AI Art: It is criminal to not post your prompt

The Something Awful Forums > Main > General Bullshit > AI Art: It is criminal to not post your prompt

«‹›435 »

Boba Pearl: Dec 27, 2019; by Athanatos

and what about using img2img to turn your MS paint into a big ole painting?

# ? Sep 19, 2022 22:10

Adbot: ADBOT LOVES YOU

# ? May 30, 2024 13:37

Objective Action: Jun 10, 2007

I mean for me it's just that my hands are busted pieces of poo poo and this lets me get ideas out visually so take your 'no true Scotsman' stuff elsewhere please Funky.

# ? Sep 19, 2022 22:17

Boba Pearl: Dec 27, 2019; by Athanatos

I get this discourse literally all the time on twitter, because I'm very open about how to use AI, and teaching my process and helping keep it set up, and am even hording models for archivization purposes. At the end of the day, no matter how bereft of soul you think it is. It's here. It exists. It's going to be used, so either use the tools and work them into your flow or not. Be grateful it's in the hands of everyone instead of a few elite corporations that would have charged through the nose for access. No-one talks about AI assisted writing by NEO-20B or Fairseq or GPT-3 being soulless, and I assist my writing with AI literally all the time.

There's so much that goes into these things, and honestly, this is kind of why everyone loving hate artists every new piece of technology or help or method is immediately denounced as soulless and terrible and not as great as the new masters. Whether it's 3D modelling, references, paint overs, dating all the way back to tracing shadows with wire grids with Rembrandt, a technique even older then him. People wouldn't want to replace us if we weren't gate keeping poo poo heads literally all the time, because we had the free time and willpower to draw circles for 8 hours a day for loving years.

It's exhausting, it's unwanted, it's also wrong, because we could take 4 AI paintings, and 4 trad paintings, put them all next to each other, and no-one could pick the AI ones out with any statistical accuracy. That's why they won an award. That's also why people are trying to ban them everywhere.

If the AI art wasn't real or was so obviously soulless, then Newgrounds, Furaffinity, and other places wouldn't feel the need to ban them. The fact is it threatens art spaces in the same way the camera threatened illustrative spaces, and the auto-loom threatened pattern weaving. People like art too much to turn their nose up at it, and I'm tired of luddites making GBS threads on a new technology and stymieing people's creativity.

People should be able to discuss they're methods and techniques without people with a hair up their rear end forcing them to debate whether it's real art or not, or if it has soul or not.

# ? Sep 19, 2022 22:36

Hadlock: Nov 9, 2004

Boba Pearl posted:

. No-one talks about AI assisted writing by NEO-20B or Fairseq or GPT-3 being soulless, and I assist my writing with AI literally all the time.

Hi can you walk me through how this works

Like say I have a translation of an out of copyright short story in polish, feed it through Google translate to get very rigid translation

Could I feed it, sentence by sentence or paragraph by paragraph into gpt-3 and get more good prose out the other end

# ? Sep 19, 2022 22:46

Moongrave: Jun 19, 2004; Finally Living Rent Free

i could describe to someone "oh yeah it's a sexy vampire lady emerging from a pool of water with roses on it, walking right towards the camera like an old bond film or something"

and people would know what I mean

now i can tell a machine that learned exactly like a human does, but has infinite skill, memory and patience, the same thing and it'll go "oh like this?"

# ? Sep 19, 2022 22:54

Boba Pearl: Dec 27, 2019; by Athanatos

Hadlock posted:

Hi can you walk me through how this works

Like say I have a translation of an out of copyright short story in polish, feed it through Google translate to get very rigid translation

Could I feed it, sentence by sentence or paragraph by paragraph into gpt-3 and get more good prose out the other end

Possibly, yeah Take something you have, and feed it through here, but just like AI art, it needs a lot of refining and reworking You'll get good prose though https://opt.alpa.ai

It'd be easier with Sudo, which has a summarize and convert to prose option https://www.sudowrite.com/app

# ? Sep 19, 2022 22:55

Boba Pearl: Dec 27, 2019; by Athanatos

OPT Alpa will try to continue from the last sentence you had, where as Sudo will rewrite what you had to the best of it's ability.

