Hacker News Reader

Comment by quenix | Hacker News Reader

Reading about advances in AI art is always bittersweet. Generative text-to-image systems have come very far in the past 2 years. Impressively so. Frankly, I am in awe at the (cherrypicked!) outputs on this page.

Years on, it's still a little hard to fully grasp the imminent, momentous impact this (has yet to have?) on commercial artists. I fear it will become pretty much impossible to make any sort of living off art in the next decade.

I mean: the outputs on the page are just awesome. Leagues ahead of the stuff we have now. And I'm already seeing old-gen generated AI images on corporate blog posts—everyone will jump on DALL-E 3.

Being in the profession right now must be very discouraging indeed. My heart goes out to those artists who will eventually be replaced by cheap, intuitive text prompts.

Commercialising art for corps was one of the last ways to exist as an artist in today's economy and get by. I fear the extinction of the profession will have a big impact on our cultural capital.

View on HN

> I fear it will become pretty much impossible to make any sort of living off art in the next decade.

Software developers have been having technology completely eat their work out from under them since the dawn of the industry. But jobs aren’t being lost over it; there’s more demand for software developers than ever. When software advancements reduce the work needed to produce output, the world has moved on by taking that as the new baseline which software developers build upon and demanding more software to be built with bigger and better features.

Back when I first started working, it was somebody’s entire job to take pages and pages of content and transcribe them into HTML manually so that they could be published on a website. People used to do that all day, every day. Then CMSs came along and completely eliminated that work. Now sure, if a developer decided that all they wanted to do was write static HTML and refused to adapt their skills, they would be out of a job. But we all used the new technology to build more dynamic websites that provided more value. The new technology didn’t take away our jobs, it provided an opportunity to do a better job.

It’s the same with this. These aren’t tools to replace artists – these are tools artists can use to do more and better work. These aren’t tools that will reduce demand – when everybody can get on-demand, totally custom artwork, people will want more of it, not less.

This certainly seems to reduce how much experience and study an artist needs to have in order to produce art. And I think it's likely to reduce how much the viewer values the product as well.

Automation does not always increase demand for the jobs being automated. Tractors didn't create more demand for oxen than ever.

The automation of tractors over manual plowing didn’t replace the Oxen, but the people driving them.

Farm automation allowed a single man to plow many fields in the time one field would have taken.

Automation allows one person to accomplish more with less.

Tractors didn’t reduce the demand for product, but increased it many fold.

Automation increases demand; every time.

The number of people working in agriculture was divided by c. 6 in the US [0], despite the population being multiplied by c. 3.5.

Sure, automation increased demand for food, but demand did not increase enough to not destroy 5/6th of farmers jobs.

Now this may not be a bad thing at all, that's not the point, but I am not sure "hey demand for art will increase but 5 out of 6 artist jobs will no longer exist!" is a super hopeful statement to artists.

[0] https://ourworldindata.org/grapher/number-of-people-employed...

> Automation increases demand

Yours is a different definition of demand than the economics definition, which is basically the amount customers will buy as a function of price. Higher demand means buyers will pay a higher price for the same art.

Automation increases supply, which increases the volume of art consumed, mainly because it is available at a lower price. That is shifting to a different point on the same demand curve, and doesn't mean that demand has increased.

Automation increases demand.

Just look at the dairy industry. Pre-automation sales numbers vs. post-automation sales numbers.

This is true for every industry on earth, and doesn't care about the pedantic definition of demand. Consumer demand = more purchases. More purchases = more revenue.

What's the alternative? Outlaw AI art? Too late for that.

It's inevitable, but that doesn't mean we should delude ourselves into thinking it's win-win for everybody.

Who's deluding themselves?

To refresh your memory, the context of this thread is that JimDabell argued that software automation increased demand for software programmers, and the same will be true for art. I responded that there's no guarantee of that, and then you pointed out that it can't be stopped. Which may be the case but it has no bearing on the question of whether artists will lose their careers. Just because it's inevitable doesn't mean it's not a cause for legitimate concern.

