This entire article misses critical developments in the AI art space.
Controlnet was just released last week. It allows for precise fine-tuning of the image with skeletal positions, depth maps, outline sketches, etc.
It has been exceptionally received in the AI art community, because pure-text-prompting precisely runs into the issue described in the article. You can only convey so much information with text prompts.
That being said, artists in general still don't want to touch controlnet. Because despite controlnet solving their complaints about how the AI isn't controllable precisely, it doesn't solve the real problem, that drawing skills are massive devalued.
Artists will still exist, but most likely as hybrid 3d-modellers, AI modelers (Not full programmers, but able to fine-tune models with online guides and setups, can read basic python), and storytellers (like manga artists). It'll be a higher-pay, higher-prestige, higher-skill-requirement job than before.
And all those artists who devoted their lives to draw better, find this to be an incredibly brutal adjustment.
PS: Despite how people made fun of 'proompters', or predicted that prompting would be automated away. The skill ceiling to good AI art has radically increased. There's now 10 different fine-tuned models you need to learn the basics of, each with different strengths. There's thousands of LORAs to insert into prompts to precisely reproduce a subject that the base model has no information on. There's 20 parameters to tune, each with different effects. There's VAEs which affect coloring and fine-details. Now there's controlnet, and you'd better learn blender to rig basic skeletons to feed it.
This likely suggests the future of an AI-augmented economy: People become AI wranglers. Wordpress was supposed to make blogging easier, instead it spawned an industry of WP wranglers.
This is the critical insight. Human artists having nothing to worry about. Human drawers do.
The decoupling of concept from execution is old news in most forms of art. Writing music and playing an instrument are two very different skills, and indeed any session musician will tell you that playing instruments well is not as valued as writing amazing music.
In fact we went through a similar thing in pop music in the 70’s and 80’s, with the brief moral panic over electronic music and people who “just push a button and don’t even know know to play their instrument.” Turns out nobody cares, as long as the music is good.
AI art decouples concept from execution for visual art. That’s all.
You're not taking into consideration how the actual artists actually feel. And, from the first hand - the feeling is horrible. Most artists love to draw, the process itself is incredibly satisfying. Just like a musician loves his instrument.
This is destroying them, stealing the joy from the thing they devoted decades of life. My wife is extremely depressed by this, to the point I think she'll need serious therapy. She still has most of her job, but yeah recent developments like ControlNet? Well, shit.
"This is the critical insight. Human artists having nothing to worry about. Human drawers do."
The reason why amazing music is valued so much is because there's so much music that you have to be amazing to be noticed. Art has had that problem for a long time. Someone spends 100hrs of talent on a masterpiece and we say "meh, seen a thousand of that quality"
Learning theory is easier than applying it so becoming an artist is easier and now artists have more competition. Not advocating against AI art but it's obviously going to have a negative effect.
Even human drawers might do all right, if they're selling physical drawings. People will still put hand-painted canvasses on their walls, just like people still enjoy an acoustic guitar concert, even if the guitarist isn't world-class. The authentic connection with a human being is where the art survives. If you're not an amazing artist, you can still be an authentic one.
But if you're just churning out commercial illustrations, then sure, an AI can do that now, or soon.
> Turns out nobody cares, as long as the music is good.
Actually, all those people who lost their professions they spent their lifetime working on, they care, rather a lot.
> indeed any session musician will tell you that playing instruments well is not as valued as writing amazing music.
Yes, in 2023 they say this, but I assure you when "session musician" was a common profession, that no one said this.
When I first started in music, if you wanted your piece of music, you had to pay instrumentalists to play it, and this was a valuable professional skill.
> Turns out nobody cares, as long as the music is good.
And the music is not good either.
When did everyone become so sociopathically detached from the welfare of others? It's just horrifying.
I'm not an artist, but my fiance is an animator. Generally, something similar exists in stock assets (photos, assets like vector files, other such things). We had a discussion on whether the existence of stock assets means drawing is no longer necessary to be an artist, and her answer was it still is something you should be able to do because knowing how to draw is the best way to learn the fundamentals of art in general, knowing how parts of a body work in a drawing or animation, and generally something is "good" vs. "bad," etc. However, as someone who isn't a fine artist, she doesn't have time to generate everything frame by frame, so she doesn't do it for every project she works on, but that knowledge is invaluable and one you can only obtain from knowing how to draw.
I think the best analogy I understand it is as is it is like assembly language or low level programming in general. I certainly do not have the time to program every piece of code I write in asm but having done projects in asm is invaluable as a coder[0], given how much it informs my mental model of how the code I write actually works. That understanding is beyond valuable and is something that puts you a rung above everyone else who just copies things from SO without knowing what they do. I think AI for devs, to the extent that it will evolve, will still be as such. People who primarily find it amazing today I find either 1) use it as a productivity boost for things that need a lot of boiler-plate[1], or 2) are SO-copy-pasters who are just amazed they have to think even less. A LOT of the "AI artist" community are the artist equivalent of the 2nd, honestly, and it's easy to detect AI art because it's generated by people who are not really artists, and in similar fashion either don't know the fundamentals or who only create "good" work by nearly directly copying other art pieces.
Also, to critique your analogy, what you guys are saying is along the lines of "given DAWs, why learn to play an instrument at all?" Except a plain look at any good composer will show they know how to play at least one instrument, even if they don't play all their music on that instrument, there is no doubt knowing how to play clearly makes you a better composer. I mean, can you even imagine a composer who cannot even play piano, or guitar? Sure a composer need not be a virtuoso concert pianist, but they should at the very least be able to play the chords of the very song they've composed. Nothing thus far teaches the human mind deep understanding of something more than doing that thing does. It is so clearly obvious in music and programming and the only reason people keep saying you can be an artist having never learned to draw is because such people simply do not understand art or composition at all.
[0] I'm not really a software developer but a computational scientist, and so I won't say I'm a developer, but still understanding of data structures, algorithms, discrete math in general, and yes asm is invaluable for me.
[1] As someone outside of the dev space, I don't know why you people don't just make better interfaces instead of juggling 9 yamls and 3 environments for every project, or just roll your own interface and reduce the boiler-plate yourself.
> Because despite controlnet solving their complaints about how the AI isn't controllable precisely, it doesn't solve the real problem, that drawing skills are massive devalued
At this point if I was an artist I'd probably be looking for a way to leverage my existing skills into AI art, because it seems like drawing skills will go the way of the horse and carriage.
