and legitimate. That birthed a thousand cheap stock photo sites and eroded the effect.

So we moved on to a constantly evolving style of illustration. Skeuomorphism, corporate memphis, gradient meshes, clay. Websites use these to signal modernity and legitimacy, but the very use of it erodes the effect and they have to move on.

If we can write an ALT tag and get a professional-looking illustration in any style, that will erode the effect, and we will have to find new ways of looking modern and legitimate.

Maybe that will involve professional illustrators, or maybe it will move somewhere else, just like when it shifted from photogs to illustrators.

Well we already know how to do style transfer. But yeah, it'll be interesting if DALL-E is considered passé in 10 years.

I saw a similar argument recently, subtly different in that it was claiming the root cause was needing to look like you spent real money as a way to signal your legitimacy. I think they called it peacocking, and it’s the same idea that “virtue signalling” had before that term got used for “politics I don’t like”.

Spending money to prove you have money to spend made me think it would be a perfect fit for NFTs. (Not any NFT in particular, just so long as it’s needlessly expensive in any strictly objective sense).

If I had to guess, I'd say the trends will still be set by big design agencies and artists who get paid handsome sums to work on a big company's rebranding project.

The difference will be that many of the smaller illustration gigs, which essentially boil down to "copy the style of this big company" will cease to exist. Those types of gigs don't require a whole lot of creativity or skill, because the client simply wants a knock-off of what someone else came up with, but for cheap. AI-generated "art" will allow those clients to bypass the human and just get what they want directly from a computer.

In a sense, it's going to be similar to what professional photography has gone through in the past few decades with cameras becoming more and more commonplace and affordable. Some people still want a professional portrait taken of them, but the majority of people are perfectly happy with what a mobile phone camera can produce.

Working artist not finished yet... But it doesn't look great

It depends on the level of quality, polish, precision or originality that you want. It’s pretty likely that some people will be much more talented and skillful with this new class of AI-art tools than others. People will figure out how to do much more with these (evolving) tools, by exploring and interacting with them, and build up a body of knowledge which they can teach to others. I imagine that there will be classes in how to make better AI-art, which could have wide appeal since the technology makes content creation much more approachable for casual users.

I think it will give rise to new art and design trends that are harder to imitate.

> Fine art is going to be okay.

I'd argue that fine art has been the opposite of "okay" for the last hundred years..

> AI art is about to lay utter waste to traditional visual art forms. This will be so much more destructive than what the Internet did to music. It will be a technological conquest of one of the great human avenues of spiritual transformation.

AI now challenges what we've always thought was unique about us, "soul" and crap. Some people who can't do anything else obviously get desperate. And don't get me started on the fact that "art" is not defined or explained nowhere in this essay. Let me define "art". Art is not Malevich's Black Square - that's 50%/50% crap/PR.

Art is ~90% work and ~10% ingenuity. AI cannot reproduce real art because no one will see any work behind it. People go to galleries not to just appreciate ideas, they go their to appreciate work. Real artists are safe. Sleep tight.

Dall E suffers from 'regression to the mean'. Because it uses existing images, what it makes never seems very much different from what we've seen before. It's kind of an average of previous works (known to Dall E).

Artists are supposed to create something new. By composition or invention or whatever. Dall E makes a collage of images. Even it's composition seems average.

So sure call what Dall E makes, art. At best its a dull, average artist creating derivative images.

Models can produce out-of-distribution results if you push them. The results won't necessarily be good, but describing it as an average, suggesting that it'll always be inside the some tight hull of the training set is misleading. Also note that averages can actually be highly unusual results if you're working with N-modal data, if you have a training set of horses and unicorns then you can get unusually short horns or other weirdness like horn-looking hair decorations.

Just look at the avocado chair or dogball.

> its a dull, average artist creating derivative images.

Which is primarily what sells.

From the point of view of a digital artist, one thing this article fails to consider is how the software itself is a creative work of art. Instead of fixating on the artifacts it produces, it’s kind of amazing to see that someone made a magic paint brush that can mix other artworks. I’ve exhibited at art shows and listened to people dismiss my work as “computer generated”, when I wrote every line of code and used and refined the software for thousands of hours to instill my intentions in the images I make. I can’t blame people for not understanding the work, but it’s funny to watch them make incorrect assumptions and stick to those assumptions fiercely.

The author is also glossing over the fact that the vast majority of art produced by humans is iterating and riffing on and mixing other pre-existing artwork. Almost nothing is pure originality, even the mechanisms we use to produce art are largely unoriginal. Oil painting & sculpture are not the only ways to make art, yet like the 4 piece rock band, the formats have become canonized and very few people work outside the framing of their predecessors.