But despite the work, it still takes a programmer to glue all of our crap together even if that crap is mostly preassembled. That's our fault and frankly a failing of our industry but it's why automating most of programming doesn't mean that your team can get by with 0 programmers.

Clip art/stock photography though, you can do that with 0 artists if you have the actual art bit automated.

>you can do that with 0 artists

you could do that with zero artists for decades because it's already fully commoditized. Everyone who is cheap already just buys a subscription to some gigantic vault of stock imagery owned by some digital rights holder for pennies.

the mechanical reproduction of art did not start in 2020

It's a good point, I'm not a consumer of that market so I don't know it well and am projecting my lack of even rudimentary mspaint skills that in that situation. I wouldn't be able to remotely customise something I found in the gigantic vault of stock imagery so I'd need somebody to help me with even that, but presuambly somebody with a recurring need for stock art probably doesn't lack those skills to the extent that I do.

That's true, but I still worry because it feels like we are reaching some kind of inflection point. For instance, Vercel's latest product (https://v0.dev) makes me nervous about the impact to front-end development.

What's interesting about seeing these conversations on Hacker News is that 10 minutes ago, our profession was putting thousands out of work on a daily basis without even a second thought.

Ever stopped at one of those self-checkout registers at the grocery store that didn't exist in your childhood and wondered what happened to the cashier who used to be there?

A lot of people mumble something about "jobs not worth doing", but I'm not sure if web front-end development (how many unique, differentiated, worthwhile websites exist? how much of that code is glue and boilerplate, and people solving the same problem in parallel every day?) fares much better in that gauntlet.

Software developers right now are, for the first time in their lives, experiencing a little backlash to their privilege -- for hundreds of millions, they've been riding that rollercoaster all their lives.

> Ever stopped at one of those self-checkout registers at the grocery store that didn't exist in your childhood and wondered what happened to the cashier who used to be there?

They're still there, they're restocking shelves.

Similarly, the invention of ATM machines increased the number of bank tellers.

My local supermarket normally seems to be staffed by maybe two or three people total. They stock shelves until someone needs them to fix the machines, do an age check or temporarily use the old checkout line. I remember supermarkets of that size used to have several people stacking shelves and then the same number again on the tills. So I think it has reduced the number of those jobs.

I feel embarrassed to admit that I've never thought about it from that perspective. Totally agree.

You're right, and I'll gladly bite that bullet and say that the changes you describe are good. It's almost always a net positive to be able to produce more output with less human effort.

I’ve been slowly warming up to this idea, and I’m not sure I’m totally convinced, but I think it does make sense to compare the web development industry to the art industry here.

Have web builders totally replaced developers? No, not at all. It’s definitely cut out the lowest end of the market, but that’s because your local restaurant really doesn’t need more than a squarespace template.

But warping a website builder to do something more complex is complicated. To the point where you need to hire someone trained in using that builder tool. Then that person finds the edges of the customizability of that tool and reaches for just writing HTML/CSS/JS.

I could totally see a similar trajectory here. We’ll have prompt engineers, and they’ll realize sometimes photoshop is just easier to tweak the output, then maybe they’ll realize it’s easier to prompt for components of an image, and composite them manually… and then we’ve reached a point where it’s just a tool that makes art easier.

Though… it still does feel somehow different from web development. Will just have to see how things shake out.

I wish this were the case but I'm not really sure I buy it. I'm a software engineer with absolutely no eye for design or illustration. Now I can write a prompt to get me a good-enough image in 90% of cases whereas before that was totally unreachable without an artist. I don't need any training to use this tool and in many cases it gets me to the full end result without any modifications needed.

The reverse doesn't hold -- there are no tools that allow an artist with no training or study to be an effective developer. I'd say the CMS example is more of a data entry job than a development job now, it's two different things.