I feel bad because this has happened so quickly. Previously it seems like there was a longer transition period.
This kind of transition has got me thinking the same about coding though. If AI programming does indeed take over, my programming skills will likely be massively devalued.
If you believe this is going to happen, how should you prepare for it? Start learning prompting now? Get involved with building these AIs? Etc.
I've tried to convince artists to transition to the AI-augmented future, since October.
There's almost no successes. I've seen artist communities on multiple websites, in multiple languages. At first they tried to laugh at the AI. Now the AI has improved so radically so fast, they universally prefer to stick their head in the sand, and ban discussion of AI altogether, its all denial and rage.
On the more optimistic side. I see an extraordinary explosion in artistic innovation, just look at websites like CivitAI. The massive community all training subcomponents of the models for each other to share. The models rapidly improving every month just through fine-tuning and theoretical innovation, without stabilityAI's involvement (They are distracted by lawsuits now).
There are many 3d-artists intensely experimenting with AI art, to say make AI-anime, which has illustration qualities on every frame (A previous impossibility due to the costs involved).
It seems with AI, it'll really cleave communities in two. The ones who eagerly embrace it, seem to enjoy it extraordinarily, and achieve quite a lot of popularity and success. But the rest just want to pretend it doesn't exist, waiting till employers realize that they are no longer needed.
Regarding programming, it doesn't appear that AI programming can replace humans. Programming is very similar to novel writing in terms of complexity for AIs. And AIs are still extremely terrible at long-form storytelling. The lesson is to aggressively use AI tools as much as possible, to understand the long-term weaknesses of AIs, and deliver your values in those areas as a human.
> seems like drawing skills will go the way of the horse and carriage.
The exact same thing happened to technical drafters when CAD destroyed the entire (sub-)industry.
And to typesetting, for that matter[1].
Which is a pity in the human sense and the sense that both were forms of artistry and produced things of great beauty, both the product and the machines used to enable them, but they're simply not economical in the face of Solidworks and digital composition. And yet the replacement technologies have also enabled a lot more creativity and further advances. Objects with complex geometries are now possible to specify and manufacture, when they previously could not even be accurately drawn.
1: https://en.wikipedia.org/wiki/Wapping_dispute, where not only could the unions not simply oppose the rising sea-level of technology, but their failure seriously damaged union credibility in general.
> At this point if I was an artist I'd probably be looking for a way to leverage my existing skills into AI art, because it seems like drawing skills will go the way of the horse and carriage
Drawing skills were made redundant for producing most types of high quality image before the motor car but people still pay for hand drawn items
Not sure AI lowering the skill barrier for and speeding up the generation of digital art is really going to change that
> my programming skills will likely be massively devalued
I see this in the context of the drawing vs. art analogy. Yes, your typing and maybe syntax skills will be devalued, but your higher level, creative programming skills are probably safe for a good long time yet.
>If you believe this is going to happen, how should you prepare for it? Start learning prompting now? Get involved with building these AIs? Etc.
If you still want to work as someone who produces code - except your personal code factory has changed from brain and fingers to AI - then yeah, you should probably do both of those things.
Even if you spend 1-2 years and this AI hype doesn't work out, you still have those skills to go back to.
Frankly, this mainstream adoption by Google and Microsoft is, uh, not going great. So you can afford to observe for now, but AI advancements have been made very rapidly, so it's not wise to completely ignore it either.
> If you believe this is going to happen, how should you prepare for it? Start learning prompting now? Get involved with building these AIs? Etc.
Short of advocating for artists' unions...
Law is still going to govern anything AI creates, so anyone who develops a skillset in both art and law would be well-positioned to gatekeep as a copyright troll. This carves out a career for yourself and enacts revenge by making it a liability for employers to use AI to replace artists.
HR is the theocratic version of that, since law school isn't cheap. You can still gatekeep, but you'd be playing by more-arbitrary rules ("you can't use that AI because it incorporates images of Women Without Penises," etc.).
Security is pragmatic, but I worry about the long-term stability of it. Every time a breach is announced, there are no consequences, so why even pretend to need it?
I used to suggest pivoting to tangible works and experiences (architecture, sculpture, etc.), but those are easily displaced as well-- I would bet against it now.
Look at porn-- what started as a handful of performers in studios is now done by any college student with a webcam in their bedroom (next up to be replaced with AI-generated content).
For now, you can still add value as a sex worker by offering the GFE, but even that's about to be obsolete. You can get your fix of flirting from a chatbot (TTS or sexting) and 3D print a copy of anybody else's genitalia. Remote-controls, vibrators and fluid pumps add some life to it. And you don't even have to leave the house, which affords you privacy to pursue darker subjects without oversight and save you more time for repeat consumption.
Go with the flow. AI will need humans to be effective, humans will need AI to remain competitive. But everyone will have the same base models, just like we all have the same web search and electricity. AI won't be a competitive advantage, it will be a basic requirement.
Do you believe the number and complexity of software applications will decrease in the next 10 years because of AI, or that it will spawn whole new ecosystems of software and new types of jobs? I believe the second is more reasonable, we will have higher expectations from software in 2033 than in 2023. The easier AI makes it, the more difficult we make the tasks.
Human desires fill the available space like air, AI exponential is slower than our entitlement. So we still need to work.
I believe this is going to happen (and is happening right now, in front of us) not just for programmers but for any role that can be automated.
> How should you prepare for it?
Get equity. I mean ownership stake. If you don't have a legal/financial claim on the output of the machine then you're about to be part of the worthless surplus from the POV of the system.
> Start learning prompting now?
No, the machines will do that well enough in a minute or two.
> Get involved with building these AIs?
No, the machines will do that well enough in a minute or two too.
- - - -
Artificial Intelligence destroys scarcity, which is the fundamental basis of our societies and economies.
Think about it: scarcity is the very problem that societies and economies evolved to solve in the first place.
Now science and capitalism have delivered technology and wealth. There is enough to go around if we just worked out the logistics, and computers can do that for us in a matter of moments. In other words, the "World Game" is not hard! https://en.wikipedia.org/wiki/World_Game we just have to get over our hangups.
Now here's where it gets really interesting: ChatGPT et. al. don't have glands, they don't have emotional trauma, no PTSD from being humans-on-Earth for generations, etc.
We can program them to be sane and perhaps even wise.
We can also attach empirical feedback devices to them, make them scientists...
