9 ms·

Back in the 90s I bought a book that came with some floppy disks. The book was about 500 pages of clip art, and the disks were the actual images. At the time,

Back in the 90s I bought a book that came with some floppy disks. The book was about 500 pages of clip art, and the disks were the actual images. At the time, people said such things would put graphic artists out of business. What actually happened is that it put mediocre graphic artists out of business and highlighted the difference between one that was mediocre and one that was good. I feel like this will happen again here with digital artists. The mediocre ones will be indistinguishable from AI, but the good ones will still stand out.

z3c0 4y ago
While I'm overall of the mind that nobody can really know what to expect, I think this may be a little more accurate of a reality than the aspiring Cassandras are trumpeting about. I'm willing to bet that most of the people who are going to utilize these technologies are people who never could afford a graphic designer to begin with. I'm sure some of the market will be lost to this tech, but a blackbox that shoots out passable images just isn't going to cut it for certain areas of the art market. I also wouldn't be shocked if a big portion of the market for this technology ends up being the artists themselves. I personally know a painter whose creative process has been overhauled by DALL-E. Brainstorming the next project is easier than ever, and unlike the DALL-E images inspiring them, the resulting paintings actually have the "human-touch" necessary to illicit the deeper emotional response that a good painting should bring about. Adding depth to a model doesn't necessarily add depth to the output. But like I said, I don't think anyone really knows what's going to happen. We'll see, I guess.
- ItsMonkk 4y ago
  When new technology is introduced people always talk about how existing workflows are cheaper. That's true, but the biggest impacts are always driven by things that were previously cost prohibitive that now become possible. YouTubers are a perfect example of that, it used to take entire television broadcast studios to do what they do, and now it can be done solo or with just a few people. And YouTubers are exactly the audience for this new tech - the smaller ones who want to put out branded merch can not currently do so at a level of quality. But you hire one artist to go out and make 5-100 pictures of your branding, tell them to take their time and make those few images to perfection, and that can now be molded to anything they want to create. Indie game devs must be thrilled. Aspiring indie directors looking to make green-screen backgrounds are thrilled. VRchat users looking for 3d models are thrilled.
  - bombcar 4y ago
    What’s interesting is the comparison- YouTubers can fit a niche that TV never could - specifically because they can produce content easier without people wandering off. Though much of YouTube can be summed up as what was the three cheapest TV shows to produce - standup comedy, talk shows, and howto. The amount of YouTube sitcoms is a much lower number.
- pbhjpbhj 4y ago
  There was an example on a Splendid Diffusion post earlier. Someone using SD for generating recipe pictures, https://news.ycombinator.com/item?id=32644800 https://news.ycombinator.com/item?id=32644800 where they almost certainly wouldn't hire an artist.
xbryanx 4y ago
Yes! We will still need new creatives to dream up something entirely new, which then can be fed into AI as new seeds. The cycle continues.
VoodooJuJu 4y ago
I think that's a good way to put it, but there's still a problem. The path of a digital artist is long and arduous. For a time on this path, the artist may be considered mediocre, or to put it better, they are an apprentice. Just as in other physical trades, an apprentice who is mediocre at their craft can still practice aspects of that craft well enough to be useful and earn some money. It is also through practice that the apprentice improves their skills. In this way, the apprentice is financially supported and even incentivized to improve at their trade, until one day they become truly good at it. So what things like DALL-E and Github Co-Pilot and your clip art package do is displace the apprentice. With no path of mediocrity for the apprentice to walk, to earn a stipend for training, how then can they receive the financial support necessary to train until they're a master? They would need to already be independently wealthy or receive financial assistance. In order to train more master artists and programmers, we would need to provide them with financial support while they train without us receiving anything useful in return.
- CuriouslyC 4y ago
  With tools like stable diffusion, mastery means something different than it used to. Now a master is understands style and composition, who knows how to use the tools effectively to produce stylish, well composed images, then has sufficient editing skills to clean up/paint over tool output in order to produce professional results.
  - salawat 4y ago
    All of said tools having the most godawful interfaces and documentation known to man if the JS ecosystem is any evidence of the direction things are going.
  - rdedev 4y ago