Most of the advancements in tech still need trained developers to utilize them. GPT-4 cannot create non-trivial programs itself and does not seem particularly close to doing so, it's mainly used for scaffolding and bug fixing and guiding, all of which still need a trained developer at the wheel.

I'm worried for the future of artists.

Were you using artists for anything you can use DALLE for now? Probably not, because what they're best at is too different. But making them more similar will actually increase demand for them because you'll have higher standards for whatever you get out of them. (Jevon's paradox)

Basically, there is no circumstance under which automation reduces demand for labor. It's like if you got a pay raise and started worrying it might make you poorer.

> Were you using artists for anything you can use DALLE for now?

I mean sure yeah, in the past I would pay for logos from artists and now I just make them with AI for my projects instead.

> I'd say the CMS example is more of a data entry job than a development job now, it's two different things.

Right, but it used to be a development job, which is the point. The CMS took a development job and turned it into a data entry job -- exactly what DALL-E does to some limited forms of artistry.

But development didn't disappear, and neither will artistry.

There are two different questions:

- will be more artwork be bought (probably)

- will the increase in the quantity of artwork being bought (assuming there is one) compensate the decrease in price for each piece of art?

Clearly for software, _so far_, the increase in productivity / decrease in price has led to such a big increase in demand for software that some programmers are better off - I say _some_ because many jobs within the software industry stopped existing.

Yet if we look at something like agriculture, there has been some increase in demand in food products(e.g. much more meat is being sold), but one can only eat so much - so most of the increase in productivity has led to way fewer people working as farmers, and not much increase in farmers income.

No "these are tools to replace farmers - these are tools farmers can use to do more and better work".

Basically people are making statements about the elasticity of demand with respect to prices for prices & quantities no one has ever observed. If making a piece of art is now 1/2 (random number) the cost thanks to AI, will people buy 50% more? 100% more? 1000% more? I have no idea and I am not sure why people think they do.

The fine art world, which is typically what people mean by "art", has basically nothing to do with generative AI and probably never will. It's an entirely different set of metrics.

What you're talking about is illustration, which will indeed have a difficult time in the future.

> The fine art world, which is typically what people mean by "art"

Maybe in your circle, but that's certainly not the case generally.

Art galleries? The art world? These are pretty universal terms and they refer to the system of museums, art fairs, galleries, etc. that have nothing to do with commercial illustrations.

The book illustrator is an artist who works in publishing, the art gallery painter is an artist who participates in the "fine art" world.

I've never heard these terms used like you are using them.

https://en.wikipedia.org/wiki/Art_world

See also: Art Forum. ArtNews. MoMa. The Met. Etc.

I didn’t invent the term.

"Art world" is a term connected to "fine art" not "art" or "artists". That's where we disagree.

>Maybe in your circle, but that's certainly not the case generally.

Great, now that's solved, let's agree on a definition of 'punk' next

Some “artists” I know are scraping by on work that could be eaten by generative AI. Even if they don’t self identify as “illustrators”.

We can argue all day about the definition of Artist. My point is that for artists, painters, sculptors, etc. that show their work in galleries and museums, generative AI will have about as much effect as digital art: a little, but not much.

Maybe in the short term. In the longer term, who is going to bother developing these skills? Not everyone “hits a home run” and is successful in galleries and museums, but there have always been fallback options like commercial illustration. If those options are gone, then it’s either 1) be good/lucky enough to succeed in galleries/museums 2) starve or 3) be independently wealthy before becoming an artist.

I mean, the vast majority of artists operating in the gallery system are not moonlighting as commercial artists. It’s an entirely different skill set.