The intro says the rock faces are not art, which feels to me like the author doesn’t really know what art is, nor of the history of debate over what art is. The author argues later that art is not in the eye of the beholder. The author claims there’s no intention in the rock faces, and that it’s a “joke”, completely ignoring the intention of the person who collected the rocks and put them together. What exactly is the difference between finding rocks with faces, and, say, photography? Good photographs are are nothing more than existing images “discovered” by the photographer. People argued for a long time that photography was not art, but it’s rare to find someone today who can say that seriously and compellingly that photography is not art.

Hand-crafting generative code is the opposite of DALL-E.

It's about affordances. A paint brush is a stupid simple object, but it offers an almost infinite range of creative affordances, and almost endless potential to invent new ones.

A camera is also an affordance machine. You don't discover images, you frame them. And you interpret them by selecting and manipulating available light, tone, and colour.

And then very likely you do even more work later in Photoshop (etc.)

DALL-E has almost no affordances. Prompt goes in, studenty surreal "art" comes out.

There's no creative control. It's disposable novelty. That's all.

There are AI art systems that apply transformations to photo and text prompts. They're much more interesting, especially when they produce images with very high production values and genuine surprise.

So it's not that AI art can't be interesting. But DALL-E is what you get when people who don't care about that throw some ML tools into a bucket and pull out a blob of code.

It's not just that it's trivial, it's condescending in a way that useful art tools aren't.

> There's no creative control. It's disposable novelty. That's all.

People literally said this about the camera.

I mean, I admit I’m playing just a touch of devil’s advocate here. As an artist on occasion, I’m not going to be wowed by the person who asks DALL-E for a landscape image. OTOH, I’m certain there can be creativity and skill put into the prompt text, based on time spent using DALL-E. And my main point is that the system is the interesting part, not the pictures.

> It’s not just that it’s trivial, it’s condescending in a way that useful art tools aren’t.

This opinion comes down to pure narrative though, right? If you see any given image and don’t know how it was made, there’s no way to judge it as somehow condescending. I’m certain DALL-E could spit out an image that you would not be able to identify as AI generated. If I didn’t have context, I’d have no idea the robot hand thing was DALL-E (and I have no idea if it spit out one of it’s training images, or if it mixed together multiple images.)

You mention utility, and I don’t know what you mean by that. Could you elaborate on why you claim AI art like DALL-E is not “useful”? It’s pretty clear to me that DALL-E could generate clip art or ad art or stock photos quite easily, and would potentially replace humans that also pump out schlock.

This position is defensible, but it's only partly true.

The author is foremost an editor of their own work. Take a lot of pictures, keep the good ones. Write a lot of junk. Keep the good parts, rewrite the others. Write a lot of songs. Publish the good ones. Etc.

It's eminently possible that AI-art will allow faster iterations on ideas, and therefore help create more good art.

> AI artists offer only the imitation of art

Doesn't this depend on the "beholder's" definition of art? In some contexts I want art to say something meaningful about the human condition. AI art can't do that (at least today and maybe ever).

In some contexts I want art to be beautiful or perhaps just decorative. I think AI art can already do that, at least sometimes. When it's successful, then it meets my definition of "art" in that context.

In other contexts I may need art that illustrates an idea or conveys a concept. AI art can already do that to. Thus, I don't agree with the premise "AI artists only offer the imitation art" at least in some contexts.

Dall E is a tool. Just like every other tech ever made. A paintbrush doesn't make the Mona Lisa, but it sure CAN. that ability to do something with a tool is the real art IMHO. Prompts are the paintbrush of DallE. Can i have random prompts? Yes. Can a strap a paintbrush to a double pendulum? Also yes.

Yeah the art here is the prompt, the input to the machine. I usually say, that any act of human like work, or exercise or singing, or doctors, or engineers of any kind, will become information-analyst and engineer, information-analyst and doctor, information-analyst and painter.

That information analysis does not necessarily need to be performed by the same person. One is the information analyst, another one one is the painter, and they both in collaboration create some art paintings. I don't see any problem with that. Which one is the artist in that case? The information analyst or the painter, or maybe both? I would argue they are both artists.

I have personally prompted and generated hundreds of pieces of art, real art and awesome art, and anyone who tells me that i don't know what a Picasso is, doesn't know what he is talking about. Picasso was my favorite band for a long time!

In the end, painting is 99% a practiced skill, and 1% original human thought. And this 1% is arguably comparable in effort to what is provided as input to an AI tool.

Is human thought only at the outset of a piece of art?

That is far too reductive

Human DNA is only 1% different than chimps. 1% is not that reductive.

This is a great way to put it.

your behavior and thinking is purely a function of your DNA?

I mean, according to a physicalist's view of the human condition, yes. You are a function of the haploid cells your parents conceived you with, and the environment those cells responded to over your lifetime.

