AI Generated Art - Impact on Artists

by Simon 22 Replies latest social current

• 

  Simon

  Unless you've been hibernating under a rock for the last couple of years, you're probably aware that AI is the new hotness.

  ChatGPT gets a lot of attention but equally amazing is how capable AI is at generating art. If you'd been ask to name anything that would be the last holdout for humans, creative arts such as writing, painting, and music would probably be some of the things at the top of the list.

  Yet those seem to be the things that AI can do best.

  It's fascinating to use the tools, but one thing you realize is that they are just combining patterns and shapes, there isn't yet any real "intelligence", artificial or otherwise.

  As an example, you can ask it to create a painting in the style or a particular artist, with a whole raft of things you want to include. It will do a great job on first glance, but can often give people more arms or fingers than is normal, and lamp-posts can grow to different heights, like trees. All the pieces are there, and totally in the style you asked for, but it doesn't really know how that it's a drawing of something, and what that thing is - it just knows that those works often go together with certain shapes and patterns and how those shapes and patterns are combined.

  But with some work on the prompts (and negative prompts to prevent certain things) you can actually get really good results. No, you can get GREAT results. Most people wouldn't know it wasn't a real photo or painting.

  So what about artists? Automation and technology (beyond better brushes, paints, and canvases) is finally going to hit the world of art and it will be devastating to many. How can one human compete with masses of images that can be generated in seconds that likely have your name associate with them from being done "in the style of" you?

  Of course this has happened to other industries. Work changed, people work at a higher level and use the new tools. When power tools came out, carpenters didn't disappear - they become more productive and those with the best skills at using the new tools were still in demand, but there were fewer employed overall (per population, the increases probably mean there are more than ever).

  It's going to be interesting to see what happens to artists over the next few years. If I were going to create a book for instance, I don't think I'd be paying anyone for illustrations or cover art because I can generate it myself. So it's democratized creativity but there are lots of ethical questions - it's been trained on the art produced by real people, so is derivative ... but then all art is anyway, it's just being evolved faster than ever before thanks to technology

  One group I wouldn't mind seeing devastated are the art-licensing companies who really are leaches or parasites that for some reason artists tolerate and keep in business. It's not unusual for them to take 50% or more of any payment for someone to use some artwork and often they do nothing to really promote the artist or their work at all.

  If you want some examples of what it's capable of, try this:

  https://civitai.com/images/160240?modelVersionId=13259&prioritizedUserIds=3659&period=AllTime&sort=Most+Reactions&limit=20

  NOTE: that site can be very NSFW, especially if you sign-up so beware. The other major thing that AI will impact is porn, and no doubt there will be a ton of legal issues relating to generated and manipulated images.

  If you want to try AI art out, the easiest way it to join a discord that will generate images in response to your prompts. There are also some web based generators to use too.

  https://discord.gg/bluewillow

  https://dreamlike.art/create

  If you have a suitable GPU and want to install something locally, the easiest might be https://stable-diffusion-ui.github.io/

  There are some others that require more computer smarts to install and get running but provide greater control - more difficult to get started with but ultimately way more powerful.

  Anyone else tried it yet?
• 

  TonusOH

  Some of the early efforts I've seen have been mixed, but I think AI art will continue to separate those who make the effort from those who don't. I've seen people spamming art sites with AI generated works, apparently as a cheap and easy way to make a quick buck, and I've seen people who use AI to enhance their work and add a depth to it that they couldn't previously. Sure, it may seem as if they're cheating a bit, but they're taking advantage of a new tool to produce better artwork. Because they make the effort, their work looks very nice. The spammed artwork is full of glitches and just plain bad design.

  As someone who dabbles in both art and writing, I'm more ambivalent about it than you might think, but that is because the market has changed. Already, people are able to steal both artwork and writing to make easy money at the expense of creators. I think that the future opportunities for creators are to build a small but loyal following that wants your work, not facsimiles or poorly-designed copies. Art/writing mills will make some money in the short term, but those who build a real following will have sustained income and sustained engagement for as long as they want.

  And yes, AI might shake up some of the industries or communities where there was easy money to be made by stealing from others. When you can generate art without technically stealing it, and when you can get 'good enough' results with a few queries... this could lead to hard times for art/writing thieves. I'm not going to shed any tears over it.
• 

  Simon

  Yeah, I can see the value of genuine original art going up - people want things that are authentic. And in a way, I see it like music - should people really be getting ultra rich and living off the work they did one afternoon? Or should they be paid for when they actually work and do a concert / performance? I know it's a long-tail, and the few who make it big suck up all the money, but people still want to see good local musicians even though we live in a world with Spotify. Art will be the same, people will pay for the real thing, and real artists can produce that, copiers can't.

  And yes, you can often spot the AI art because people are not selective enough to filter out the ones with weird anomalies that are the giveaways. If you are selective and work at the prompts the results can be outstanding, and sometimes the "happy accidents" (Bob Ross, right?) are what make things better.

  Ultimately, it is just another tool to use and future artists may be the ones who master the new tools better.

