Why don't indie devs use AI-generated images as art?

[Davaris]

Stable Diffusion output is objectively worse. Anyone with eyes can see it.

What I meant to say there is, I suspect they have an agreement with Discord. Since you cannot use MidJourney unless you join Discord.

"Stable Diffusion also ran via Discord."

Stable Diffusion aren't forcing you to use Discord.

MidJourney is forcing you to use Discord.

 

[Non-Edgy Gamer]

Stable Diffusion output is objectively worse. Anyone with eyes can see it.

You're trying the same inputs on two different models. They're not going to work the same.

Also, one is older and has probably been finetuned and/or retrained a half-dozen times, and the other is still awaiting the release of its first version update.

Both of them basically use the same technology and are subject to similar technical limitations. The difference is that Stable Diffusion is free and doesn't force you to have Discord trannies reviewing every image you generate.

What I meant to say there is, I suspect they have an agreement with Discord. Since you cannot use MidJourney unless you join Discord.

There is no agreement needed. I could setup a Discord bot too. It's not actually generated on Discord. Just posted there.

The same would apply to Stable Diffusion if they also wanted to go the community-assisted moderation model and were also too cheap to make a website.

But there's really no point to them doing that, since they use a censor on their web platform's images and already let people generate uncensored nude images in the open-sourced model.

 

[Davaris]

You are slapping your head there. I said they are forcing people to join Discord to use their product, which indicates to me they have a financial agreement with Discord.

Then you said another company is offering Discord as an option, which means they are are not forcing anyone to use it.

So what?

Bad Sector is right about you:

"Dude, stop. These responses show that you clearly did not understand what i wrote at a fundamental level and i am not interested in explaining to you your gross misunderstanding. Re-read what i wrote from the beginning without trying to "prove me wrong" or whatever mode your brain is in right now because you are reading things that do not exist."

Are you an AI?

 

[Non-Edgy Gamer]

You are slapping your head there. I said they are forcing people to join Discord to use their product, which indicates to me they have a financial agreement with Discord.

And I said it doesn't. They are using Discord as a platform the same as anyone can. What they're doing isn't unique at all and literally anyone can do it.

Then you said another company is offering Discord as an option, which means they are are not forcing anyone to use it.

Ok, you're misunderstanding the cause and effect here:

MidJourney doesn't want people to use Discord. They want an easy way to manage thousands of images and have community content moderation. Because hiring people to police all these images would bankrupt them.

Bad Sector is right about you:

Bad sector wasn't half as retarded as you seem to be. I honestly skipped over his sentence about the license "not being compatible with open source" (even though it is and plenty of Debian-compatible licenses restrict use for things like "hate speech") because he spammed "muh license" so many times, but since he was being kind of a dick about it, I didn't apologize.

You are just flat out drawing dumb conclusions based on nothing.

"This company has a Facebook page! They must have an agreement! Durrrrr!"

Whatever, nigger. Keep thinking that.

Are you an AI?

If I am, I'm probably one that's smarter than you.

 

[Blaine]

Upscaling will continue to suck for the rest of eternity because the visual information (fewer pixels = less information) simply isn't there, and guesses will never be good enough, because they can only ever be based on the existing, surrounding information.

Spoiler

Are you trying to prove my point here, or pull a fast one? There are entire structural elements missing, or added, in the "upscaled" version. It's a complete re-fabrication, not true (emphasis on "not true") upscaling, like an artist painting a larger copy of the Mona Lisa using the original as a reference, except less faithful in this case as entire omissions and additions have been made.

https://postimg.cc/k2chVgJ2

What you've posted is technically and correctly called copying (in the art and graphics world), not upscaling.

Now try upscaling a postage stamp to letter size while seemingly retaining all details intact at a proportional DPI/PPI.

 

[Davaris]

You are slapping your head there. I said they are forcing people to join Discord to use their product, which indicates to me they have a financial agreement with Discord.

And I said it doesn't. They are using Discord as a platform the same as anyone can. What they're doing isn't unique at all and literally anyone can do it.

Then you said another company is offering Discord as an option, which means they are are not forcing anyone to use it.

Ok, you're misunderstanding the cause and effect here:

MidJourney doesn't want people to use Discord. They want an easy way to manage thousands of images and have community content moderation. Because hiring people to police all these images would bankrupt them.

Bad Sector is right about you:

Bad sector wasn't half as retarded as you seem to be. I honestly skipped over his sentence about the license "not being compatible with open source" (even though it is and plenty of Debian-compatible licenses restrict use for things like "hate speech") because he spammed "muh license" so many times, but since he was being kind of a dick about it, I didn't apologize.