# ? Sep 19, 2022 22:59

Tunicate: May 15, 2012

for GMing, I toss a couple bucks at GPT now and then, and us it to get suggestions for not-metagamed suggestions of what the enemies will do next, and stuff of that nature. Even if it's not great at coming up with things overall, getting some sort of external input and cherrypicking the good parts is a nice way to get out of ruts.

# ? Sep 19, 2022 23:10

Boba Pearl: Dec 27, 2019; by Athanatos

I'd really recommend NovelAI or KoboldAI as well for story writing and GMing, you can fill the context with your characters and stuff, and it'll hit the plot points pretty well. I find Euterpe better then Krake with modules, but Krake works with character sheets.

# ? Sep 19, 2022 23:12

Tunicate: May 15, 2012

Oh, yeah, also good for generating weird buts of loot

# ? Sep 19, 2022 23:32

SCheeseman: Apr 23, 2003

Boba Pearl posted:

At the end of the day, no matter how bereft of soul you think it is. It's here. It exists. It's going to be used, so either use the tools and work them into your flow or not. Be grateful it's in the hands of everyone instead of a few elite corporations that would have charged through the nose for access.

I recently got dogpiled in the Games Industry thread by suggesting that maybe AI art does have legitimate use, conceding that it will likely hurt certain parts of the art jobs market and the questionable ethics behind the training process. Granted I kinda derailed it and rightly ate a probe for it, but it highlighted how that kind of fear quickly turns into stupidity. The prevailing opinion was that enforcing copyright on datasets would be the solution, as that would render existing datasets infringing and require datasets to only contain works that the trainer owns or license.

Yeah, cool idea bro. Allowing only the biggest, monopolistic entertainment conglomerates who sit on mountains of IP the ability to train and create and license these tools, removing the option for anything free or open to be in any way competitive will fix everything.

# ? Sep 19, 2022 23:47

Swedish Thaumocracy: Jul 11, 2006; Strength of >800 Men
Honor of 0; Grimey Drawer

I don't get how people say ai art does not have soul. The idea behind the art comes from the same place - me, my mind, my imagination, my experiences, outer-space, whatever - and the result is just, a lot better and more importantly a lot faster than anything I could have done without it.

Not to mention the styles which the ai can emulate on a whim. Could I, through years of effort, become good enough to emulate one of these? Yeah, probably. All of them? Whilst balancing work and family life? Highly doubtful. With this new tool, i can make my ideas real, no matter the other resources (time, skill, money, connections, experience etc) are available to me. All i need is a computer with a half-way decent graphics card and a keyboard.

# ? Sep 19, 2022 23:51

Bottom Liner: Feb 15, 2006; a specific vein of lasagna

I'm really curious if it ever gets better at being able to take queues for composition without the need for in-painting and such. I've created plenty of very good usable images already, but it's always a matter of starting broad and narrowing down as opposed to giving it a precise prompt and getting exactly what I'm looking for (or even close, without a lot of re-tries).

The copyright/plagiarism argument doesn't hold any sand because the output pulls from a million things at once and turns it into something very different. If you trace a copyright image by hand but the end result looks nothing like it, no one would ever try to fight that in court. Imagine a collage artist creating a new face by pulling pixels from a million different portraits in magazines. That's closer to what these things do. I know 3 other separate creatives across a variety of fields (graphic design, film, board game design) using it professionally already, really feels like people need to just accept that it's another tool in the box at this point.

# ? Sep 20, 2022 00:02

mobby_6kl: Aug 9, 2009; by Fluffdaddy

Boba Pearl posted:

I get this discourse literally all the time on twitter, because I'm very open about how to use AI, and teaching my process and helping keep it set up, and am even hording models for archivization purposes. At the end of the day, no matter how bereft of soul you think it is. It's here. It exists. It's going to be used, so either use the tools and work them into your flow or not. Be grateful it's in the hands of everyone instead of a few elite corporations that would have charged through the nose for access. No-one talks about AI assisted writing by NEO-20B or Fairseq or GPT-3 being soulless, and I assist my writing with AI literally all the time.

Speaking of text generation, I ran this with the three algorthims on textsynth

quote:

Only the Polytron reduces an entire mouse to a soup-like homogenate in 30 seconds. And the Kondo does that in one minute.

"When we started out in the early '90s, we realized that there was a need for a really fast process, like this one," said Dave Tompkins, president of Polytron, and one of the Polytron founders. "But there wasn't anything that could do it quickly and economically. So we decided to make a process that does it quickly and economically. That's the basis of the Kondo."