    But the problem presented in GP still exists right? Not everyone starts out with a good understanding of style and composition from the start. They need time to master those skills but also need a line of income to survive till then. If mediocre level work is all automated someone starting out might not get the time to reliably skill up
theptip 4y ago
I think putting some graphics artists out of business is one side of the coin. The other side of the coin is creating new jobs for people with good visual taste and imaginative ideas, who did not pass the skill barrier previously required to actually create good art. Presumably we need fewer of the new more efficient jobs to displace the work done by the old jobs. But that lowers the price of ”mediocre” graphics, and therefore increases demand; maybe we actually end up with more jobs at the more accessible level. These market dynamics are quite hard to predict but either way, bad time to be an entry level professional artist, great time to be just about anyone else.
- jameshart 4y ago
  This is a good take. Look at music production: historically the barriers to music creation were the dexterity and practice needed to master an instrument, or multiple instruments. But when you put tools in the hands of more people without the filter of needing the time, training and skill to coax the sounds in your head from the instrument you have… and suddenly you get stuff nobody was making before.
- Jevon23 4y ago
  >The other side of the coin is creating new jobs for people with good visual taste and imaginative ideas, who did not pass the skill barrier previously required to actually create good art. I mean. Do you think that there are people out there with good taste in software and imaginative ideas, who can’t pass the skill barrier currently required to write code? Why would art be any different?
  - theptip 4y ago
    I think art is a combination of physical skill and conceptual taste/ability, among other things. The skill part is non-trivial. Software seems like mostly thought-stuff to me, there’s no mechanical skill-based piece. But even so, when AI can generate a full app with the ease of iteration displayed in the OP, then sure, I think you’ll see some people with app ideas generating those apps themselves, instead of having to hire contract developers. Right now just completing stubs of functions doesn’t seem useful enough to allow someone that can’t code to make an app.
mrtksn 4y ago
What happens is, story tellers get empowered with each and every advancement that makes their process easier. Only the technical people with niche focus get screwed. Over the history of humanity the printing press, the photograph, the computers etc. destroyed some profession only to make something else flourish.
quadcore 4y ago
but the good ones will still stand out. Yes, for a day or a week. The AI wont stop scanning and learning.
- archagon 4y ago
  This AI is not intelligent. It will not “learn” in the same way that humans do.
  - Gigachad 4y ago
    What is this based on? So far these image generation AIs are learning to draw at a faster rate than I am.
    - nanonomad 4y ago
      [dead]
  - quadcore 4y ago
    There isnt a result it wont be able to mimic in a short period of time.
    - archagon 4y ago
      Highly unlikely, unless by “mimic” you mean “vaguely evoke”. There is no actor behind these models. Recombination is only a very limited form of intelligence.
      - quadcore 4y ago
        Its also supervised in the sense that for-profit companies work on it to grab as much value as they can.
    - Jevon23 4y ago
      Well it’s already been trained on millions (billions?) of images and it still has trouble with plenty of things. Scenes with more than one person in them, or multiple people and objects interacting in complex ways, are the most obvious cases. I’m sure the technology will keep improving but I think it’s possible to be overly optimistic as well.
k__ 4y ago
Sounds a bit harsh. It's like software development. Back in the day you could make good money with websites. Knew HTML? You got the job! But the goalposts are constantly moving. You had to add CSS and later JS to your skills to keep up. Today, you won't get paid for writing HTML anymore. Same for other industries. You have to move with the industry, gather new skills, etc.
april_22 4y ago
Yes and I think it will especially push the boundaries of art forward like we have never seen before. Also, it will be interesting to watch whether and how people will be assessing art created by AI. Will there be something like Connoisseurship for AI art?
ellis0n 4y ago
Back in 2012, I made a simple model that can generate multiple 512x512 images from an iPad 2 and 512MB of RAM in minutes. Here are the samples https://imgur.com/a/XMnMi https://imgur.com/a/XMnMi Then I found that the end of the era of handmade digital art is coming. Only transistors are limited and future digital artists will differ in memory size and teraflops like bitcoin miners.
Keyframe 4y ago
You're right about that. At the end of the day it is a tool in the craft domain of art. Artist take a long-ass time to perfect their craft end of the business, but ultimately what sets great artists from mediocre ones is not craft, it's their taste. Own (good or bad) taste pushes artist to not stop working on their image/painting until their own taste is satisfied. Same will happen no matter the tool. I am actually terribly excited by these tools to expedite sketching, not in a sense of speed as much as volume while hunting for those directions that satisfy that inner taste.