You are just flat out drawing dumb conclusions based on nothing.

"This company has a Facebook page! They must have an agreement! Durrrrr!"

Are you an AI?

If I am, I'm probably one that's smarter than you.

I bet you can't pass a Google captcha.

I'm putting you on ignore now, so we don't have to do this any longer.

 

[Non-Edgy Gamer]

I'm putting you on ignore now, so we don't have to do this any longer.

And another snowflake melts away.

 

[Dexter]

Are you trying to prove my point here, or pull a fast one? There are entire structural elements missing, or added, in the "upscaled" version. It's a complete re-fabrication, not true (emphasis on "not true") upscaling, like an artist painting a larger copy of the Mona Lisa using the original as a reference, except less faithful in this case as entire omissions and additions have been made.

https://postimg.cc/k2chVgJ2

Yes, that's the point. The first image was already an AI fabrication that was hallucinated by from noise. Further steps do the same but add more and more detail trying to make it look like what I want while staying faithful to the initial composition, and I told it to do that via Denoise. That's what an intelligent AI Upscaling Algorithm does. You can also tell it to Copy the image as accurately as possible without adding any detail (e.g. no Denoising), in which case the final result will obviously kinda suck at that size since the detail in the Original isn't there:

Spoiler

 

[Blaine]

Certainly the end result is aesthetically superior to algorithmic upscaling.

You know, this suggests an interesting experiment: If you begin with a postage stamp-sized image (size = bytes in this case, as it's the easiest context-agnostic measurement of information density) and instruct the AI to "copyscale" it progressively upward through, say, a dozen iterations until you've reached an image that's more on the order of tabloid-sized, how greatly would the end result differ from the original postage stamp?

Considering how many additions the AI made at each step in your sequence above, I'm guessing it'd be quite different.

 

[Dexter]

You know, this suggests an interesting experiment: If you begin with a postage stamp-sized image (size = bytes in this case, as it's the easiest context-agnostic measurement of information density) and instruct the AI to "copyscale" it progressively upward through, say, a dozen iterations until you've reached an image that's more on the order of tabloid-sized, how greatly would the end result differ from the original postage stamp?

Considering how many additions the AI made at each step in your sequence above, I'm guessing it'd be quite different.

I would guess it would highly depend on what you tell it to "see" in the noise and how autistically accurate you could describe the stamp while trying to skirt the edge between Original and new details. I told it I want a Cyberpunk Cityscape based on a specific image and painter and that's what it hallucinated. If you tell it that's not actually a city at all but a Fantasy Woodland it'll produce entirely different results:

Spoiler

 

[Darkozric]

Davaris You should be wishing that an AI coding will never exist because non coders like me will destroy you for good, arrogant fuck.

 

[Davaris]

A couple of days ago I saw an AI coding in Python, from text prompts people gave it. I saw one person ask it to make a program to 'Simulate the Universe', and it wrote a program with comments and physics formulas.

So here is my updated plan for making money with AI art:

Tell the AI that programs Python, you want a Stable Diffusion website, with advertising banners linked to your GPU farm.

Next create lots of accounts on forums, and social media (you will need a human to pass the captchas).

Then tell your Python programming AI, you want a program that links these accounts to the latest AI chatbots. These chatbots will argue/answer questions from humans, and other chatbots, about AI generated art for you.

When that is up and running, which will take about 1 second for the AI to code, you can use the money it generates from add clicks, to relax in your favorite holiday destination.

Twitter is estimated to be 80% bots by the CIA, so a lot of people have been making money doing this kind of work for some time.

A word of warning, if you can't code, you won't know if the AI is making your AI art website, or if it is emptying out your bank account. Even then, if you can code, it may do this and you might miss it.

 

[Blaine]

Davaris You should be wishing that an AI coding will never exist because non coders like me will destroy you for good, arrogant fuck.

A coder could just as easily use AI coding as could a non-coder—better, actually, since the coder would not only be able to use the AI, but would also know how to code.

It's not that you aren't a coder that's holding you back from destroying Davaris, but rather that you're not very bright.

 

[Darkozric]

It's not that you aren't a coder that's holding you back from destroying Davaris, but rather that you're not very bright.

You're not wrong, I'm not bright enough to code, my limited intelligence was mostly busy with conservatories and jamming with prestigious musicians. Each to their own domain.

But there's one thing that I can guarantee you with confidence, I will never be an arrogant cunt when AI is starting to compose equally or even better music than me.