I encourage you to check out the work at museums and galleries in whichever city you live in. It should be immediately obvious that these aren’t really skills that transfer to Photoshop and blog post images.

art/fine art, will only be at risk when ai learns to physically paint or interact with the world. I think it needs to actually use a paint brush. Pictures of paintings are never the same as seeing them in person. They just can't be captured the same way. Ai is being trained on photographs of these physical paintings.

a lot of art isn't always so much about the technique but the content. ai might make the gap from beginner and crap to pleasing to look at in a technical sense, but that doesn't mean the art it self will be interesting or meaningful.

if you really think art as a whole is at risk of being destroyed by ai you should go to your nearest high quality museum and really think if ai would make that. i know like half the posters are in the bay area lol, go to sf moma this weekend and think about this.

Yes, your opinion seems like the obvious truth to me and I don’t quite understand why this is eluding people.

probably just people commenting on fields they dont have experience with

> art/fine art, will only be at risk when ai learns to physically paint or interact with the world.

So you just need to pair DallE with an axidraw?

it would need to understand shading, layering, strokes etc.. and how utilize them to get pleasant effects.

idk how a computer would deal with the unpredictability of a medium like water color. axidraw looks more like a printer thats using a pencil instead of ink.

skip to 36min in here, https://www.youtube.com/watch?v=OokWxQPU5kw

In the first (cherry-picked, as you say) example, the man's moustache is doubled, the telephone handset is in two places, and his sideburns are on sideways.

It would be a terrible outcome if authorship and illustration are mostly reduced to editing and touching up errors in AI-generated statistically likely art.

(Although oddly, for programming, I'm really looking forward to that outcome).

This is going to open new doors for future artists, at least in the near term.

I can easily imagine a future where a single dedicated individual or at least a very small independent team can make a full-length movie without leaving their apartment on a tribal budget that rivals a big Hollywood production costing hundreds of mullions to produce today.

Right now you've got the writers striking, worried that the studios are going to replace them with AI, I think this is totally backwards, it's the studios who should be worried because the barriers to entry that protect them today are about to come crashing down.

I am already seeing a few people make short films using these tools. Right now anything "AI generated" has a certain novelty factor, like back in the day when 3D CGI was new and we were all rendering chrome 3d spheres and shiny red cubes and cylinders on black and white checkerboards, this phase is going to pass soon enough. Perhaps that surreal Midjourney glow will be part of 20's nostalgia in the decades to come. There's a whole new set of skills a new generation of artists are going to master and do things largely unimaginable a year or two ago. They're going to make art that expresses their own perspectives and ideas and not just what's currently allowed by the current consensus, just as the artists before them did.

> I can easily imagine a future where a single dedicated individual or at least a very small independent team can make a full-length movie without leaving their apartment on a tribal budget that rivals a big Hollywood production costing hundreds of mullions to produce today.

This is definitely exciting to contemplate, but it's still quite tricky for the economics: Right now, the fact that you need 4000 people to make a blockbuster movie limits the number of movies being made, giving each movie enough of an audience (potentially) to make their money back.

Better tools will enable more awesome creatives to make content alone - but will there be enough eyeballs to consume that content, even if it's cheap to make?

We went through this with YouTube. The answer is yes. There's more than enough demand for stories that the current movie industry isn't interested in telling to sustain millions of cottage-industry movie or TV shows producers.

Artist don't create art to make a living, but to experience the joy of creation.

Making a living from art is something that the majority of artists even before GenAI did not reach.

I agree that it will change things a lot, but fearing the "exctinction of art" is a bit dramatic IMO.

I think you're defining a very particular type of 'artists', if you broaden the definition then think of graphic designers or concept artists, now thier job has been completely automated.

if you broaden the definition then think of graphic designers or concept artists, now thier job has been completely automated.

Sure, but will that have an effect on art? Most artists I know have day jobs completely unconnected to their art. Even the ones that draw stuff for a living don't consider that drawing their art. After spending all week working in the "drawing factory" they spend their weekends working on their 'real art'.

The only argument I can see is that fewer people will be willing to really learn the craft of drawing/illustration etc. if the chance of getting a day job doing that diminishes greatly.