z3c0 4y ago

While I'm overall of the mind that nobody can really know what to expect, I think this may be a little more accurate of a reality than the aspiring Cassandras are trumpeting about. I'm willing to bet that most of the people who are going to utilize these technologies are people who never could afford a graphic designer to begin with. I'm sure some of the market will be lost to this tech, but a blackbox that shoots out passable images just isn't going to cut it for certain areas of the art market. I also wouldn't be shocked if a big portion of the market for this technology ends up being the artists themselves. I personally know a painter whose creative process has been overhauled by DALL-E. Brainstorming the next project is easier than ever, and unlike the DALL-E images inspiring them, the resulting paintings actually have the "human-touch" necessary to illicit the deeper emotional response that a good painting should bring about. Adding depth to a model doesn't necessarily add depth to the output. But like I said, I don't think anyone really knows what's going to happen. We'll see, I guess.

ItsMonkk 4y ago
When new technology is introduced people always talk about how existing workflows are cheaper. That's true, but the biggest impacts are always driven by things that were previously cost prohibitive that now become possible. YouTubers are a perfect example of that, it used to take entire television broadcast studios to do what they do, and now it can be done solo or with just a few people. And YouTubers are exactly the audience for this new tech - the smaller ones who want to put out branded merch can not currently do so at a level of quality. But you hire one artist to go out and make 5-100 pictures of your branding, tell them to take their time and make those few images to perfection, and that can now be molded to anything they want to create. Indie game devs must be thrilled. Aspiring indie directors looking to make green-screen backgrounds are thrilled. VRchat users looking for 3d models are thrilled.
- bombcar 4y ago
  What’s interesting is the comparison- YouTubers can fit a niche that TV never could - specifically because they can produce content easier without people wandering off. Though much of YouTube can be summed up as what was the three cheapest TV shows to produce - standup comedy, talk shows, and howto. The amount of YouTube sitcoms is a much lower number.
pbhjpbhj 4y ago
There was an example on a Splendid Diffusion post earlier. Someone using SD for generating recipe pictures, https://news.ycombinator.com/item?id=32644800 https://news.ycombinator.com/item?id=32644800 where they almost certainly wouldn't hire an artist.

xbryanx 4y ago

Yes! We will still need new creatives to dream up something entirely new, which then can be fed into AI as new seeds. The cycle continues.

VoodooJuJu 4y ago

I think that's a good way to put it, but there's still a problem. The path of a digital artist is long and arduous. For a time on this path, the artist may be considered mediocre, or to put it better, they are an apprentice. Just as in other physical trades, an apprentice who is mediocre at their craft can still practice aspects of that craft well enough to be useful and earn some money. It is also through practice that the apprentice improves their skills. In this way, the apprentice is financially supported and even incentivized to improve at their trade, until one day they become truly good at it. So what things like DALL-E and Github Co-Pilot and your clip art package do is displace the apprentice. With no path of mediocrity for the apprentice to walk, to earn a stipend for training, how then can they receive the financial support necessary to train until they're a master? They would need to already be independently wealthy or receive financial assistance. In order to train more master artists and programmers, we would need to provide them with financial support while they train without us receiving anything useful in return.

CuriouslyC 4y ago
With tools like stable diffusion, mastery means something different than it used to. Now a master is understands style and composition, who knows how to use the tools effectively to produce stylish, well composed images, then has sufficient editing skills to clean up/paint over tool output in order to produce professional results.
- salawat 4y ago
  All of said tools having the most godawful interfaces and documentation known to man if the JS ecosystem is any evidence of the direction things are going.
- rdedev 4y ago
  But the problem presented in GP still exists right? Not everyone starts out with a good understanding of style and composition from the start. They need time to master those skills but also need a line of income to survive till then. If mediocre level work is all automated someone starting out might not get the time to reliably skill up

theptip 4y ago

I think putting some graphics artists out of business is one side of the coin. The other side of the coin is creating new jobs for people with good visual taste and imaginative ideas, who did not pass the skill barrier previously required to actually create good art. Presumably we need fewer of the new more efficient jobs to displace the work done by the old jobs. But that lowers the price of ”mediocre” graphics, and therefore increases demand; maybe we actually end up with more jobs at the more accessible level. These market dynamics are quite hard to predict but either way, bad time to be an entry level professional artist, great time to be just about anyone else.