So we have physical abundance and benevolent, sane, empirically-grounded AI advisors, how much longer will it take to sort things out? I think we could be looking at the start of a Golden Age?
Yeah this post seems like it will not stand the rest of time.
The fundamental conceptual shift is the emergent behaviors displayed by these models. And that is highly unforeseeable.
We’re not working with machines that do what they’re built to do. They’re doing more. And every day we discover something new they’re capable of doing.
Making any sort of statement about the limits of this technology is going to be shortsighted.
The opposite is also true. People overestimate what is possible and underestimate how much more effort and time is required to start replacing industries.
> Artists will still exist, but most likely as hybrid 3d-modellers, AI modelers (Not full programmers, but able to fine-tune models with online guides and setups, can read basic python)
Why should artists have to learn fucking Python!? There are so few jobs open to people who don't want to "learn to code", this was one of the last, and that's going too.
> This likely suggests the future of an AI-augmented economy: People become AI wranglers.
So we don't get to create ourselves anymore. No drawing, no making music, no programming even. We just give prompts to AI programs. Which means we are worthless, because any bozo can do this.
I hate this timeline so much. It's like all the worst and most mediocre parts from every SF future, with an extra dose of stupidity and greed on top.
I wouldn't worry too much. We've pushed the biosphere past the point of repair. This "utopia" they dream of won't have a chance to come to fruition before we all go down in flames.
And given this dichotomy as choice, I would choose the flames too.
At the end of the day, we are not going to see AI art in the Guggenheim, or any of those other museums where the finest art in the world resides. AI art is and will always be a novelty. Real art comes from the human spirit. It is not about technical execution, and not just about painting a pretty picture. In actual art we want to see humanity, something that AI will never, ever truly know about.
Seems like a lot of people are missing this point, and it seems like a lot of people have a chip on their shoulder about being unable to create art, and believing they can express the deep recesses of their feelings and psyche and everything else that makes us human by proxy through AI, and thus completely missing the point of creating art in the first place.
These are great platitudes, and as a fellow human I like the idea that only we can create "real" art, but I don't know how to justify that idea. I have no chip on my shoulder about art, it's not part of my identity at all. Computer programming is, though, and I fully expect it to affect my industry in the next decade.
The current fine art market is extremely unmeritorious so I agree that AI art will never make it to the Guggenheim. That isn't a good metric since the human curators will never allow it to happen. The metric will be the industry. Will ai start composing scores for movies? Promotional posters? Art that's sold outside of expensive curated galleries (etsy, ect)?
It's possible that humans will keep their stranglehold on these and I hope that's the case... But I wouldn't bet on it.
I think even without AI art we are well past the point of art in famous galleries making sense. Isn't most of this just thinly veiled money laundering anyway?
I imagine we WILL see AI art in a famous gallery for exactly this reason. Except it will be lauded for the imagination of the creator and the tweaks they made, not their drawing ability.
> This likely suggests the future of an AI augmented economy: People become AI wranglers. Wordpress was supposed to make blogging easier, instead it spawned an industry of WP wranglers.
Good counter for "AI will steal our jobs". When competition starts to use AI+human, you got to level up. And since everyone has the same base AIs, the differentiating factor is still human.
Artist here: Article is an idealistic take not based in reality- the Jeff Koons of the world will be safe with their 30 million balloon art sales.
Everyone else is going to be rekt.
All of the work my artist friends do is for businesses- businesses trying to make the biggest profit. Ai art generators are faster/cheaper/more flexible. The artists are done and now it's all about the prompt engineers that don't need to be artists to excel.
Market will decide what to pay "artists" but seems like it will be a race to the bottom. Artist types I know are preparing accordingly.
But you should realize that any artist learning prompt engineering should be a better AI-artist than a non-artist learning prompt engineering. After all, art is not jsut about technique but also creativity, composition, theme etc., right? And an artist can furthermore touch up the AI generated product manually if necessary.
Yep. I spent a good hour trying to generate some stuff with DALLE. I then asked a friend that works in media and he was able to give me ideas for a prompt that worked almost immediately.
AI art is fun and _sometimes_ it can replace some things but at the end of the day, I don't really want to mess around prompting over and over. If I pay an artist $100 for something, I couldn't care less how they make it. I feel like AI can be another tool. I'll even make a bold prediction. In 10 years, we will still have "Graphic designers" and "digital artists" the ceiling will just be raised similar to how it was when photoshop became more commonplace.
What leads you to believe those problems won't be solved in the next few years? What stops me from having ChatGPT provide feedback on my prompts to make them more artistic? Maybe I just pay some artist to touch my stuff up for me if they're desperate for work or maybe the output gets good enough that I don't.
Out of interest, how are they preparing accordingly? It's not remotely my own field but I'm starting to wonder how I should be guiding my children who have tallent for art, in light of all this.
Draw comics, or really manga in particular.
The only real challenge in AI art, that won't be solved any time soon, is storytelling: chaining multiple panels together for coherent stories.
Even ChatGPT cannot tell a coherent story longer than 2 paragraphs. Being able to write a long story well, requires understanding world modelling, human motivations , etc, AGI tier abilities.
Also, expose them to AI-art. If they lose interest in art after seeing AI art, that means they were never meant to be artists, they merely like to draw, not to make art. And drawing alone is not really economically useful anymore.
AI is shifting the balance of power from technical ability and quality of tools to creativity/vision (and ability to market). The only "art" job left is going to be something akin to an art director.
Expose your children to a wide variety of art, books, music, food, travel and so forth. Teach them why (aesthetically) good things are good and bad things are bad. Also, encourage them to create stuff for an audience so they learn how to present things, gauge people's tastes and become comfortable with failure young.
The harsh reality as far as I can tell is that the prospect of making a living as an artist, which was already slim if we are being honest, has now shrunk to near zero.
They should be learning about the AI field in order to create their art. That might be the only skill worth money when they are older.
Focusing on learning how to leverage these new tools to produce their own works faster/better/cheaper- focusing on their original ideas/designs/models- coming up with strategies to protect their source art from being used in other people stable diffusion models- learning about/utilizing nft tech to find new pathways to monetize their art-
In the end though its going to be all about original ideas imo- so creatives that don't have original IP really need to get on that and develop it- the new tools will allow anyone to create anything in realtime- people will be frozen by the prompt if they don't have original ideas/concepts/characters/worlds etc-
The temptation for those without original ideas will be to leverage chatGpt etc for "ideas" but humans are not required on that path-
So for your children- ignore the tools/technicalities and focus on ideas- original ideas-
And older artists said the same about computer art, things like :" this new artists do not know to mix paints or prepare a canvas".