I've heard more than once an artist say doing art is the only way to keep themselves from committing suicide. Artists gonna do art.

Another example would be someone who loves to garden. If a gardening robot was invented those people would still garden themselves.

Generative AI allows people without mechanical skill to do art, by iteratively interacting until the AI gets it just right for them, so in that sense generative AI opens up art to many more people.

As for making money from art, I believe we're going to have to implement some sort of universal basic income. I currently don't have to work because I collect social security, so I pursue my hobbies the way most other people on UBI will do.

Why do we even want to contribute to a culture so austere and calculating and uncaring anyway? Like, kind of a chicken/egg situation.

It is all just impetus to go and find the prior form of "art" we have had all along. Art need not be about the Artist and her Works, or about creating something for some kind of vague consumption. Art can be about collectivity and shared truths. Art used to be about speaking to God or whatever, and whoever actually placed the pieces of glass were very secondary. At my most optimistic, I feel like some turn like this (but probably more secular) is inevitable. People need to get stuff out, and if this pressure is not relieved by our current ideas of labor and such, it will find a way nonetheless like water through stone.

http://web.mit.edu/allanmc/www/benjamin.pdf

My guess is this will devalue art overall. That is indeed momentous it’s been a corner stone of modern culture for about 500 years. What does a world without art look like.

Does a new form of “art” evolve that makes use of these seemingly omnipotent brushes?

Sadly, I feel it has already started to devalue art. I find myself unironically wondering whether some pieces are AI-generated on my feed. When I can't tell, I mentally devalue the image.

I'm becoming more and more redpilled on the generative art stuff. The risk-benefit analysis checks out in favour of LLMs, but maybe art is something we can keep our Markov chains off of.

I think it will devalue visual art the same way photography has devalued photorealistic paintings

It’s simply a new tool for artists to create art with. It will change art but I doubt it will destroy art. Like you said, a new form of art will evolve from it.

We'll still have theater.

Film/video already devalued theater and generative video is marching right behind generative imagery.

> Film/video already devalued theater

Is that true? Are there fewer live music shows, plays, live comedy, etc... than there were before cinema and television?

In total, probably not but per capita?

Looking forward to the day I have a little stage in my living room where little androids act out scenes from Shakespeare, Beckett. Ha I bet Beckett never saw that one coming.

Alternatively, it creates the oppertunity for research into different art styles, such as the ControlNet optical illusion spirals that went viral on Twitter/X last week: https://arstechnica.com/information-technology/2023/09/dream...

Those spirals got backlash for being "too easy to do", but others have done more creative things with the same technology: https://www.reddit.com/r/StableDiffusion/comments/15wdoqa/be...

That's the importance of accessibility, and has happened all the time in art history (e.g. photography did not kill painting on a canvas)

Graphic arts have always been about the money, so I'd guess whatever happens on the technical side will have little effect on the art market in general.

For instance the emergence of extremely good reproductions didn't affect the price of the original paintings. Same way the evolution of photoshop doesn't have much effect on the price of picture prints.

Now stock photography could be impacted. But otherwise generative AI only opens the door to more picture production, and won't be put in competition to the traditional painting/prints market/direct artist support market.

> What does a world without art look like.

As an aside I always find it funny when 'art' is used as short form for 'pictures', especially in the world's context.

> I always find it funny when 'art' is used as short form

I think, given the context being an generative AI for images it should be fairly obvious what I’m talking about.

I’m sure it’s only a matter of time before they do it for music, dancing, fighting and whatever else you care to tag with the term.

I dunno. The value of the type of art that is a "corner stone of modern culture" (i.e. the type of art you see in museums, etc.) doesn't seem to be related to how hard it is to produce.

For example, Flag by Jasper Johns sold for $110M in 2010 [1]. The fact that almost anyone could make something like this (the subject matter is public domain, and the technique is within reach of anyone who has taken an art class) doesn't seem to have diminished its value.

And as for all of the art you see in the everyday world, being able to design it more easily would likely lead to more of it, not less. Instead of seeing the same 20 things that they carry at IKEA over and over, you'd have custom artwork everywhere.

[1] https://en.wikipedia.org/wiki/Flag_(painting)

EDIT: Added the last paragraph while you were replying.

> i.e. the type of art that you see in museums, etc.)

Nope. I mean the visual art we see all around us. If it’s trivial to generate it’ll be everywhere and nobody will value it. It will become a annoyance to people - so what fills that space then? Bare walls? Or will things go to a new level. Was kind of my line of thinking.

It becomes all about the artist; something that can not be replicated. What the art says about artist, what the artist intended, and the mere fact that a specific artist created it. Photography played a part in engendering this transition to "modern art" in the last century.

I have paintings hanging in my house produced by artists I personally know. If anything, I think ubiquitous, throw away and infinitely reproducible digital art makes my paintings even more special to me (although none are worth all that much to start with).

Until we see successful video games/movies/tv shows built using 100% generative art I will remain sceptical that this will happen. What's more likely is that generative AI art will be its own category like digital art, where creative folks will use their creativity to create amazing works of art with the help of generative AI. After all, it takes creativity to know what to generate. You can't ask the AI to make you a whole movie or set of video game levels.

100% might be a high bar for just skepticism. If there was something massively successful that was even 50 or 60% AI (and the novelty of AI wasn't the sole reason for its success), I think I would be firmly convinced.

It might increase the preciousness of art made with physical media. Such real art will be seen as the real deal, handcrafted, etc.

I think the market for affordable prints and copies will definitely suffer, because I can create prints of Midjourney art that looks as good as anything I see in our small local galleries and gift shops.

Artists would have to find creative ways to separate themselves from AI-generated art.

Also, it would be nice if there was something akin to a watermark that was perhaps invisible but upon inspection using a certain tool will reveal if it was created using the Generative AI model (similar to how real currency can be inspected and differentiated from fake currency notes)

All the AI generated images still make my head hurt when I look at them for any length of time. You start seeing oddities with shapes and lines, or some bizarre errant merge happened and my brain struggles to comprehend it. There's also a sort of "natural messiness" hand made images have that I don't think AI will ever capture.

i feel the same way, but for big money movers, unless you can put it on a spreadsheet, it might as well not exist.

Would you pass a blind test? I'm pretty sure I wouldn't.

For now, all we know is this technology replaces the human capacity of drawing with good technique what a human wanted to be drawn. I think the relevant part of art is not the ability of generating an image, but to put thought on what you want to be there.

The language on this page also suggests they want to put the prompt engineers out of a job.

Luckily AI art will most likely be in the public domain so artists will probably always be involved but you'll become some editing cog in a machine trying to change a piece of ai art enough that it becomes copyrightable.

There's still a significant lag time before uncensored generative AI catches up with this. So gracious of OpenAI to keep porn artists employed a little longer with all their moralizing.

Seems like a problem for generative systems too though, fewer artists creating original artwork results in less input and variety to generate work from.

Images aren't all created by "artists", and you don't need to train on images created by "artists" in order to produce something that could be called "art".

I don’t really see how this has any impact outside of digital art. People are still buying paintings even though photos exist, and still buying sculptures even though cnc machines exist.

I also think the output isn’t yet good enough to be usable without some human intervention (fixing hands, spot treatments, etc.)

That’s just until AI combines with CNC to produce unique sculptures. Which won’t take long.

The impact has already happened, this will accelerate, paintings and even sculptures can be now bought dirt cheap online as it's being mass-produced by machines in China. Now fine art, graphic design etc are next to go.

Probably being produced by factory workers, not machines. (It's like clothing - clothes are often cheap now, but every one of them was man-made.)

And Chinese wages are rising, so it'll eventually be too low-level for them.

Reply on news.ycombinator.com