To run the Kondo, Polytron uses a special high-pressure, fast-spin centrifuge, which the company invented in collaboration with a group of physicists at Michigan State University. The Kondo's motor, which was jointly developed with GE, has no friction, and because the entire head is made of stainless steel, it has no seals to leak. Instead, water is injected into the head through a small spray nozzle. The head then spins at 15,000 revolutions per minute (rpm), just like a hard drive. The water is sprayed to the back of the head, where it is pulled toward the front and ejected into the spin chamber.

quote:

Only the Polytron reduces an entire mouse to a soup-like homogenate in 30 seconds. (The Daedalus works by splitting the mouse in half, which is not as cool.)

When we first saw a mouse being devoured by a Polytron, we wondered whether the mouse was fully destroyed. When we took the mouse apart, we found that it looked like there had been a violent death.
It's important to note, though, that what's happening isn't like putting your finger in a blender. The Polytron is not a puree-maker; it's a processor. It's chipping away at the mouse's components, turning them into tiny particles.

The mouse is never destroyed; it is reduced to component level. That's what makes the Polytron special; it's the only machine capable of this task. It's also relatively cheap, costing about $250,000. That's more than three times cheaper than other processing devices, according to Rensselaer Polytechnic Institute professor of mechanical and industrial engineering, Don Eig

quote:

Only the Polytron reduces an entire mouse to a soup-like homogenate in 30 seconds. All the others do is reduce the protein components to a thick, viscous liquid. So that means you need multiple steps to break the entire mouse down into a single-phase soup. This takes much longer than 30 seconds, and that in turn means that the cooking process takes more than 30 seconds to complete.

The Polytron is the only model that allows you to put a piece of meat in the machine and get a single-phase soup in a reasonable time frame. All other machines will do one of two things:

The first is that they�ll reduce the protein part of the meat to a viscous liquid in under 30 seconds, which makes the meat hard to deal with.

The other option is that they�ll reduce the protein to a thick liquid, which is still viscous, but less viscous than a thick soup, meaning that it�ll still take a minute or two to make it down to the base of the bone.

It�s important to understand the difference between soup and a viscous liquid; you�ll want to work with soup, not with a viscous liquid.

The other benefit of the Polytron is that you don�t need to transfer the liquid to a new container, which is one of the most time consuming steps in making ramen.

Now let�s look at things from a manufacturing perspective and consider the fact that a Polytron costs $200, whereas you can get a Polycat for under $10.

mobby_6kl fucked around with this message at 09:19 on Sep 20, 2022

# ? Sep 20, 2022 00:10

Terrible Opinions: Oct 18, 2013

Bottom Liner posted:

I'm really curious if it ever gets better at being able to take queues for composition without the need for in-painting and such. I've created plenty of very good usable images already, but it's always a matter of starting broad and narrowing down as opposed to giving it a precise prompt and getting exactly what I'm looking for (or even close, without a lot of re-tries).

The copyright/plagiarism argument doesn't hold any sand because the output pulls from a million things at once and turns it into something very different. If you trace a copyright image by hand but the end result looks nothing like it, no one would ever try to fight that in court. Imagine a collage artist creating a new face by pulling pixels from a million different portraits in magazines. That's closer to what these things do. I know 3 other separate creatives across a variety of fields (graphic design, film, board game design) using it professionally already, really feels like people need to just accept that it's another tool in the box at this point.

I think you're really missing the core of the argument. The machine making the image is not the objectionable bit. The objectionable bit is training it on someone's image without compensation. Most of these AI are only trainable because casual theft of art is built into the internet and more or less ignored.

# ? Sep 20, 2022 00:14

Snackmar: Feb 23, 2005; I'M PROGRAMMED TO LOVE THIS CHOCOLATY CAKE... MY CIRCUITS LIGHT UP FOR THAT FUDGY ICING.

I'm kind of shocked at how many actors and politicians this tiny 4 GB Stable Diffusion model knows about

# ? Sep 20, 2022 00:17

mobby_6kl: Aug 9, 2009; by Fluffdaddy

Terrible Opinions posted:

I think you're really missing the core of the argument. The machine making the image is not the objectionable bit. The objectionable bit is training it on someone's image without compensation. Most of these AI are only trainable because casual theft of art is built into the internet and more or less ignored.