Artists can adapt and use this new tools and do their job faster, you might have to sell a logo cheaper but you might be able to make them 20 times faster.
There will just be people who are employed to wield these tools, select, curate and combine these images into whoever commissioned the work.
I don't know who all these tools are really aimed at? As a software engineer, I'd still prefer to pay someone to mess with any type of "art" or design while I focus on other things.
I mean even if I had an AI program that I could use to code, to do SRE, to do images, to do accounting etc, I still think someone would need to be in charge or making these things happen, or else I'd just be busy prompting machines all day, which sounds mad fatiguing and boring.
I think the real question is what does "making these things happen" realistically pay, pennies on the hour to sit in front of a screen and maybe elevate that to the 1/10,000 person who has to come hit a few keys to get the station running again?
On the other hand, where was the backstop previously? Companies could have saved money on art by hiring unpaid college interns or teams from where the median wage is less than a few hundred dollars a month. Evidently there was some reason to favor hiring actual artists versus the cheapest possible person who can draw. Maybe that reason hasn’t died out yet.
My point is that unique digital art won't be replaced even by a hypothetical perfect AI image generator.
When it comes to the digital art workforce - it is likely to be strongly affected.
> Others might fall into a gap of “nice skills, but not yet that offer a business advantage”. Furthermore, the lower entry barrier to create any art is likely to result in the average quality going down - not unlike that plastic made manufacturing cheaper, but also less durable.
And you are right that, sadly, in many kinds of digital art, it will be a race to the bottom.
There's a mistaken assumption here. The author seems to believe that these generators operate on the basis of random chance, hence a 200x200 pixel image in RGB has a possibility space of 10^300000 images.
...But an overwhelming majority of those images are random noise, and are effectively homogeneous and interchangeable. They're perfectly identical.
Like human artists, the image generators are (very strongly) biased to create "structured" images, which are derivative of existing artistic works or natural representations in the form of photographs. In principle, there's no a priori qualitative difference between the AIs and humans in this respect -- no human art is created de novo, but is always a continuation of, or a reaction to, existing forms.
Further, there's no reason to believe that (a) the size of the possibility space matters in a quantitative sense, and (b) that AIs will never be as capable of qualitative "originality" -- in conception, composition, or subject matter representation -- as a human artist would be.
I put my caveats on indistinguishable images. And three is the challange - to make a better estimation of an effective number of distinguishable images.
> (b) that AIs will never be as capable of qualitative "originality" -- in conception, composition, or subject matter representation -- as a human artist would be.
Sure, AI already creates original things. The point is that questions like "create a drawing that presents friendship" can be approached from many, many different, and what is crucial - subjective, ways.
They are not worried about making all artists redundant, they are worried about making most artist redundant.
Instead of needing to pay for mediocre artist to get mediocre art that's good enough for a purpose, now the AI taught on the good artist (all subjective of course) can produce similarly mediocre art that's fit for purpose.
It's equivalent to replacing all of the simple CRUD web devs with AI that does good enough job
Interestingly, it might be easier to replace mid-skill artists than mid-skill developers (not to say both won't happen over time). The key difference is that when generating art, you can say "Make A in-the-style B.", get what you want, and move on.
Meanwhile, you have to live with the consequences of AI generated code.
Case in point: My wife was trying to build a system to generate POs per vendor from a series of workorders a couple weeks ago. She was using AirTable, which vastly simplifies the creation and use of relational databases. Given a lack of database experience, the structure that she made was misaligned with what she actually needed (and completely denormalized). Putting it into production would have caused major problems over time.
This made me realize that writing code is only one of the barriers to entry in software. Eliminating this barrier effectively gives everyone in the world access to a small team of incredibly junior developers who have no idea what they're doing.
Just like mismanaged DIY home projects create more work for professional contractors, I think AI might -oddly enough- create more positions for mid-tier and higher developers.
It most definitely is for simple reason - non interactivity. Image doesn't need to be told what clicking this button does or how that pop up menu needs to animate or million other interactions.
> Just like mismanaged DIY home projects create more work for professional contractors, I think AI might -oddly enough- create more positions for mid-tier and higher developers.
It might also create a lot of jobs for "AI crafters" - know enough about it to craft good prompts about what client wants and have enough graphic editing knowledge to tweak and mix AI input to create what is needed. It probably will also empower "one man shops" dealing in small customized websites for customers, replacing or augmenting what now stock images are used for.
Reading the comments it seems that everyone assumes that art is always and only presented as pixels on a screen. Whereas most of the art in my house is hand painted on canvas and wood, hand embroidered, hand knitted, hand woven, hand thrown and decorated pottery. One picture was hammered out of a sheet of pewter.
One of my favourite places is the Henie-Onstadt Art Centre sculpture park, another is Vigelands Anlegg in Oslo, no pixels in either place.
So why does everyone seem to think that art is only flat, lit, pixels on a flat surface?
I think the exact same thing, obviously because we're on hacker news where most people seem to be hoping they can spend their days prompting machines for art so they can replace "artists" and save some money.
It's actually funny when I think about it because most people wouldn't know what image to use and how to use it even when it was generated for them, there's even a skill in selecting art.
You could argue enough good photos have already been taken that all practically all photographers should have already been made redundant since the year 2000, we still have photographers.
I think if people are ignoring AI's limitations, it's more out of fear that programming is next than having something to gain. I doubt having the money/labour that's currently used creating art for something else would help me in a noticeable way:
-the money would just trickle up
-not that much money is spent on it anyway
-world becomes more depressing, nothing you look at had any effort put into it, no-one had to believe in an advert on any level, instead it's the output of a machine optimised to trick you (as one of a shrinking number of people with any agency) to spend and therefore make the machine stronger.
-artists/potential future artists decide to learn to code instead?
That reminds me of Ira Glass' quote about taste: artists know it when they see it.
Or why it's a mystery some folks prefer tabs vs spaces, or recoil from what they consider shoddy code.
Somewhere in the discipline of point, line, and perspective, in the composition of shapes and their organization, is the artifact of amalgamated neurons, to be observed by yet another consciousness.