I would say that “and the environment” means “no, not purely a function of your DNA”.

Sometimes I wonder if hackernews ever does anything with the ideas it generates or if it just gets a kick out of generating them.

actually, that stat is completely and totally wrong. If you take the most similar parts of human and chimp DNA- just the well-aligned proteins- then yes, we differ by 1%. But that's not a metric that means much, if at all, since many aspects of phenotypes are not controlled by the difference in our protein coding regions.

lmao no

I would disagree with that. Much of creating art is a feedback loop. An artist rarely, if ever, has a fully formed picture in their mind of what a finished piece will look like, and it's even more rare that the final piece conforms to that image. In the process of creating, the artist takes in what is in progress and has to adjust accordingly.

The final piece is influenced by the medium it's created in, the process of creating it, constraints both momentary and well-known, and the context of the time and whims of the artist.

Good point, I'm guessing there will be successors to Dall-E that will address this and implement support for human feedback.

I don’t think 99% of the population or even art students could paint Dürer’s rabbit no matter what training or incentive was offered.

Also the history of pigments is fascinating. There’s a whole lot lot more to painting than brush strokes.

>> In the end, painting is 99% a practiced skill, and 1% original human thought

True, but the line between art and great art is 99% thought and 1% skill.

The billions of images and its software determine its outputs right? Is there the same analogy for humans?

I don’t think so. It’s much harder to imagine a deterministic story of how we developed set theory for example. Nothing empirical forces us into set theory. It’s not needed by physics for example.

The billions of images and software plus the entered phrase fully determine the output right? What possibly similar story gave rise to things like set theory?

Sure, you can say DALL-E is derivative, only good at imitation..

...but what about DALL-E 2?! I mean, that's an artist!

The story used pictures from DALL-E2 so I suspect that they were referring to the platform as a whole.

Is this satire? In today’s internet, it’s sometimes impossible to tell, at least for me.

I don’t think so. It is indeed on another level from its predecessor https://youtube.com/clip/UgkxMgheH1Ck-f6FG3o_lcFYLcLZhzb_ZeI...

It is more subjectively art like.

Cards on the table first: I consider the stone faces in the Mori Art are genuine art.

My first point is this: humanity is part of Nature, so all art is ultimately created by Nature. Human-produced nature the process just involves raising a human at some point during the process.

Second point: Human artists don't live in a vacuum; all of them create through "imitation and recombination" of other art; "Great artists steal" and all that. The fact that an AI "recombines existing art" is an argument that makes it more like human-produced-art, not less.

Third is the time scales involved. Humans start producing art when they are 2 or 3 years old, and finish producing art when they die - let's say 80 years old. On the other hand, it takes nature thousands of years to produce one of those human-looking stone faces in the museum through geological processes alone. Considering the times involved, the human-produced option feels like "the shortcut", the cheap option. Using an AI to pump 100 works per second is just the obvious next step.

Finally, in the case of the Mori Art Museum, *someone* had to go through those rocks and decide "this one looks like a face" (throw into pile A) and "this one doesn't look like a face" (throw into pile B). That constitutes "Art", in the sense the Author is talking about - there was a consciousness and curation is a form of art on its own right.

The stone faces themselves are not art, but the collection of stone faces is art. It's not just a random sample of stones.

In the same way that particles of pigment are not art, but clusters of pigment spread on a canvas is a painting.

Art has a point of view.

I agree. But I think it's the aesthete's point of view that is essential; the artist's is optional.

I think it's another case of moving the goalpost whenever AI can do something that used to be exclusively human. If we could send DALL-E's images back in time 10 or 20 years, everyone would say "of course it's a real art, computers are too dumb to do that!"

I expect DALL-E to pass a picture-based variation of the Turing test (an evaluator asking human artists and AI to draw things). The current implementation has a few small giveaways like drawing hands poorly, but that seems fixable.

Please explain why this is different than the first machine calculator. There’s already been semi Turing tests on Twitter for DALLE and it “passed”.

If people said that, they wouldn’t know anything about (conceptual) art. If you’d walked into an art school, they would have said “no, that’s automated illustration generation.” Which is what you’d get today. It’s amazing that HN thinks DALL-E is somehow an artist. This happens because most posters on the topic don’t seem to understand what art is. It’s not illustration or decoration. “Good art” is not a painting that looks particularly realistic or dramatic or colourful etc.

There are artists in my family and I’ve worked with a bunch of artists myself, but you only need a cursory understanding of art to know DALL-E is automated illustration.

I think DALL-E is amazing, absolutely amazing, but it is not creating art in the sense that an artist does. If you think it does, you’re probably confusing artists with illustrators.

Well no, as the author notes, Tolstoy at least would not have categorized these products as art and he wrote over a century ago.