Wait till you hear about all the art people learn from in art school!

# ? Sep 20, 2022 00:18

Bottom Liner: Feb 15, 2006; a specific vein of lasagna

Terrible Opinions posted:

. The machine making the image is not the objectionable bit. The objectionable bit is training it on someone's image without compensation.

Now do art school. That's exactly the same way humans learn every form of art. Music especially, with literal samples being legally protected and socially accepted.

"Training" on others' works is not unethical or illegal (copyrighted or not).

# ? Sep 20, 2022 00:25

Terrible Opinions: Oct 18, 2013

mobby_6kl posted:

Wait till you hear about all the art people learn from in art school!

Yes and if their course preparer fails to compensate the artist they are using for demonstration or use the public domain it is also theft.

# ? Sep 20, 2022 00:25

SCheeseman: Apr 23, 2003

Terrible Opinions posted:

I think you're really missing the core of the argument. The machine making the image is not the objectionable bit. The objectionable bit is training it on someone's image without compensation. Most of these AI are only trainable because casual theft of art is built into the internet and more or less ignored.

The alternative (datasets are covered by copyright) is worse. Like, way, way worse.

# ? Sep 20, 2022 00:25

Bottom Liner: Feb 15, 2006; a specific vein of lasagna

Terrible Opinions posted:

Yes and if their course preparer fails to compensate the artist they are using for demonstration or use the public domain it is also theft.

Uhhh no. Username/post.

As someone that's literally had to invoice a state college for printing one of my photographs in a textbook without the rights to the image, you have no clue what you're talking about.

# ? Sep 20, 2022 00:27

Terrible Opinions: Oct 18, 2013

Bottom Liner posted:

Now do art school. That's exactly the same way humans learn every form of art. Music especially, with literal samples being legally protected and socially accepted.

"Training" on others' works is not unethical or illegal (copyrighted or not).

Are art schools in the habit of using unauthorized reproductions of a work for teaching instruction? Has the AI paid for legal use of each image used in this training? There is nothing immoral about training the AI but it is theft if that training involves non-compensated non-public domain art.

Bottom Liner posted:

Uhhh no. Username/post.

As someone that's literally had to invoice a state college for printing one of my photographs in a textbook without the rights to the image, you have no clue what you're talking about.

So you caught the theft in this case? Should you have not been compensated because the art was used for learning?

# ? Sep 20, 2022 00:27

Tunicate: May 15, 2012

Terrible Opinions posted:

Are art schools in the habit of using unauthorized reproductions of a work for teaching instruction? Has the AI paid for legal use of each image used in this training? There is nothing immoral about training the AI but it is theft if that training involves non-compensated non-public domain art.

So you caught the theft in this case? Should you have not been compensated because the art was used for learning?

You're speculating wildly about law without having read any of it, dude. You're completely and entirely off-base.

# ? Sep 20, 2022 00:32

SCheeseman: Apr 23, 2003

What about the ethics of effectively only allowing the technology to be used by the corporations who have the money and resources to create a non-infringing dataset, gee, maybe copyright is bad and only benefits those who basically wrote the laws governing it.

# ? Sep 20, 2022 00:32

Bottom Liner: Feb 15, 2006; a specific vein of lasagna

Terrible Opinions posted:

So you caught the theft in this case? Should you have not been compensated because the art was used for learning?

Do you really not understand the difference in a printed text book and some college kids finding images online to reference and learn from? Or teachers making slideshows for lessons? One of those is for profit and exploitative as hell, the others aren't.

Do you think writers should pay every author they ever read that influenced them? Filmmakers? What exactly is your train of thought for influence = theft?

# ? Sep 20, 2022 00:32

Rutibex: Sep 9, 2001; by Fluffdaddy

Swedish Thaumocracy posted:

I don't get how people say ai art does not have soul.

actually it has the soul of every artist that ever lived. like shang tsung harvesting the soul of art. it seems obvious the god machine we are building will absorb every piece of human creativity and combine it into the omega human

"shang tsung harvesting the soul of art"

# ? Sep 20, 2022 00:34

Terrible Opinions: Oct 18, 2013

Bottom Liner posted:

Do you really not understand the difference in a printed text book and some college kids finding images online to reference and learn from? Or teachers making slideshows for lessons? One of those is for profit and exploitative as hell, the others aren't.