I wonder if AI art will just help generate more gacha games. And one may wonder, of the limited time left on this planet, what really could we spend it on?
The library of every book contains no meaningful work in the search. The space of all generated art is oblivion.
Because that's what 99% of the people employed for their artistic skills are doing.
I doubt AI will have much effect on the tiny number of people who can make a living by producing art to hang up in your house. It's the people doing illustrations for magazines/websites/packaging/etc. who are fucked in the medium/long term.
People will 100% start hanging AI art on their walls. The killer feature is being to customize the artwork to exactly what you want. It's just a question of time and not a lot of it either.
>So why does everyone seem to think that art is only flat, lit, pixels on a flat surface?
Everyone doesn't think that. It just happens that digital art, both as a medium and as an industry, is the only kind of art relevant to conversations about AI, because that is the medium of artwork that AI generates, and that is the industry that is being disrupted by it.
Yes, that is interesting: so digital artists can go back to traditional methods. Like a game where you scan in crayon textures. It's a unique look. More handcrafted.
Maybe AI will bring about a sea change in how we express these traditional works in the digital context.
It's true we miss a huge chunk of art by just considering pixels. Museums are dedicated to the idea that art and expression are intertwined, and that context--both past and present--brings a unique experience to the observer.
Curators certainly aren't generating those longform descriptions next to art pieces. Someone had to think deeply, analyze, and type that out.
AI will make many, but not all, artists redundant.
As usual, the lower skill/value work will be automatized sooner. Platforms like fiverr will probably suffer a lot, but so will many "for hire" artists. High value work will remain to be exclusive for humans.
Many high value artists (programmers ...) started as low value and worked themselves up. But if the low value segment is eaten up by AI then the path will be destroyed as well. I wonder if this will make classical education more valuable (after graduation you're starting as high value producer) and further increase class division.
I don't think education has anything to do with this. AI tools are pretty easily accessible to anyone and you don't need a "classical education" to wield them.
If anything, I think the necessity of education is going to be weakened further and more quickly in the next decade.
> I wonder if this will make classical education more valuable (after graduation you're starting as high value producer)
I haven't noticed a correlation between formal education and the people that I would describe as "high value producers". That could be a bias because the median person without a degree doesn't get hired at all, but if we're removing mid-tier and below devs from the work pool, that's going to eliminate most of the people I know with CS or SE degrees.
> I haven't noticed a correlation between formal education and the people that I would describe as "high value producers".
I agree, but I would argue that people without formal education have to usually work their way up.
Basically, if I'm a high school drop out, then I'm very unlikely to land a job at FAANG right away. Quite possibly I will have to start in some low value job to get some credibility. But if AI eats most of these jobs, then it also removes the possibility for the high school drop outs to move up.
Only society can make artists redundant. If people want to get their "art" by digitally rehashing what real artists did in the past then we don't need artists.
How exactly things will play out is not clear: photography did not eliminate painting. Algorithms will not eliminate more "manual" creative work. Some new genres might emerge.
Ultimately what is a more important problem is that the commercialization of artistic production was always challenging. When you can create infinite replicas with semi-random variations it only makes the problem worse.
Exactly. Art as an economic industry may die out. As a career, it may be reduced to a much smaller scope. But we'll always be able to express ourselves artistically - maybe even more so if AI frees up more time in which to do so.
...or enjoy making art with other people to use in our commercial endeavors ?
Like, it might actually be fun to work with creatives to make art. It's one of the areas of my work I actually really enjoy. I can't imagine prompting DALL-E to be as fun and enjoyable, nor would it be "creative".
While people are (to varying degrees) generous, its hard to build everything on generosity and recirocity. On the other hand once you start bean counting and transactionaling everything and in particular not even accept different types of beans, out entire existence becomes the one-dimensional, money driven disaster it has become.
Many artists are really scared by what these apparently uncontrollable interests are unleashing. Who can blame them.
I wouldn't say thanks to information theory, but rather on human tendency to be bored with repeated styles. It doesn't matter how good the tech is, but if it comes out like Michael Bay movies or anything else that we learn to recognize, people are going to want 'more original' art. The other difference is that the point of art is to express and evoke emotions, it's an open question whether that can be done effectively without having feelings during the process.
This is a good take and under appreciated generally. Already theres an identifiable generic flavour to much AI art. Just knowing that it’s AI generated makes it feel somewhat lifeless. Not because of the end product is lacking , but because of the awareness of the production method. it’s less engaging if you know a machine spat it out, especially once you recognise and become bored by the stylistic markers that indicate machine generation.
That will play into the spending calculation of, for example, advertising agencies, particularly high end brand work.
I’ve seen some screenshots of ai ads already. They are usually somewhat horrying, like the person will be smiling, but you start to notice their mouths are open too large and they have too many teeth and lack eyelids, and the people in the background are even more distorted and nightmareish.
I wouldn't make that strong a statement anymore. The level of play by AlphaGo/Zero seems to demonstrate creativity and understanding that we would only attribute to humans. That is to say I can't say for certain that emotion couldn't be faked, and we don't know that human emotion is super-special, only that it's complex.
Readit News
That being said, artists in general still don't want to touch controlnet. Because despite controlnet solving their complaints about how the AI isn't controllable precisely, it doesn't solve the real problem, that drawing skills are massive devalued.
Artists will still exist, but most likely as hybrid 3d-modellers, AI modelers (Not full programmers, but able to fine-tune models with online guides and setups, can read basic python), and storytellers (like manga artists). It'll be a higher-pay, higher-prestige, higher-skill-requirement job than before. And all those artists who devoted their lives to draw better, find this to be an incredibly brutal adjustment.
PS: Despite how people made fun of 'proompters', or predicted that prompting would be automated away. The skill ceiling to good AI art has radically increased. There's now 10 different fine-tuned models you need to learn the basics of, each with different strengths. There's thousands of LORAs to insert into prompts to precisely reproduce a subject that the base model has no information on. There's 20 parameters to tune, each with different effects. There's VAEs which affect coloring and fine-details. Now there's controlnet, and you'd better learn blender to rig basic skeletons to feed it. This likely suggests the future of an AI-augmented economy: People become AI wranglers. Wordpress was supposed to make blogging easier, instead it spawned an industry of WP wranglers.
This is the critical insight. Human artists having nothing to worry about. Human drawers do.
The decoupling of concept from execution is old news in most forms of art. Writing music and playing an instrument are two very different skills, and indeed any session musician will tell you that playing instruments well is not as valued as writing amazing music.
