Hacker News new | past | comments | ask | show | jobs | submit login
Ask HN: What will be the large-scale second-order consequence of AI Art?
30 points by chenxi9649 on Aug 31, 2022 | hide | past | favorite | 36 comments
The main use case that comes immediately to mind when using tools like Midjourney/DALL-E/Stable Diffusion is graphic design/stock photos, and most of the discussion surrounds this theme of making art more efficient(like that SD Photoshop plugin). However, somehow, these discussions give me the feeling of predicting that the main purpose of an automobile is to deliver milk(the faster horse idea).

For example, Photography, something that was only accessible to a few, became massively accessible when the iPhone came out. The large-scale second order was not simply that we had more photos we could reminisce with(the previous major use case of photos). But rather that we started sharing photos/videos with each other as a form of communication/storytelling. (with platforms like IG/Snapchat/Youtube/Tiktok)

This new generation of art tools fundamentally *expands* the set of people who can express their thoughts/feelings visually from a tiny fraction of the population to the masses. Something that used to take years of training is now achievable with a few weeks of practice. This is INSANE! And it's hard to know how this will actually change the world. (Note: I am by no means saying amateur artist + AI == professional artist, just like how iPhone didn't kill professional photographers.)

Will there be a new medium of storytelling? Will the next generation of Wattpad writers become "comic book artists"? Will we start to communicate through generated art that represents how we feel? What do you think can be the world-changing second-order consequence when everyone can express their thoughts and feelings visually?




I'm surprised by the comments on this topic. The art generated by AI is not the product. AI generation is a tool that artists (and non artists to some extent) can use to take away a lot of the tedium. I think of it as AI assisted art.

My bets on second order consequences:

1. 3D world generation based on prompts

2. Ambient music and sounds

3. Character modeling

4. And yes, interactive storytelling.

Imagine getting these models to a decent enough shape and having a 3D printer hooked.

Speech to Text -> Prompt Validation and Cleanup -> 2D Image -> 3D Models -> Printed Object.

Humans add, correct, guide, direct the vision and AI does the gruntwork all through the pipeline.

I wrote more about this in another thread: https://news.ycombinator.com/item?id=32660252


"Tedium" is an odd choice of words.

There will be AI-based music generation soon. Will you say "it takes away the tedium of playing music"? Playing music is a joyful activity. Learning to play it can be tedious but the performance itself is pleasant.

People learn to play a song and then they play it hundreds of times throughout their life even though they could just record it once and let the recording do the "tedious" work of making the sounds!

Drawing is also pleasant. It isn't "gruntwork", though there are tedious tasks within it sometimes. Mostly it is meditative.

The fact that you don't like to draw and you want something that will draw for you is a whole issue in itself.

Would you want something that will dance for you? Ski for you? Flirt and make love for you? Don't worry, you'll be "directing the vision".

It sounds like you don't want to make art, you just want credit for art having been made.

Like, there are people who will use this tool to do amazing stuff, but they're going to have to spend months or years "tediously" learning how to make it do more than what it does out of the box. You know that, right?


> take away a lot of the tedium

What AI takes away is all the hard work and application of talent that happens between articulating "ghost ship flying over futuristic city under moonlight" and having a quality picture of exactly that before your eyes.

That talent and work, with years of practice behind it, is the bulk of what art is.

Not just the talent to execute, but creativity. The AI has a deep imagination, or a facsimile there of, which is being relied upon. For the same prompt, it can generate a vast stream of completely different pictures incorporating different concepts in different ways.

There is still tedium: deciding which of the 200 such pictures that were generated for you do you go with.

Evaluating the output of AI and tweaking prompts and parameters isn't art, but more like constructive art criticism.

If a person A were giving prompts and changing requirements to another person B who produces successive refinements of a picture, B would be readily identified as the artist and not A.

Here is how an artist could use AI: use it to generate ideas, and then incorporate them into their own work.


Likewise, I was surprised by the pessimism in this thread surrounding generative art. Humans are extremely visual creatures, yet creating visuals is incredibly difficult. I hope this can drastically increase the number of visual storytelling in the world through the means that you mentioned.


The latest season of Westworld had a great vision om what prompt based character modeling could look like.


Explosion in indie animated series. Going to go from $15-$20 million for a season of an animated show (requiring industry buy-in, expectation of profit, etc.) to something 2-3 people can make as a passion project in a year. As with indie games and YouTube channels, a couple of these shows will become cultural gems.