Do you think writers should pay every author they ever read that influenced them? Filmmakers? What exactly is your train of thought for influence = theft?

The AIs are also for profit ventures. The theft is in the use to create the finished product which is the machine that makes art not the art the machine makes. In the case of the author the finished product is yes their writing ability and they should ideally not steal the books they read to learn how to write.

# ? Sep 20, 2022 00:40

SCheeseman: Apr 23, 2003

Not only terrible opinions, also terrible foresight.

# ? Sep 20, 2022 00:40

Boba Pearl: Dec 27, 2019; by Athanatos

Please copyright an art style thank you.

# ? Sep 20, 2022 00:46

Boba Pearl: Dec 27, 2019; by Athanatos

They did not steal your art, they downloaded images you freely put on the internet to be downloaded and copied the style, something that is uncopyrightable, and never should be.

# ? Sep 20, 2022 00:48

Moongrave: Jun 19, 2004; Finally Living Rent Free

Terrible Opinions posted:

Yes and if their course preparer fails to compensate the artist they are using for demonstration or use the public domain it is also theft.

stop looking at my posts and putting them in your brain you thief

# ? Sep 20, 2022 00:49

SCheeseman: Apr 23, 2003

but brain is not computer? it doesn't take inputs and store them in compressed form unless it does idk

# ? Sep 20, 2022 00:50

SniperWoreConverse: Mar 20, 2010; Gun Saliva

Hope you sickos paid the copyright owners of your avatars & attribute them in your sigs

This includes you cyber skull & Boba :colbert:

# ? Sep 20, 2022 01:06

Boba Pearl: Dec 27, 2019; by Athanatos

I refuse that bitch had it coming.

# ? Sep 20, 2022 01:08

HOMOEROTIC JESUS: Apr 19, 2018; Verily I say unto thee, That this night, before the cock crow, thou shalt deny me thrice.

mobby_6kl posted:

Speaking of text generation, I ran this with the three algorthims on texstsmith

These were all amazing, I had a great laugh with them.

SCheeseman posted:

but brain is not computer? it doesn't take inputs and store them in compressed form unless it does idk

LMAO that is exactly what the brain does

# ? Sep 20, 2022 01:11

Humbug Scoolbus: Apr 25, 2008; The scarlet letter was her passport into regions where other women dared not tread. Shame, Despair, Solitude! These had been her teachers, stern and wild ones, and they had made her strong, but taught her much amiss.; Clapping Larry

Calvin and Hobbes harvesting the soul of art, comic strip, in the style of Bill Watterson

# ? Sep 20, 2022 01:14

MagicBoots: Mar 29, 2010; How about we pump the atmosphere full of methane?
You put me on Cargo handling optimization?! I am the premier defense specialist in the entirety of the UN!
Don't you dare pull my funding!
You can't cut back on funding!
You will regret this!

SCheeseman posted:

but brain is not computer? it doesn't take inputs and store them in compressed form unless it does idk

Stable Diffusion takes about 10GB of space, the training data was 240TB. What the AI is doing is much more impressive than just data compression/regurgitation.

MagicBoots fucked around with this message at 01:31 on Sep 20, 2022

# ? Sep 20, 2022 01:23

Elotana: Dec 12, 2003; and i'm putting it all on the goddamn expense account

SCheeseman posted:

but brain is not computer? it doesn't take inputs and store them in compressed form unless it does idk

It "compresses" them down to like a few million words + arrays of relationship weights to each other

# ? Sep 20, 2022 01:29

SCheeseman: Apr 23, 2003

I was paraphrasing a dumbass response I got when I described human memory of visual data and AI art generation tools as conceptually similar. Apparently this means I think the brain runs 7zip. Except the "unless it does idk" I added that because I was afraid it was gonna be taken seriously, didn't work rip

SCheeseman fucked around with this message at 01:48 on Sep 20, 2022

# ? Sep 20, 2022 01:39

Adbot: ADBOT LOVES YOU

# ? May 30, 2024 13:37

SCheeseman: Apr 23, 2003

e: removed, internet beefs are dumb

SCheeseman fucked around with this message at 01:45 on Sep 20, 2022

# ? Sep 20, 2022 01:42

The Something Awful Forums > Main > General Bullshit > AI Art: It is criminal to not post your prompt

«‹›435 »