In fact we went through a similar thing in pop music in the 70’s and 80’s, with the brief moral panic over electronic music and people who “just push a button and don’t even know know to play their instrument.” Turns out nobody cares, as long as the music is good.
AI art decouples concept from execution for visual art. That’s all.
This is destroying them, stealing the joy from the thing they devoted decades of life. My wife is extremely depressed by this, to the point I think she'll need serious therapy. She still has most of her job, but yeah recent developments like ControlNet? Well, shit.
The reason why amazing music is valued so much is because there's so much music that you have to be amazing to be noticed. Art has had that problem for a long time. Someone spends 100hrs of talent on a masterpiece and we say "meh, seen a thousand of that quality"
Learning theory is easier than applying it so becoming an artist is easier and now artists have more competition. Not advocating against AI art but it's obviously going to have a negative effect.
But if you're just churning out commercial illustrations, then sure, an AI can do that now, or soon.
Actually, all those people who lost their professions they spent their lifetime working on, they care, rather a lot.
> indeed any session musician will tell you that playing instruments well is not as valued as writing amazing music.
Yes, in 2023 they say this, but I assure you when "session musician" was a common profession, that no one said this.
When I first started in music, if you wanted your piece of music, you had to pay instrumentalists to play it, and this was a valuable professional skill.
> Turns out nobody cares, as long as the music is good.
And the music is not good either.
When did everyone become so sociopathically detached from the welfare of others? It's just horrifying.
I think the best analogy I understand it is as is it is like assembly language or low level programming in general. I certainly do not have the time to program every piece of code I write in asm but having done projects in asm is invaluable as a coder[0], given how much it informs my mental model of how the code I write actually works. That understanding is beyond valuable and is something that puts you a rung above everyone else who just copies things from SO without knowing what they do. I think AI for devs, to the extent that it will evolve, will still be as such. People who primarily find it amazing today I find either 1) use it as a productivity boost for things that need a lot of boiler-plate[1], or 2) are SO-copy-pasters who are just amazed they have to think even less. A LOT of the "AI artist" community are the artist equivalent of the 2nd, honestly, and it's easy to detect AI art because it's generated by people who are not really artists, and in similar fashion either don't know the fundamentals or who only create "good" work by nearly directly copying other art pieces.
Also, to critique your analogy, what you guys are saying is along the lines of "given DAWs, why learn to play an instrument at all?" Except a plain look at any good composer will show they know how to play at least one instrument, even if they don't play all their music on that instrument, there is no doubt knowing how to play clearly makes you a better composer. I mean, can you even imagine a composer who cannot even play piano, or guitar? Sure a composer need not be a virtuoso concert pianist, but they should at the very least be able to play the chords of the very song they've composed. Nothing thus far teaches the human mind deep understanding of something more than doing that thing does. It is so clearly obvious in music and programming and the only reason people keep saying you can be an artist having never learned to draw is because such people simply do not understand art or composition at all.
[0] I'm not really a software developer but a computational scientist, and so I won't say I'm a developer, but still understanding of data structures, algorithms, discrete math in general, and yes asm is invaluable for me.
[1] As someone outside of the dev space, I don't know why you people don't just make better interfaces instead of juggling 9 yamls and 3 environments for every project, or just roll your own interface and reduce the boiler-plate yourself.
At this point if I was an artist I'd probably be looking for a way to leverage my existing skills into AI art, because it seems like drawing skills will go the way of the horse and carriage.
I feel bad because this has happened so quickly. Previously it seems like there was a longer transition period.
This kind of transition has got me thinking the same about coding though. If AI programming does indeed take over, my programming skills will likely be massively devalued.
If you believe this is going to happen, how should you prepare for it? Start learning prompting now? Get involved with building these AIs? Etc.
On the more optimistic side. I see an extraordinary explosion in artistic innovation, just look at websites like CivitAI. The massive community all training subcomponents of the models for each other to share. The models rapidly improving every month just through fine-tuning and theoretical innovation, without stabilityAI's involvement (They are distracted by lawsuits now). There are many 3d-artists intensely experimenting with AI art, to say make AI-anime, which has illustration qualities on every frame (A previous impossibility due to the costs involved).
It seems with AI, it'll really cleave communities in two. The ones who eagerly embrace it, seem to enjoy it extraordinarily, and achieve quite a lot of popularity and success. But the rest just want to pretend it doesn't exist, waiting till employers realize that they are no longer needed.
Regarding programming, it doesn't appear that AI programming can replace humans. Programming is very similar to novel writing in terms of complexity for AIs. And AIs are still extremely terrible at long-form storytelling. The lesson is to aggressively use AI tools as much as possible, to understand the long-term weaknesses of AIs, and deliver your values in those areas as a human.
The exact same thing happened to technical drafters when CAD destroyed the entire (sub-)industry.
And to typesetting, for that matter[1].
Which is a pity in the human sense and the sense that both were forms of artistry and produced things of great beauty, both the product and the machines used to enable them, but they're simply not economical in the face of Solidworks and digital composition. And yet the replacement technologies have also enabled a lot more creativity and further advances. Objects with complex geometries are now possible to specify and manufacture, when they previously could not even be accurately drawn.
1: https://en.wikipedia.org/wiki/Wapping_dispute, where not only could the unions not simply oppose the rising sea-level of technology, but their failure seriously damaged union credibility in general.
Drawing skills were made redundant for producing most types of high quality image before the motor car but people still pay for hand drawn items
Not sure AI lowering the skill barrier for and speeding up the generation of digital art is really going to change that
I see this in the context of the drawing vs. art analogy. Yes, your typing and maybe syntax skills will be devalued, but your higher level, creative programming skills are probably safe for a good long time yet.
If you still want to work as someone who produces code - except your personal code factory has changed from brain and fingers to AI - then yeah, you should probably do both of those things.
Even if you spend 1-2 years and this AI hype doesn't work out, you still have those skills to go back to.
Frankly, this mainstream adoption by Google and Microsoft is, uh, not going great. So you can afford to observe for now, but AI advancements have been made very rapidly, so it's not wise to completely ignore it either.
Short of advocating for artists' unions...
Law is still going to govern anything AI creates, so anyone who develops a skillset in both art and law would be well-positioned to gatekeep as a copyright troll. This carves out a career for yourself and enacts revenge by making it a liability for employers to use AI to replace artists.