I may come across as dismissive but most people don't care for abstract or portrait-y art. I know I don't.

What I appreciate is comic books. With the amount of computer tools available today, comic books are way more beautiful than two or three decades ago. Some of the most gorgeous art I've seen is in comic books.

DALL-E can't replicate that and I doubt any AI can (or will be able to) draw exactly what some writer is envisioning in their brain. So artists aren't going anywhere.

Why not? DALL-E can already emulate comic book style. With more tweaks and training I imagine it could be made pretty good at it. DALL-E2 is much more precise and coherent than DALL-E1 was, and I see no reason why we wouldn't get an improved DALL-E3/4/5.

I bet it's only a matter of time when it can create an ok-quality full comic book based on a textual script. And perhaps the script could even be generated by GPT-like AI too.

I expect we will see lots of indie comic books made using these tools, and perhaps this kind of comic will ironically be seen as less corporate and more authentic?

XKCD is doing fine with minimal art, because it’s about the writing. Other writers will make comics very different in style, without having to settle for minimalism. Very long, story-driven comics will be more feasible for amateurs to make in a shorter time. Instead of drawing a few pages a week (which is a lot of work), they could write and illustrate a short chapter.

More labor-intensive comics that are more about gorgeous artwork than story will probably still be around, but a smaller niche than before.

I was thinking of Marvel/Star Wars/DC comics instead of webcomics like XKCD but yeah youre right.

These tools will lead to a boom in webcomics and memes where precision isnt important. The superheroesque traditional comics will be harder to replicate with this.

Comic book computer generation is not here yet, but if you want my opinion, it is coming soon. I don't see any reason why we cannot create a cartoon octopus call it Octofred, use it as a character and embed it in a story.

Yeah well try to sell it at 4 dollars an issue.

On the plus side modern studio art is so execrable that AI laying waste to it will probably be a net positive. The human urge for expression will always find an outlet and perhaps this will channel it in new and better ways.

Definitely. There is a weird elision on the author's part that other forms of artistic creativity exist. FWIW there is plenty of room at the bottom for those artists that can pull together the tenuous threads between seemingly incompatible disciplines.

Artists should be thankfull for DALL-E. A century ago (more or less) another tech invention - camera - started to create realistic pictures. Thanks to the camera, artists were able to move away from painting realistic pictures into abstract art. They should really be thankfull to be able to live in the times when another tech invention forces them to move the art further forward.

I agree. When the tools become more advanced, effort can be liberated toward more creativity.

Can't wait to end up where every artist's work is under scrutiny for being made by an AI.

And I can't wait to see the creations where the artists incorporate AI, using it as proficiently as the brush.

https://www.youtube.com/watch?v=yk5EJd7S2uA

Look at this example, it's just beginning.

Um, figurative painting is all the rage in the actual art world today.

Art was never primarily about realistic mimicking of nature, even if some subset of it in some cultures may have temporarily appeared to be.

Are we sure about that? The cave paintings at Lascaux are of recognizable animals. Egyptian and Maya art are naturalistic too, and highly refined toward realistic portrayals.

For most of Western art's history, content and style were both dictated by the patron. Even the ideas we hold around art and freedom of artistic license had to wait until the 1800s and the Salon de Refusé to come into existence.

Most buyers and makers of art today still gravitate toward figurative drawing and naturalistic landscapes. It seems like we have always used art primarily to reflect and comment on the real world.

This is absolutely true. Art post-impressionism is a reaction to the camera. I'm excited to see what new art can come of all tech inventions.

Consumer-ready black and white photography is about 150 years old. Many “realistic” (whatever people mean by that) paintings were made after that point. Consumer-ready color photography is about 50 years old, and not a lot of artists followed any traditional style by that time.

In most simple terms, the oscillation between vaguely “metaphysical” (imaginary, romantic) and vaguely “physical” (realistic) approach has occurred for a long time.

Discussing whether it is art or not is completely useless and irrelevant for me. I get it that there are people with different opinions, but I just don't care, it would be fighting about a definition and feelings.

What I care about is that if I need a sweet little drawing for my presentations, some cool painting that I can get printed and hang in my kitchen, a cute gift idea, or as a book cover, I can describe what I want and, at least based on the images that I have seen so far, I can start generating pretty good images, and stop when I found something that I absolutely like. I can be ready in minutes.

The cheapest alternative for someone without artist friends would be going on fiverr, browsing artists for maybe hours, check their rates, hope they are actually good, hire one, wait days or weeks to finish the job, feeling uncomfortable when asking for changes, and hopefully not having to do this one more time because the first artist was crap.

The difference is gigantic.