  Software development also has the same challenges - many feel threatened with ChatGPT (that can write code) but unlike art the viewer doesn't really know if it makes sense or is full of nonsense anomalies (sometimes it is). You still need expertise, and it isn't writing the entire app or integration or talking to the business to extract what they really need (which rarely matches what they ask for). The people who copy-and-paste code off StackOverflow will still be the same clueless people creating crap when they use ChatGPT if they use it as a shortcut and not a tool to learn with.
• 

  joey jojo

  I have played with around with it. My first thought was that for anyone that creates digital art, they have just unlocked easy mode for their creations. The only limiting factor is their imagination and how well they learn to phrase requests to get what they want.

  For anyone that creates art, it's an amazing way to get what's in your head into a visual form if you want a reference pic for ideas you have. That makes it a massive time saver.

  It still relies on human imagination to produce the images, so to me that makes it a valid art form. There is talk about royalties being paid to artists because their style may be used by the program in generating the piece but maybe thats just the leeches talking.

  Art has always been about absorbing inspiration from other artists and producing something uniqie.

  'Steal like an artist.'
• 

  Anony Mous

  Here’s the thing, it uses existing inputs, it can only give existing outputs, it won’t generate anything novel in the sense that you won’t get a new style like Cubism in the early 20th century (Picasso).

  Here are the things that will be replaced: generic artists, the people in advertising that look to influence the average person with safe, generic boilerplate ‘template’ art. It can generate a website, copy and images with a simple prompt, where current website systems (eg. wordpress) have oodles of templates that no sane person could ever pour over, you’ll basically ‘search’ for your style and it will generate that style. Big companies like Coca Cola will never use it, but instead of having a 1000-person website team, they’ll now have to seek out select true artists and pay them a lot more.

  I’ve used it in the field of computer and neurosciences, where I currently work, it’s great IF you know what you’re looking for, it improves your productivity but it can’t do anything new. So your code is now better commented because you have to form better sentences to have it generate semi-decent boilerplate. But it won’t implement your logic, it doesn’t understand starting points and end goals.

  So what will happen: people at the low ends of each industry, the ‘gophers’, they’ll have to quickly adapt and either get better or get out. You won’t need to fill offices with useless people like HR, PR. In most fields however, people won’t be allowed to stay ‘junior’ for very long as the junior roles will be completely replaceable. People at the top will get better and more productive as they won’t have to think about and work out small details anymore.

  It’s going to have a huge impact, people that haven’t taken up the ‘knowledge economy’ in the last few decades, the accountants that stubbornly stick to Excel and QuickBooks instead of specializing in R and SQL, the 50 year old entry level IT person making $80k+ because he creates his own work, the HR and legal at your average company that just search through hundreds of pages of rules, those are all gone because I can ask a trained model the same questions and get the same answers.
• 

  slimboyfat

  AI art is derivative. But human art is derivative too. No artist starts from scratch. They all copy and adapt and evolve.

  The malformed hands are really ugly, but they are already working on the problem and I hear it is fixed already on some models.

  That’s what we’ve got to remember, these are very early programme. The deficiencies in the current images will be vastly improved. The ability already, and the potential for further improvement is mind boggling.

  I suppose I sympathise with artists to some extent. But you can’t stop AI learning from existing art just as you can’t stop a human artist from learning from existing art. There is no practical or reasonable way you could constrain that.

  Personally I think it’s fantastic that we can all produce images we want with increasing ease.

  But I do worry what the whole thing means for society. I think we are looking at massive restructuring of society, work, the money system - everything. And I don’t think we have the political or public institutions ready to cope with the challenge.

  Here is one of my AI generated images using Midjourney. It’s a scientist in ship yard in the style of cyberpunk.

  
  
• 

  Simon

  The malformed hands are really ugly, but they are already working on the problem and I hear it is fixed already on some models.

  Hands are interesting. Apparently, lots of human artists struggle with them, so don't include them in paintings. Factor in that a lot of pictures are portraits, and maybe don't include them, the number of examples in the training set is going to be a lot less than for faces which it's often amazing at, certainly by comparison.

  Remember, it's not really AI, there is no I ... it doesn't know what fingers really are or how they work, how a hand can change shape whether it's fingers curled round an umbrella, the handle of a mug, pointing or waving ... which is why it often get's the general idea of where hands go but they don't always make much sense.
  
• 

  TonusOH

  I agree about hands. I think AI will eventually be fine-tuned enough to capture even something as complex as human hands and similar structures. I think it's less about the amount of data fed to it, and more about smarter algorithms.
• 

  Terry

  In one day I generated the following images:
  
  

  Monday Bing A.I. using my text prompts
  
  
  
  
  
  
  
  
  
  

  
  

  

  

  
  

  
  
  
  
  

  
  

  

  \\
  
  
  
  
  
  
  
  
  

  
  
• 

  enoughisenough

  well it seems impressive! how do those who have done this feel about what "you" have created? Do you feel like it is your artwork?