 

[Blaine]

You're not wrong, I'm not bright enough to code, my limited intelligence was mostly busy with conservatories and jamming with prestigious musicians. Each to their own domain.

There are many different domains of intellect, among which musical intellect is as top-echelon as any other form, and arguably higher than many. Just because you don't make a good engineer or coder doesn't necessarily mean that your intellect is limited.

Your intellectual failure to which I was referring was one of very basic logic, something that creative types often struggle with (and perhaps necessarily, because rigid logic rarely goes hand-in-hand with creativity). It simply stands to reason that coders are better equipped to harness, monitor, and customize coding AIs than non-coders, in much the same way that musicians would be better equipped to monitor AI-produced sheet music than would the musically illiterate (not a direct parallel, but a good one nonetheless).

That having been said, AIs completely lack true creativity, as far as we can measure. They mimic only. Therefore, creative minds have something to contribute to machine learning and AI systems.

 

[Darkozric]

It simply stands to reason that coders are better equipped to harness, monitor, and customize coding AIs than non-coders, in much the same way that musicians would be better equipped to monitor AI-produced sheet music than would the musically illiterate

It's a good example, fair point. But since coders/artists are better equipped to harness coding/drawing AIs, then what's the cause of the amount of butthurt and panic by many? I'm just curious.(I'm not referring to Davaris specifically, I was just messing with him.)

 

[Blaine]

It simply stands to reason that coders are better equipped to harness, monitor, and customize coding AIs than non-coders, in much the same way that musicians would be better equipped to monitor AI-produced sheet music than would the musically illiterate

It's a good example, fair point. But since coders/artists are better equipped to harness coding/drawing AIs, then what's the cause of the amount of butthurt and panic by many? I'm just curious.(I'm not referring to Davaris specifically, I was just messing with him.)

I have no idea. I just "discovered" this thread yesterday, and frankly I was bewildered at how contentious and aggressive everyone has become over a discussion about AI. I would've thought it'd be a neutral or safe topic, but there are several people here fighting like cats and dogs.

Now granted, I joined in a bit, but uh... you were talking about destroying people. Figurative or no, that's a bit aggressive.

 

[Machocruz]

It's a good example, fair point. But since coders/artists are better equipped to harness coding/drawing AIs, then what's the cause of the amount of butthurt and panic by many? I'm just curious.(I'm not referring to Davaris specifically, I was just messing with him.)

Speaking for artists, the joy of handcraft coupled with the commercial dominance of AI generated images that is on the horizon. Putting in words to create images and creating images through physical manipulation of a tool and medium are not remotely alike. It's the tactile act of drawing or painting or sculpting that many love, not necessarily the result. But they will, if they wish to be employed by someone else, be forced to use the new thing, and they will loathe it. And like it has always been, most are not skilled, original, or savvy enough to succeed outside of the established order. People who are more about the final result and coders who just want to make images shouldn't have a problem.

Generally the people who are panicking about the future of (paid) art making are very ignorant about the art world, and art in general. They lack knowledge, imagination, ability, and purpose.

 

[Blaine]

Machocruz There is value in originality, however. These AI tools produce nothing wholly original, alhough it can be argued that no one truly does: after all, the first cave painters painted the animals that they could see, and all artists worth their salt begin painting from life and references. No one is God, and no one can create something out of literal nothing; but what geniuses do create from the accumulated imagery, emotions, sounds, life experiences, etc. that are all mixed up and stewing around in their brains can be so close to original that we needn't try to nitpick the difference. The ability to reconfigure and combine all of those memories in certain ways and to then recreate (in a medium) and present them in a certain way is actually incredible.

The extremely elaborate "distorted and embellished collages" that AI creates can achieve aesthetically-pleasing results much more easily than human beings, I'd say, and although they only mimic and create nothing original per se, these "collages" can often present an appearance to us that seems original, like nothing we've ever seen before. At the moment I think most people can detect the falseness behind it all (after all, the AIs basically scan a ton of images on Google based on various criteria, cut out various portions, then paste/morph them together and distort them based on certain criteria), but that may eventually change, and nevertheless there's still a huge novelty factor.

In the same way that both an eagle and a man can arrive at the same mountaintop via very different means, including the way that they evolved as living organisms as well as the method of transit, it may be that AI will get there someday, and maybe even beyond. However, man also made it to space, whereas the eagle never will (except for the Eagle lander).

 

[Non-Edgy Gamer]

@Machocruz There is value in originality, however. These AI tools produce nothing wholly original

The problem is that we live in a materialistic world where companies value production over creativity.

Most of the films and games released today are rehashed ideas of rehashed ideas. The spark of human creativity has long since become an afterthought. The money today is in production.