HR is the theocratic version of that, since law school isn't cheap. You can still gatekeep, but you'd be playing by more-arbitrary rules ("you can't use that AI because it incorporates images of Women Without Penises," etc.).
Security is pragmatic, but I worry about the long-term stability of it. Every time a breach is announced, there are no consequences, so why even pretend to need it?
I used to suggest pivoting to tangible works and experiences (architecture, sculpture, etc.), but those are easily displaced as well-- I would bet against it now.
Look at porn-- what started as a handful of performers in studios is now done by any college student with a webcam in their bedroom (next up to be replaced with AI-generated content).
For now, you can still add value as a sex worker by offering the GFE, but even that's about to be obsolete. You can get your fix of flirting from a chatbot (TTS or sexting) and 3D print a copy of anybody else's genitalia. Remote-controls, vibrators and fluid pumps add some life to it. And you don't even have to leave the house, which affords you privacy to pursue darker subjects without oversight and save you more time for repeat consumption.
Truly, what a wonderful world...
Do you believe the number and complexity of software applications will decrease in the next 10 years because of AI, or that it will spawn whole new ecosystems of software and new types of jobs? I believe the second is more reasonable, we will have higher expectations from software in 2033 than in 2023. The easier AI makes it, the more difficult we make the tasks.
Human desires fill the available space like air, AI exponential is slower than our entitlement. So we still need to work.
I believe this is going to happen (and is happening right now, in front of us) not just for programmers but for any role that can be automated.
> How should you prepare for it?
Get equity. I mean ownership stake. If you don't have a legal/financial claim on the output of the machine then you're about to be part of the worthless surplus from the POV of the system.
> Start learning prompting now?
No, the machines will do that well enough in a minute or two.
> Get involved with building these AIs?
No, the machines will do that well enough in a minute or two too.
- - - -
Artificial Intelligence destroys scarcity, which is the fundamental basis of our societies and economies.
Think about it: scarcity is the very problem that societies and economies evolved to solve in the first place.
Now science and capitalism have delivered technology and wealth. There is enough to go around if we just worked out the logistics, and computers can do that for us in a matter of moments. In other words, the "World Game" is not hard! https://en.wikipedia.org/wiki/World_Game we just have to get over our hangups.
Now here's where it gets really interesting: ChatGPT et. al. don't have glands, they don't have emotional trauma, no PTSD from being humans-on-Earth for generations, etc. We can program them to be sane and perhaps even wise.
We can also attach empirical feedback devices to them, make them scientists...
So we have physical abundance and benevolent, sane, empirically-grounded AI advisors, how much longer will it take to sort things out? I think we could be looking at the start of a Golden Age?
The fundamental conceptual shift is the emergent behaviors displayed by these models. And that is highly unforeseeable.
We’re not working with machines that do what they’re built to do. They’re doing more. And every day we discover something new they’re capable of doing.
Making any sort of statement about the limits of this technology is going to be shortsighted.
Why should artists have to learn fucking Python!? There are so few jobs open to people who don't want to "learn to code", this was one of the last, and that's going too.
> This likely suggests the future of an AI-augmented economy: People become AI wranglers.
So we don't get to create ourselves anymore. No drawing, no making music, no programming even. We just give prompts to AI programs. Which means we are worthless, because any bozo can do this.
I hate this timeline so much. It's like all the worst and most mediocre parts from every SF future, with an extra dose of stupidity and greed on top.
(And I'm a computer programmer.)
And given this dichotomy as choice, I would choose the flames too.
Seems like a lot of people are missing this point, and it seems like a lot of people have a chip on their shoulder about being unable to create art, and believing they can express the deep recesses of their feelings and psyche and everything else that makes us human by proxy through AI, and thus completely missing the point of creating art in the first place.
The current fine art market is extremely unmeritorious so I agree that AI art will never make it to the Guggenheim. That isn't a good metric since the human curators will never allow it to happen. The metric will be the industry. Will ai start composing scores for movies? Promotional posters? Art that's sold outside of expensive curated galleries (etsy, ect)?
It's possible that humans will keep their stranglehold on these and I hope that's the case... But I wouldn't bet on it.
I imagine we WILL see AI art in a famous gallery for exactly this reason. Except it will be lauded for the imagination of the creator and the tweaks they made, not their drawing ability.
Good counter for "AI will steal our jobs". When competition starts to use AI+human, you got to level up. And since everyone has the same base AIs, the differentiating factor is still human.
SD has been great at giving lots of new ideas but frustratingly difficult/impossible to iterate on anything.
Everyone else is going to be rekt.
All of the work my artist friends do is for businesses- businesses trying to make the biggest profit. Ai art generators are faster/cheaper/more flexible. The artists are done and now it's all about the prompt engineers that don't need to be artists to excel.
Market will decide what to pay "artists" but seems like it will be a race to the bottom. Artist types I know are preparing accordingly.
AI art is fun and _sometimes_ it can replace some things but at the end of the day, I don't really want to mess around prompting over and over. If I pay an artist $100 for something, I couldn't care less how they make it. I feel like AI can be another tool. I'll even make a bold prediction. In 10 years, we will still have "Graphic designers" and "digital artists" the ceiling will just be raised similar to how it was when photoshop became more commonplace.
Also, expose them to AI-art. If they lose interest in art after seeing AI art, that means they were never meant to be artists, they merely like to draw, not to make art. And drawing alone is not really economically useful anymore.
Expose your children to a wide variety of art, books, music, food, travel and so forth. Teach them why (aesthetically) good things are good and bad things are bad. Also, encourage them to create stuff for an audience so they learn how to present things, gauge people's tastes and become comfortable with failure young.
They should be learning about the AI field in order to create their art. That might be the only skill worth money when they are older.
In the end though its going to be all about original ideas imo- so creatives that don't have original IP really need to get on that and develop it- the new tools will allow anyone to create anything in realtime- people will be frozen by the prompt if they don't have original ideas/concepts/characters/worlds etc-
The temptation for those without original ideas will be to leverage chatGpt etc for "ideas" but humans are not required on that path-
So for your children- ignore the tools/technicalities and focus on ideas- original ideas-
I don't know who all these tools are really aimed at? As a software engineer, I'd still prefer to pay someone to mess with any type of "art" or design while I focus on other things.