jameshart 4y ago
This is a good take. Look at music production: historically the barriers to music creation were the dexterity and practice needed to master an instrument, or multiple instruments. But when you put tools in the hands of more people without the filter of needing the time, training and skill to coax the sounds in your head from the instrument you have… and suddenly you get stuff nobody was making before.
Jevon23 4y ago
>The other side of the coin is creating new jobs for people with good visual taste and imaginative ideas, who did not pass the skill barrier previously required to actually create good art. I mean. Do you think that there are people out there with good taste in software and imaginative ideas, who can’t pass the skill barrier currently required to write code? Why would art be any different?
- theptip 4y ago
  I think art is a combination of physical skill and conceptual taste/ability, among other things. The skill part is non-trivial. Software seems like mostly thought-stuff to me, there’s no mechanical skill-based piece. But even so, when AI can generate a full app with the ease of iteration displayed in the OP, then sure, I think you’ll see some people with app ideas generating those apps themselves, instead of having to hire contract developers. Right now just completing stubs of functions doesn’t seem useful enough to allow someone that can’t code to make an app.

mrtksn 4y ago

What happens is, story tellers get empowered with each and every advancement that makes their process easier. Only the technical people with niche focus get screwed. Over the history of humanity the printing press, the photograph, the computers etc. destroyed some profession only to make something else flourish.

quadcore 4y ago

but the good ones will still stand out. Yes, for a day or a week. The AI wont stop scanning and learning.

archagon 4y ago
This AI is not intelligent. It will not “learn” in the same way that humans do.
- Gigachad 4y ago
  What is this based on? So far these image generation AIs are learning to draw at a faster rate than I am.
  - nanonomad 4y ago
    [dead]
- quadcore 4y ago
  There isnt a result it wont be able to mimic in a short period of time.
  - archagon 4y ago
    Highly unlikely, unless by “mimic” you mean “vaguely evoke”. There is no actor behind these models. Recombination is only a very limited form of intelligence.
    - quadcore 4y ago
      Its also supervised in the sense that for-profit companies work on it to grab as much value as they can.
  - Jevon23 4y ago
    Well it’s already been trained on millions (billions?) of images and it still has trouble with plenty of things. Scenes with more than one person in them, or multiple people and objects interacting in complex ways, are the most obvious cases. I’m sure the technology will keep improving but I think it’s possible to be overly optimistic as well.

k__ 4y ago

Sounds a bit harsh. It's like software development. Back in the day you could make good money with websites. Knew HTML? You got the job! But the goalposts are constantly moving. You had to add CSS and later JS to your skills to keep up. Today, you won't get paid for writing HTML anymore. Same for other industries. You have to move with the industry, gather new skills, etc.

april_22 4y ago

Yes and I think it will especially push the boundaries of art forward like we have never seen before. Also, it will be interesting to watch whether and how people will be assessing art created by AI. Will there be something like Connoisseurship for AI art?

ellis0n 4y ago

Back in 2012, I made a simple model that can generate multiple 512x512 images from an iPad 2 and 512MB of RAM in minutes. Here are the samples https://imgur.com/a/XMnMi https://imgur.com/a/XMnMi Then I found that the end of the era of handmade digital art is coming. Only transistors are limited and future digital artists will differ in memory size and teraflops like bitcoin miners.

Keyframe 4y ago

You're right about that. At the end of the day it is a tool in the craft domain of art. Artist take a long-ass time to perfect their craft end of the business, but ultimately what sets great artists from mediocre ones is not craft, it's their taste. Own (good or bad) taste pushes artist to not stop working on their image/painting until their own taste is satisfied. Same will happen no matter the tool. I am actually terribly excited by these tools to expedite sketching, not in a sense of speed as much as volume while hunting for those directions that satisfy that inner taste.