Does a white man generally code better than an Indian? Yes. Does that mean companies will hire him instead of 50 Indians? No.

The extremely elaborate "distorted and embellished collages" that AI creates can achieve aesthetically-pleasing results much more easily than human beings, I'd say, and although they only mimic and create nothing original per se, these "collages" can often present an appearance to us that seems original, like nothing we've ever seen before.

Spoiler

Sounds like a description of almost every movie made for the last 20 years.

AI is a disruptive technology to the art field, even if it's only AI-assisted art that's truly viable, because of the way business works in this age. The companies who will fire 10,000 employees and outsource their jobs to a call center in the Philippines with people who can barely speak English care very little for ideas of God or human creativity. What they're looking at is ways to make more money.

Westerns are about to be introduced to the idea of driverless robots delivering their food. Something that sounds far-fetched, but that Russia has had for years now. No one is talking about the lack of human connection in that. They'd just be glad they don't have to tip their driver.

I agree with you on a fundamental level. But this is the mistake people make when discussing if AI is viable or what a "true" AI is: what matters is whether or not it's good enough to cost someone their job, not whether or not it's as good as a human.

 

[Davaris]

It simply stands to reason that coders are better equipped to harness, monitor, and customize coding AIs than non-coders, in much the same way that musicians would be better equipped to monitor AI-produced sheet music than would the musically illiterate

It's a good example, fair point. But since coders/artists are better equipped to harness coding/drawing AIs, then what's the cause of the amount of butthurt and panic by many? I'm just curious.(I'm not referring to Davaris specifically, I was just messing with him.)

I have no idea. I just "discovered" this thread yesterday, and frankly I was bewildered at how contentious and aggressive everyone has become over a discussion about AI. I would've thought it'd be a neutral or safe topic, but there are several people here fighting like cats and dogs.

Now granted, I joined in a bit, but uh... you were talking about destroying people. Figurative or no, that's a bit aggressive.

I think he got big mad because I said text adventures don't sell, which is objectively true, because they haven't been commercially viable since the 1980s. Either that, or being edgy is his LARP.

They are saying this is good for games, when all it can do is make pictures of landscapes and portraits. If you want to make text adventures this is fine, but they are a tiny percentage of the kinds of games people want to make. When robots start running around taking pictures of everything, this could become useful for prototyping other kinds of games.

 

[Non-Edgy Gamer]

I think he got big mad because I said text adventures don't sell, which is objectively true, because they haven't been commercially viable since the 1980s.

How is something a text adventure if it has images? This guy.

I thought he was talking about visual novels, which made the Fate series into a multi-billion dollar franchise, but now he's talking about the 80s.

 

[Darkozric]

Jesus wept...I posted just one troll-ish shit and he put me on ignore. A 2005 account and he acts like he's on RPGwatch.

I had no intention to spam him with BS or to waste his "precious" time. It seems that some ppl are softer than butter.

 

[Davaris]

It's a good example, fair point. But since coders/artists are better equipped to harness coding/drawing AIs, then what's the cause of the amount of butthurt and panic by many? I'm just curious.(I'm not referring to Davaris specifically, I was just messing with him.)

Speaking for artists, the joy of handcraft coupled with the commercial dominance of AI generated images that is on the horizon. Putting in words to create images and creating images through physical manipulation of a tool and medium are not remotely alike. It's the tactile act of drawing or painting or sculpting that many love, not necessarily the result. But they will, if they wish to be employed by someone else, be forced to use the new thing, and they will loathe it. And like it has always been, most are not skilled, original, or savvy enough to succeed outside of the established order. People who are more about the final result and coders who just want to make images shouldn't have a problem.

Generally the people who are panicking about the future of (paid) art making are very ignorant about the art world, and art in general. They lack knowledge, imagination, ability, and purpose.

I see Darkozric says, he was just messing with me. Whateva.

I haven't seen any AI art that makes me want to fanboy it. I look at it, and forget it. Good art has messages put in it by an intelligent artist with life experience.

At the moment people in general, not just artists, fear AI, because they think it will take their jerbs, and there will be no safety net for them. They are right. It will take their jerbs. They better make plans for that.

 

[Non-Edgy Gamer]

A quick experiment with everyone's favorite roastie:

Original:

Upscaled (Left), Img2Img edits with Stylistic Changes (Right)

The most annoying thing the AI does is change colors without being asked to. Especially things like eye color. Lowering the diffusion helps, but it still happens. That and mistakes like the scar on the left right (left) side of her face, but that's more to do with the upscaling and how she was drawn.