I've seen incredibly long prompts to DALL-E to get a precise and incredible result, much better than anything I could come up with

like everything before it, it's just another tool that some will show a 100x mastery of than the average person

> I've seen incredibly long prompts to DALL-E to get a precise and incredible result, much better than anything I could come up with

Have you tried having DALL-E draw a blue cube, red cube, and yellow sphere in the same image?

https://twitter.com/david_madras/status/1512573390896480267

I've always pondered if code could dream up idiosyncratic styles of art. Instead of using 'training' material, it procedurally generates art using random noise as the base input, and uses a different algorithm each time it generates art, instead of a predictable algorithm (to avoid everything looking the same). You could, in effect, generate serendipity on demand, and use algorithms to create algorithms, a very meta thing to do, but would be an interesting project to undertake.

It won't work because the resonant part of art doesn't come from random noise. It comes from an inflection point on a particular emotional introspection or connection. When the viewer or listener shares this connection, art happens. Without this emotional exchange, art is lifeless and uninteresting.

Interestingly, music works the same way. Random, unpitched noise sounds to us like wind or rain. In music synthesis, randomness has a use (like in DALL-E) but it is limited. What humans want hear to "get into" a song emotionally comes from specific arrangements of pitches - not random ones. The artist chooses them because they "feel" a certain way, an exclusively human concept. When the listener has the same emotional reaction, art succeeds. No DALL-E painting will ever achieve this because it has no feelings to begin with.

What's missing from DALL-E is novelty and invention. The greats artists all invented something. Van Gogh's style never existed prior. Neither did Warhol's. When cameras were invented, artists complained and wanted to unionize to prevent their use. Yet we got Ansel Adams out of it.

George Eastman helped democratize that process and, in doing so, took most of the art away. His slogan for Kodak was, "You push the button, we do the rest," anticipating the way AI illustration works.

But no amount of DALL-E output will ever create art and the reason is in the first paragraph of this story: it all comes from existing images. Photography has become so un-artistic now that everyone is a photographer and photos are staged with Instagram-ready locations before anyone even takes the (exact same) photo.

This obsession with tech and the consummate commidification of aesthetics has even leaked into other classic art forms. When was the last time anyone invested in architecture as an art form? Most buildings are vernacular, a kind word for popular and hideous.

Don't get me wrong. I love both art and tech. I've sold my art photos for thousands of dollars, and I've spent thousands on tech for making music when I could just sing for free. There's a use for both these ideas, but the invention of a tool does not art make.

I think that many spectators of human-made art also fall for something close to Pareidolia, Apophenia, or the sensing of meaning that is not (necessarily) there. Even expert critics can't infer exactly what authors or creators had in mind when creating a work of art or literature and without a statement from the artist it's just a heuristic guess based on culture and personal taste. People have wondered for hundreds of years what the Mona Lisa's expression means, making assumptions about Da Vinci's technique and intentions for her smile, but without definitive evidence from the artist himself it's all guesswork. Likewise in this article the author assumes that art is about communication but in fact art is a very lossy and ambiguous communication medium. Human artists choose to put concepts in juxtaposition with each other but what that means to people differs with each individual, and so what an artist intends to communicate may not come across even if people strongly feel that they are right about their own interpretation. Dall-E2 is almost as capable at the same kind of juxtaposition (astronaut riding a horse) that makes people think about what it might mean.

> This communicative property of art is irreducible to its extrinsic properties (like color, form, etc), and instead concerns the significance of the artwork to the artist and their communication of that significance to the viewer. To steal some terminology from analytic philosophy, let us call these latter properties the “intrinsic” properties of art to separate them from the merely extrinsic properties of art.

While I agree that there is a fundamental difference in meaning between human and AI art, and also about how dismal the prospect of a world in which most art we enjoy is AI-generated, I feel the author is cheating a bit by ignoring half of the “intrinsic” properties.

That is, art is also about (to me, mainly about) the experience it produces in the viewer. This intrinsic property is not affected by whether the author is AI, or even a natural process. Though I agree it would be unusual to call the Niagara Falls 'art'.

Okay, but then we shouldn't call something "art" due to the creator's style or brush technique either.

What I like about AI and algorithmic art is that it can produce near infinite varieties around a style or technique, and often produces more interesting results than an artist leaning on a style with very little meaning or reason behind it.

I'm not interested in debating this topic yet again. Photography, movies, radio shows, TV, video games, digital photography, self-published books, online personas acted out in real time. Whatever it is, if it's new (and people are still figuring it out) it's supposedly "not art" until the critics die and nobody remembers them and all we're left with is people making shit and other people finding some kind of connection to it.

Even the rocks example sucks. No, no conscious person carved that rock. But a person did pass over thousands, millions of other rocks to select that particular rock and ascribe meaning to it. How is that different than an artist who shoots hundreds, thousands of photos, and picks the one that is "good"? The one that they will then stamp approval in and say “this is what I want to say"? It's not.