First to market with an easy-onboarding decent-looking AI animation platform will build a self-perpetuating position. Think like Roblox for animators.


Vague embarrassment ten years on when you look at something from 2023 that’s filled with obvious machine-made art, the same way looking at old rave fliers full of MANDELBROTS!!! feels faintly embarrassing.


I don't see why it has to be embarrassing. I view it as a form of retrofuturism, an artifact of how society and its expectations for the future at that time influenced art.

The current ai art world is shockingly reminescent of new minecraft updates; everyone using the same new blocks in their builds, so much so that block pallettes are comprised of blocks that do not go together at all, al for the sake of "making a build with the new blocks"

I think experimenting with new art techniques in this way is a fundamental part of the human experience, and i see no reason to cringe at it, even if it means that you can sometimes tell that a minecraft build was made when the update that included its blocks came out.


Is there a term for this anticipatory future embarrassment? Some kind of German conjunction?


German here. I made this up, but I think it fits: vorausschämen

A quick search confirms that other people were already using the term. :)


How about “prediscomfitation”


Some thoughts I get once I think about this.

All marketing and design stuff will move to cheap labour countries. The company in the emerging markets will employ 1000s of people who will get the business context from people .

They will employ all these people with AI tools to get creative and generate multiple options to be sent back to main office.

To make better AI, they will need better data. I can see many Mturk kind of gamification and crypto level stuff driving millions of people to tag data .AI will get multiple different inputs for a same image and then AI can keep improving on top of that.


I think the tools and techniques for developing AI-assisted art are gonna explode, and we will see things like "Photoshop for AI-Generated art" or "Blender for AI generated models", where there are mixed tools that allow touching-up AI-generated art with more AI-based tools.

I think this will absolutely change the content space. We will see smaller groups of individuals able to make animated-films, add-special effects to their video-projects, etc. Get ready ready to see more content geared for the fringes of society, and thus amplification of their preferences in media, politics, etc.


First of all, there will be a phase of abuse. Remember when clipart was freely available and people "expressed their thoughts and feelings visually"? Every presentation had clipart, lots of it. Then people realized that's bad taste. Remember animated gifs in the 90's? Absolutely abused in websites. Remember 3D and embossed icons? Oh yeah, now we are abusing minimalism and invisible delimiters... and purple. All will later become bad taste. We just go to extremes until we are fed up.


It seems that we are already seeing that happening with people using Midjourney/Dall-E photos on their blog posts and all of them have this particular "mythical/psychedelics" vibe. In the early days of IG, all photos also had that tint filter.

However, there does seem to be a correction phase that comes after that leads us to an ultimately better place. So I am quite optimistic about what comes after with generative art.

When the reason for use goes beyond the initial novelty factor.


I do not get the fascination with AI "art", unless one is programming and tuning the AI themselves to accomplish some expression. Art is about human expression, and I'm not for sure typing words into a text box to get a picture generated from data is much expression.

So from my perspective, I think it's a fad that'll die out or ultimately be abused, just like everything other machine learning toolset out there. But who knows? Humans and fads can remain irrational longer than I can stay solvent.


> Art is about human expression

It was previously, because we were the only ones capable of it. As a counter-point, it's not hard to imagine aliens creating art.

The definition that specifies human origin is now outdated.


And we still are; computers are designed, software is designed.

“AI” as we know it does not exist without humans pushing it forward; romantic notions our inventions are independently sentient depends on ignoring a whole lot.

For the same reason I can choose to not import a religion I can not import the idea these stupid things are “thinking” independently of humans, without justifying it as I don’t have to with religion as well.

All the constraints are defined by us. It’s not magic.


I agree it's not magic. I don't think art requires magic.

As a pretty handy example, photographers are considered artists. I took photography pretty seriously at one time, and it's very dear to me, but I will say that I don't think there's very much magic in it.

I don't think the algo is thinking or sentient or anything like that. It's hard to argue that it's not producing art, though.

Tangent: I would also argue humans are designed. I don't mean any mysticism: I think all life is being designed by unthinking processes. As evidence, just look at all the specialist life-forms that are filling a niche that they were...well, designed for.