I mean even if I had an AI program that I could use to code, to do SRE, to do images, to do accounting etc, I still think someone would need to be in charge or making these things happen, or else I'd just be busy prompting machines all day, which sounds mad fatiguing and boring.
When it comes to the digital art workforce - it is likely to be strongly affected.
> Others might fall into a gap of “nice skills, but not yet that offer a business advantage”. Furthermore, the lower entry barrier to create any art is likely to result in the average quality going down - not unlike that plastic made manufacturing cheaper, but also less durable.
And you are right that, sadly, in many kinds of digital art, it will be a race to the bottom.
...But an overwhelming majority of those images are random noise, and are effectively homogeneous and interchangeable. They're perfectly identical.
Like human artists, the image generators are (very strongly) biased to create "structured" images, which are derivative of existing artistic works or natural representations in the form of photographs. In principle, there's no a priori qualitative difference between the AIs and humans in this respect -- no human art is created de novo, but is always a continuation of, or a reaction to, existing forms.
Further, there's no reason to believe that (a) the size of the possibility space matters in a quantitative sense, and (b) that AIs will never be as capable of qualitative "originality" -- in conception, composition, or subject matter representation -- as a human artist would be.
> (b) that AIs will never be as capable of qualitative "originality" -- in conception, composition, or subject matter representation -- as a human artist would be.
Sure, AI already creates original things. The point is that questions like "create a drawing that presents friendship" can be approached from many, many different, and what is crucial - subjective, ways.
Instead of needing to pay for mediocre artist to get mediocre art that's good enough for a purpose, now the AI taught on the good artist (all subjective of course) can produce similarly mediocre art that's fit for purpose.
It's equivalent to replacing all of the simple CRUD web devs with AI that does good enough job
Meanwhile, you have to live with the consequences of AI generated code.
Case in point: My wife was trying to build a system to generate POs per vendor from a series of workorders a couple weeks ago. She was using AirTable, which vastly simplifies the creation and use of relational databases. Given a lack of database experience, the structure that she made was misaligned with what she actually needed (and completely denormalized). Putting it into production would have caused major problems over time.
This made me realize that writing code is only one of the barriers to entry in software. Eliminating this barrier effectively gives everyone in the world access to a small team of incredibly junior developers who have no idea what they're doing.
Just like mismanaged DIY home projects create more work for professional contractors, I think AI might -oddly enough- create more positions for mid-tier and higher developers.
> Just like mismanaged DIY home projects create more work for professional contractors, I think AI might -oddly enough- create more positions for mid-tier and higher developers.
It might also create a lot of jobs for "AI crafters" - know enough about it to craft good prompts about what client wants and have enough graphic editing knowledge to tweak and mix AI input to create what is needed. It probably will also empower "one man shops" dealing in small customized websites for customers, replacing or augmenting what now stock images are used for.
One of my favourite places is the Henie-Onstadt Art Centre sculpture park, another is Vigelands Anlegg in Oslo, no pixels in either place.
So why does everyone seem to think that art is only flat, lit, pixels on a flat surface?
It's actually funny when I think about it because most people wouldn't know what image to use and how to use it even when it was generated for them, there's even a skill in selecting art.
You could argue enough good photos have already been taken that all practically all photographers should have already been made redundant since the year 2000, we still have photographers.
-the money would just trickle up
-not that much money is spent on it anyway
-world becomes more depressing, nothing you look at had any effort put into it, no-one had to believe in an advert on any level, instead it's the output of a machine optimised to trick you (as one of a shrinking number of people with any agency) to spend and therefore make the machine stronger.
-artists/potential future artists decide to learn to code instead?
Or why it's a mystery some folks prefer tabs vs spaces, or recoil from what they consider shoddy code.
Somewhere in the discipline of point, line, and perspective, in the composition of shapes and their organization, is the artifact of amalgamated neurons, to be observed by yet another consciousness.
I wonder if AI art will just help generate more gacha games. And one may wonder, of the limited time left on this planet, what really could we spend it on?
The library of every book contains no meaningful work in the search. The space of all generated art is oblivion.
I doubt AI will have much effect on the tiny number of people who can make a living by producing art to hang up in your house. It's the people doing illustrations for magazines/websites/packaging/etc. who are fucked in the medium/long term.
Everyone doesn't think that. It just happens that digital art, both as a medium and as an industry, is the only kind of art relevant to conversations about AI, because that is the medium of artwork that AI generates, and that is the industry that is being disrupted by it.
Maybe AI will bring about a sea change in how we express these traditional works in the digital context.
It's true we miss a huge chunk of art by just considering pixels. Museums are dedicated to the idea that art and expression are intertwined, and that context--both past and present--brings a unique experience to the observer.
Curators certainly aren't generating those longform descriptions next to art pieces. Someone had to think deeply, analyze, and type that out.
As usual, the lower skill/value work will be automatized sooner. Platforms like fiverr will probably suffer a lot, but so will many "for hire" artists. High value work will remain to be exclusive for humans.
Many high value artists (programmers ...) started as low value and worked themselves up. But if the low value segment is eaten up by AI then the path will be destroyed as well. I wonder if this will make classical education more valuable (after graduation you're starting as high value producer) and further increase class division.
If anything, I think the necessity of education is going to be weakened further and more quickly in the next decade.
It isn’t obvious at this point what the skill ceiling is, on using AI tools, since they’ve only been around for a couple months.
I haven't noticed a correlation between formal education and the people that I would describe as "high value producers". That could be a bias because the median person without a degree doesn't get hired at all, but if we're removing mid-tier and below devs from the work pool, that's going to eliminate most of the people I know with CS or SE degrees.
I agree, but I would argue that people without formal education have to usually work their way up.
Basically, if I'm a high school drop out, then I'm very unlikely to land a job at FAANG right away. Quite possibly I will have to start in some low value job to get some credibility. But if AI eats most of these jobs, then it also removes the possibility for the high school drop outs to move up.
How exactly things will play out is not clear: photography did not eliminate painting. Algorithms will not eliminate more "manual" creative work. Some new genres might emerge.
Ultimately what is a more important problem is that the commercialization of artistic production was always challenging. When you can create infinite replicas with semi-random variations it only makes the problem worse.
Like, it might actually be fun to work with creatives to make art. It's one of the areas of my work I actually really enjoy. I can't imagine prompting DALL-E to be as fun and enjoyable, nor would it be "creative".
Many artists are really scared by what these apparently uncontrollable interests are unleashing. Who can blame them.