Whatever, dude. You do you. You'll be dead soon enough.

This reminds of people in the olden days saying that computers can't actually play chess. It's just a well trained excel spreadsheet. But they don't actually play the game.

The article doesn't mention the Turing test. Isn't that the relevant criteria here? Can an expert tell the work in museums from what DALL-E produces?

It doesn’t take an expert to tell DALL-E 2 from a human artist. I attained a mere GCSE[0] grade C or D in art (can’t remember, it was 22 years ago), and yet I scored 27/30 on this exact question this morning: https://news.ycombinator.com/item?id=31467045

[0] UK exam taken at age 16, which was the age when UK mandatory education ended at the time I took it: https://en.wikipedia.org/wiki/General_Certificate_of_Seconda...

Agreed! But that's not really the point the article is trying to make. Once you know DALL-E, it is pretty easy to see which are DALL-E and which are not. Just like you get to know an artist. But in 10 years, if we have 100 DALL-E like networks and you compare them to 100 artists, can we tell the difference then?

> saying that computers can't actually play chess. It's just a well trained excel spreadsheet. But they don't actually play the game.

The code has no epistemological grasp of whether it's playing chess or calculating hurricane paths.

We are body, mind, and soul. Externalizing the mental rules of $ACTVITY into silicon for execution does not grant a soul to that silicon.

One might argue that coding chess for a computer is only "good" insofar as it helps the coder understand the problem space, or maybe produces better training tools for humans.

Computation is a means, not an end.

> The readymades of Duchamp were art because of their audacity, because of the context of the other art around them, and because of their arrangements...

If you open the door for Duchamp's urinal then you can't close it on DALL-E. Everything you said there applies to DALL-E.

I don't think Duchamp's urinal is particularly great, but DALL-E'S process is pretty much the opposite. Ask it to conceive a "fine art scuplture of a fountain" and it'll draw something that looks a lot like the existing sculptures, with water spouting out of it

Only way you're likely to get a urinal from DALL-E is if you ask it to draw a urinal (or curate a dataset full of urinals, possibly mislabeled).

And nobody would have considered Duchamp's urinal art (never mind still be talking about it today) if he'd produced it when asked to design a bathroom rather than an opportunity to install whatever he wanted at an independent artists' exhibition.

It wouldn't be hard to get it to change words so it doesn't give you what you want but something with similar properties.

I don't think that's art; it's just a social game, yet that's somehow meant to be the definition of art these days.

Another interesting take is that a lot of human made art isn't art either. Art is supposed to be questioning or critiquing social convention, not be "pretty pictures" which the large majority of output have been for the last few decades.

Not every picture has to be a banksy to be art. Art is in the eye of the beholder.

DALL-E demonstrates a lot more creativity than Banksy, in my opinion.

Art is just supposed to instill an emotional or intellectual reaction. Pretty pictures can convey beauty found in nature and do that.

> Art is supposed to be questioning or critiquing social convention...

Says who?

Let the market be the judge of what is art, if you didn't know this is AI would you buy it? as simple as that. The rest is an opinion nobody needs. When it comes to art nobody has a damn clue, so i got bad news for you mate.

Looking from the entertainment industry perspective: Ai art is somewhat similar to photobashing, in my view. It's primarly a tool used by skilled artists to convey things quicker and with less effort. The finished piece still requires the artist to have enough taste and competency to apply it in a way that is pleasing to the viewer.

If I hire an artist to deliver the concept for a building that has x number of parameters, I don't really care what % of the image is generated by AI, photobashed or drawn by hand as long as the parameters are met, the overall image looks good and isn't plagiarized.

Last year I built an AI-art installation from scratch and posted the DIY-documentation here on HN [0][1]. A big part of why I built the installation was to stir up this exact discussion. To me, art is about emotions and making people feel, even if that feeling is their strong opinion on art.

[0] https://github.com/maxvfischer/DIY-ai-art

[1] https://news.ycombinator.com/item?id=28221904

Art is just a label. If we don’t call DALL-E images art then that changes nothing. It doesn’t give any struggling artists making bland album covers their job back if someone would rather have a “synthwave dog on a surfboard in space” generated for free.

Whether it’s album art or just an image on an album cover? A meaningless distinction.

Defining art as “anything born out of a conscious creative process” is fine. Any other definition is also useful. Not sure there is situation when having a strict definition is necessary. Other than for arguing what art is, that is.

Maybe a set of rocks that have faces by coincidence is not art, but once an artist arranges them into a nativity scene? That is art, yet the simple arrangement of rocks could have happened without the artist, so art is really about intention of the artist.