I am not saying it’s not producing art, I’m saying a human instigated the machine to produce it; it’s a robot with a bunch of brushes in its hand. The difference between that and life evolving to fill a gap in a vacuum is the computer did not do that on its own. Semantics, maybe, but a computer did not evolve until humans did. Fish and moss evolved independently of humans; that’s the difference that distinguishes whether AI is intelligent to me. I would define it as a memory store for contemporary human sensibilities, not a independently bodied entity.

That said, if people want to wax poetic privately in their own jargon, whatever, of course. That’s democracy.

But the refined over thousands of years corpus of knowledge “contained” within STEM fields has been shown over and over to be our most reliable guide humans have come up with for building a society. Nothing in there attempts to define what is and isn’t intelligence; it’s a model of what matter exists and how it coalesces at various “speeds” relative to light. Needing to define if AI is this or that is a romantic human thing; very subjective.

Where a physical thing “is” is hard to argue. Where a thing fits into ones subjective mental taxonomy is easy to argue.


Not sure I'd say these machine learnings tools are capable of art.

And rather than calling the definition of art outdated, it makes more sense to call this machine learning generation stuff "artificial art" or "superficial art" or some other separate term. It doesn't make sense to mix two rather incompatible concepts.


They're obviously capable. Some of the content is indistinguishable from human-created content.

Actually I would say most of it is, since humans can produce bad art as well.

You're worried about the process, and I get that. But the result is art.


I think "programming and tuning" will enable a deeper degree of expression, but just because people are not tuning the AI itself doesnt necessarily mean the creation with AI isn't a form of expression.

Ultimately, is "thinking" not simply prompting your own brain with information(often times natural language) and see what the output is and then we fine tune?

Maybe a question to ask is, can you express yourself if you are not creating the canvas/paint brush? Or, is the Lego creation a form of expression if you are not making the Lego pieces!?


Brian Eno started his [ambient] career doing essentially that, creating and tweaking programming systems to do various musical things, and then letting them to go do their thing. Does this make his ambient pieces, many of which started as a programmatic system, not art? If not, where does the line fall?


Don't forget the human reaction to "art". It can be a derivative, soulless, and thoughtless but still make people feel and think.


See also "Ask HN: Will AI-generated images flooding the web pollute future training data?"[0]

As I mentioned thereon, "I look forward to the 'pollution'"[1]

------------

[0] https://news.ycombinator.com/item?id=32577822

[1] https://news.ycombinator.com/item?id=32581496


This technology can be applied beyond images, to virtually any data modality.

- Video + audio generation (or automatic tweening for animation) means a single animator can create an entire movie or animated series.

- 3d models and textures + automatically generated animation and behavior trees. This could enable artists and non-technical people to create games and interactive experiences.


Cheap designs. There's a viral song going around tiktok right now about a kid who loves corn. Here's a Dall-E generated tshirt design that I'm sure would sell with cleanup and small effort.

https://i.imgur.com/HdT2bWM.png

One can crank out merch much faster for current trends.


"Given the right prompt you can ask for ideas for decorations, architecture, product design, advertisements and who knows what else. It's only a matter of time before it creates 3d images and gets connected to a 3d printer then we'll be able to print some of what we see."

I've written this before but it's relative to this discussion.


The next step will be attaching one of these AI models to a robot that paints, to create unique physical pieces that will be exhibited in galleries everywhere. Much like there are art styles using recycled garbage, this one will use recycled pixels from its training data.


People will look at this and think: If this is possible, what other knowledge work can have its first 80% of work automated?

It'll create free money to go harder for all kinds of other fields.


True human would not die , creating genuine art is more then just writing prompts , requires true endeavour and human experience . Ai still can't compute that .


I for one am looking forward to AI generated RPG games.


That would be cool. But AI isn't consistent enough for a game, right?


The effect of technological innovations is always that things get cheaper and thus available to more people and often that leads to bigger changes. In this case I think you're right about this having a big impact on design; most designers won't lose their jobs but they will have new creative tools and I think they may start to use AI generated art in place of stock photograph. The average joe will be able to create stock photos for their blog posts and websites and maybe that will take away some amount of jobs, but those were probably not very good jobs anyway. Stylization using image to image techniques and inpainting are all tools that could save artists and designers time while also offering new capabilities. The AI tools definitely make it possible for a non artist to make a comic, illustrate a poem or an album cover, or make art for an indie video game. So in those ways more people will get access to things that they might not have before.