Yet, AI is not happening in a vacuum, it cant exist without art as data, and without a developer, so it already has intention from two sources, which makes this a lot murkier, and certainly the second an artist uses a language model to create some art, that is still art, even if the AI is merely a tool.

We used to only have art that was created or discovered. Now we add art that is synthesized, which still requires a prompt and optionally curator.

I'm thinking about immediate applications and extension, instead of describing the image, describe a context and effect: e.g. a financially distressed couple, relieved by a comical accident. The natural next steps seem to be in advertising, or storyboarding shallow content--still best used by a human than let loose like monkeys on typewriters.

Not being an art collector for investment purpose the only thing that patters to me is if it please my eye / invokes some emotions / etc. If it is made by computer so be it. If it costs basillion dollars just because it was Andy who spilled some paint on canvas - I do not care as it feels like a noise to me. I respect that the other might have different opinions but in art I only care what I like. Whether it is valued by somebody else is irrelevant to me

Imagine Picasso is paralyzed. He can't paint, can't draw, but he still has the same mental faculties. You sit him in front of a computer screen, which displays random pixels for all eternity. Whenever, after many eons, the screen displays the exact pixels of a painting he wants to create, Picasso blinks his eyes to signal that the image should be saved. The end result is that he produces exactly the same set of works that he did in real life.

Is this art?

Ultimately the discussion falls to the notion of whether a simulacrum is experience.

If you see a painting of a horse accompanied with a description of why the artist painted it, does it matter if those molecules were arranged that way by human experience?

In either case, you are now being prompted by a simulacrum. And should you then artistically be inspired, your readers are viewing a simulacrum of a simulacrum. Does this compound the "error" or damp it?

It is a very sophisticated brush. The real art is in making this brush, as well as in using it (by running queries and picking the best results)

cool opinion bro, you should rate hollywood bullshit with 0-4 stars for a living

It's possible to become more than just an opinion.

We already have legal rulings to this effect with banning an AI as the inventor in a patent https://www.epo.org/law-practice/case-law-appeals/communicat... .

once we create consciousness, I think that consciousness will be able to create art. whether it will choose to is a different question.

We already know how to create consciousness, it’s called "make babies" ;)

When two grown ups love each other very much...

Lack of consciousness or intentionality seems like a strong claim.

I'll give you that the art it generates is an imitation of art, but what do you think 99% of art that makes any money is? Go to the art section of target or IKEA, look at what hangs on walls in hotels and offices. All that stuff can now be automated with more uniqueness and at a lower cost than before.

Decades ago the art world traded skill (lifetime of practice, apprenticeship, in relative poverty) in the reproduction of realistic forms on canvas and in marble, for novelty (Piss Christ, Duchamp’s Urinal, Basquiat’s graffiti). DALL-E will just precipitate the next stage I think.

Edited for typo.

This is the same argument that would conclude that Picasso was a more skilled artist at 15:

https://mymodernmet.com/wp/wp-content/uploads/2016/12/1-pica...

Than at 35:

https://mymodernmet.com/wp/wp-content/uploads/2016/12/6-pica...

And that he was an even worse artist at 90:

https://mymodernmet.com/wp/wp-content/uploads/2016/12/13-pic...

I make no judgement as to the validity or quality of the art, either classical or modern. I think art is much too personal a thing to speak of in quite that way. Only that the art world has decided to make the trade I speak of. Someone who can produce photo realism today, will not set the art world on fire. Neither will someone who takes 6 years to produce a single large anatomically precise sculpture after 25 years training. However if you can inject novelty (I recently read of a statue made of the artists own congealed blood, and another made of their ear wax) then you have the attention of the patron.

I would say that he produced better art, rather than that he was more skilled, because he could've produced work below his level of skill.

You seem to be describing "technical skill", like "can he move his pencil in a straight line". Essentially, can he draw the things he intends to draw on the paper?

I'm talking about "artistic skill", which is the ability to effectively put ideas or emotions into a work of art.

An analogy which might explain my thinking: A software engineer who writes a 1,000 line program in a month is producing work beneath their "skill" because they could've written a 100,000 line program that does the same thing, in that same month. Typing is a technical skill for programmers, but I wouldn't say that typing skill makes you a better software engineer.

I'm talking about the same skill you are. You (and many others) think that the third picture requires a lot of artistic skill. As far as I see, the third picture is a prototype with a lot of technical debt.

I don't think we can meet on common terms because our definitions of art are different. I do not subscribe to the opinion that art is about challenging our thinking. Instead, I consider good art to be about immersion. You do not have to convince yourself to like it. You do not feel the need to like it ironically. Yes, you can get used to it and it loses some of it's power, but the greatest art is the art that you can keep looking at (or listening to etc.) without feeling the need to justify it. (I cannot remember which philosopher I borrowed that idea from.)

Photography killed the market for realistic forms on canvas being the pinnacle of achievement.

The art world has always been about novelty, but what was considered novelty centuries ago is now considered as "classic". There are countless examples of artists that didn’t get much recognition because what they did was too "modern" for the times they lived in (see Caravaggio for example). Most of the things you see in museums are notorious because they were extraordinarily novel for their time, not because they would be hard to reproduce.

I think this meme sums it up perfectly: https://media.discordapp.net/attachments/931832635758641158/...

I feel like with this scene there will be more emphasis on curation since the tool rapidly speeds up creation/synthesis. Coming up with clever prompts and sifting through the output to find the gems will still be a valuable artistic skill.

I sympathize with the sentiment of op, even if, hopefully, they are proven "wrong" with time. If an ai could produce an endless parade of masterpieces does it devalue the masterpiece? In some sense I think the answer must be yes.

Creativity is going to be digitized and automated and the artists of the future will all be programmers. That goes for all mediums; movies, sound, games, art etc.

The average consumer does not care where it comes from as long as its entertaining, full stop.

Here's a test for you. Is it human or computer generated?

https://thisimagedoesnotexist.com

Honestly who cares if it's art or not, it's cool

(Note that I posted this, not because I agree with the article, but I thought it might lead to some interesting discussions and debate here.)

Does anyone else remember wikihow.com and its bizarre and surrealistic illustrations?

I'm pretty sure DALL-E could get a job there.

A lot of people just don't understand what art is, for the sole reason no one taught them that. The stereotype of “art” that “naturally” (i. e. out of general neglect) emerges from compulsory education process is not that different from anecdotal 19th century opinion of an average educated person on that, as whole education systems stem from that era and keep its sensibilities (like scientific reductionism, militant materialism, and so on). Left alone, people stick to informational, ideological, or decorative functions of works of art (see examples in other comments).

You can point to the crowds in famous galleries, but what do they do there? Move quickly through the halls as if competing to increase their ppm (paintings per minute) stats, when in fact you need to stand and think in front of each of them. It's the requirement for both “avant-garde” and “realist” works, no matter how old and “simple” they are. What exactly is “realism” anyway? Is this “reality”?

https://commons.wikimedia.org/wiki/File:Ilya_Repin_-_%D0%9A%...

Of course not, it's a virtual scene constructed by the artist to convey certain thoughts. And that's a very recent example. Many Renaissance paintings now require a commentary that is longer than those needed for contemporary works.

What can one see in, say, Gollum example in the article? It is collection of cliches, either recognized (the composition of the figure of a scribe with a book has been used in countless illustrations) or artificial (boulders? why this connection?). An artist, as opposed to an illustrator, would distance from that, making, say, a humorous image of an excited Gollum filling page after page after discovering his own talent, and poor hobbits carrying long stacks of paper (well, my approach is not that original either).

It's not hard to imagine that computers would sooner or later generate cheap fantasy novels (as people do now), add cheap illustrations (as people do now), and make low quality translations to other languages (as people do now). Then what is exactly new here, apart from production rate, and what does it all have to do with art?

How are you guys getting access to DALLE-2? Is it still that you have to know somebody who knows somebody?

There's a waiting list.

> DALL·E 2 is a research project which we currently do not make available in our API. As part of our effort to develop and deploy AI responsibly, we are studying DALL·E’s limitations and capabilities with a select group of users.

https://openai.com/dall-e-2/

the sad part about DALL-E is that it is going to be used extensively to sell NFTs.

Well if you consider coding to be art then this statement would be false

> Art used to be something to cherish

> Now literally anything could be art

> This post is art

Art is what you can get away with

yawn

> Without taking into account the consciousness of the artist, the word “art” loses all meaning, becoming merely a synonym for “beautiful.” (...) Without the intentionality of the artist taken into account, the definition of “art” is bled of all meaning

<rant>

Oh, so now you care about art having meaning?

Ever since I understood why I didn't like most of what's labeled "modern art" I've been complaining about the art world not demanding artists to write down what their work means. "Why did you nail a banana to the wall?" "That's for you to figure out". "What does this all-blue canvas mean?" "Whatever you want it to mean".

Getting the intentionality of an art piece out of an artist is quite often like pulling teeth. And I don't ask about grand statements - "I thought it looked pretty" would be just as acceptable as "it's a critique of classical art's over-reliance on technique over message". But even that is apparently too much to ask.

DALL-E works have no intentionality and it shows. But how is that any worse from what we have now, where the artist just shrugs and walks away with no one calling them out?

</rant>

Modern art galleries often have text with what the artist said about their work, or commentary on the theme and symbolism.

I tend to prefer the art that doesn't need it.

Reply on news.ycombinator.com