I think the next wave will probably be video related AI tools and those will also be really impactful.

At present it still takes a lot of trial and error to create what you want, something that a knowledgeable designer can probably do better than the average person, so I still see a need for design people in the loop in most cases.

I can definitely see people starting to use AI tools for memes - it will be interesting to see if that is as sticky as memes are today. I think a lot of their value comes from recognizable formats, but AI art could definitely provide some interesting variety and new forms of expression.

Fakes and misinformation is still a threat. We have yet to have a big event triggered by a fake AI generated image, but I think that day is coming.

One thing about AI generated art - while you can make some crazy and interesting combinations (something that is fertile ground for artistic experimentation), the AI models can only ever interpolate _existing_ artwork. They can never come up with a new style or something that nobody has ever done before. One could argue that all art evolves from past art, but I think if we only ever used AI tools then creativity would eventually stagnate as the tools could only ever regurgitate a variation on something that has already been done.


When I open up a streaming service or YouTube feed, the sheer amount of currently existing content that I am compelled to click on is endless. I had no choice but to develop a thoughtless reflex to "not recommend" anything novel outside of my current sphere of topics or whose thumbnail screams "content" to me, or my sleep schedule will be perpetually ruined.

Now the opportunity for all 8 billion persons on Earth, regardless of experience, to create their own fascinating, epic, emotionally fulfilling and artistically relevant pieces in addition to the above only gives me knots in my stomach.

I don't understand why an accelerating explosion of ideas that nobody has ever thought of suddenly being realized is supposed to be an unequivocal good. You can have all the entertainment in the world, but human bodies will be chained to physical and biological limits for the time being, namely 14-16 waking hours a day. And some people like me have a much harder time regulating their attention spans as is.

To me, this intense focus on AI-related tooling in the recent past serves as a reminder that if it were possible for humanity to fill every last parcel of physical space with objects of our own creation, we would do so. I think it's just human nature; planting our objects on Mars would only be the first step of millions if it were feasible.

The digital world also suffices. It will be filled to the brim with a set of beautiful paintings of mansions with flowered gardens and fountains, painting, oil on canvas, 4k, detailed, thomas cole, trending on Artstation.

I also have to wonder what the effects will be on the culture of the humans steering the tools. I'm not sure what to feel about the connotations around the concept of "art" gradually coming to mean "the product of those with powerful enough GPUs and ten minutes of spare time a day." I don't know how else to explain it, but feels like something important is lost in the process.

It's like how when you try to practice drawing with pencil and paper, some people tell you that learning a skill like drawing is meant to be a journey, and is a process that requires patience and hard work. The implication is that it's supposed to instill virtue within the creator, so long as you're interested enough to embark on that patient, thoughtful journey for however long it takes. Now it feels like that promise has been invalidated within the span of a few years and a pile of tensors.

My worry is that the fascination with art produced algorithmically will cause people to stay in their rooms, and lose a desire to become more in touch with the outside world. It's common practice to go outside and sketch objects in the real world, using your eyesight and making spatial comparisons to what's on paper. That becomes archaic if the computer can do everything for you.

(I'm also biased in saying this, since I was never capable of enjoying the process of manual drawing long enough to see passable results. It doesn't give me a good feeling to know that I'll never be able to produce art with my hands that an AI would hypothetically never be able to produce.)

One other thing that's only tangential: Assume that any given platform like YouTube could somehow solve its biases and moderation problems, maybe by becoming a public service of some kind. Even with the "best" possible algorithm built for the sake of creators trying to build an audience, what happens when there's now billions of people vying for a slice of your 16 waking hours? An endless stream of fascinating, insightful, fulfilling information that never ends, personalized down to every single fleeting interest and sense of taste you've ever had.

This makes me uncomfortable.

In the present day, the hardest truth for me to swallow is that even without the ills of any given algorithm, what addicts me is ultimately the product of hundreds of other people's passions and drive for a fulfilling life through creating, out of a pool of millions that just happened to miss the mark of my profile of interests. Individually, the act of creation tends to be labor intensive and no single person contributes significantly to oversaturation, but the sheer number of people with the capability of telling their stories begets my perpetual insomnia.

And I'm not sure there's any chance of a solution, if this is something that will even be recognized as needing to be solved at all. I just hope that I never stumble across the perfect stream of content, because once it has been created, I will never be able to take my eyes off it.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: