- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
re: Netflix Japan uses AI in anime (artists not happy)
Posted on 2/8/23 at 9:34 pm to Fewer Kilometers
Posted on 2/8/23 at 9:34 pm to Fewer Kilometers
quote:
All of these analogies in defense of AI appropriating art, but not one of the examples have the "next step" relying on the previous creators to continue forward. It's not a case of a better mousetrap. It's a program that will require the leech to continuously feed off of the work of others or it dies.
What makes you say this? From what I understand, it’s exactly the opposite - the current iteration of the AI should continue to work just fine without new training data. In fact it should continue improving based on user input without any additional training data.
Maybe you know more than I do, but the bold part above is exactly the opposite of everything I’ve read about currently available AI art generators.
1
Posted on 2/9/23 at 9:19 am to lostinbr
quote:Art evolves. As consumers' tastes evolve from exposure to new artists, the AI will have to ingest new creative. If AI had been created in the 1950's and then starved of new art from that point on, it would now be at an improved state, but based solely on those styles. As Warhol, Frazetta, Basquiat, and Neiman became the popular artists of the day, those are the styles that AI would want to create to sell, but it wouldn't have the database to pull from. You might luck out and get a Patrick Nagel knock-off by AI accidentally combining his influences of pre-60's advertising illustration, pin-ups, and Japanese prints, but the chances of it hitting just as the complimentary 80's fashions made it so popular are extremely slim.
What makes you say this? From what I understand, it’s exactly the opposite - the current iteration of the AI should continue to work just fine without new training data. In fact it should continue improving based on user input without any additional training data.
Maybe you know more than I do, but the bold part above is exactly the opposite of everything I’ve read about currently available AI art generators.
Think of it as programmatic advertising that stopped monitoring consumers when you were 16. It's now serving you ads for items it thinks you're probably into at 46, but it has no idea how you and the world's tastes have changed or that cassette tapes and VHS have been replaced by streaming.
Posted on 2/9/23 at 11:45 am to Fewer Kilometers
quote:
If AI had been created in the 1950's and then starved of new art from that point on, it would now be at an improved state, but based solely on those styles. As Warhol, Frazetta, Basquiat, and Neiman became the popular artists of the day, those are the styles that AI would want to create to sell, but it wouldn't have the database to pull from. You might luck out and get a Patrick Nagel knock-off by AI accidentally combining his influences of pre-60's advertising illustration, pin-ups, and Japanese prints, but the chances of it hitting just as the complimentary 80's fashions made it so popular are extremely slim.
This is a simplistic strawman. First, literally a photo of every piece of art and photographed thing could be input from the beginning of mankind to the 50's. Secondly, every iteration and combination thereof could be run from 1950 to 2023, thousands per second for 73 years. And within the outputs it is very likely that you would recognize styles that have come and gone between the real world 1950 and this made-up Fallout-esque frozen in the 50s time period.
Posted on 2/9/23 at 11:58 am to blackinthesaddle
quote:bullshite.
This is a simplistic strawman. First, literally a photo of every piece of art and photographed thing could be input from the beginning of mankind to the 50's. Secondly, every iteration and combination thereof could be run from 1950 to 2023, thousands per second for 73 years. And within the outputs it is very likely that you would recognize styles that have come and gone between the real world 1950 and this made-up Fallout-esque frozen in the 50s time period.
As we speak, people are using AI to create images that are "in the style of" specific current artists. Your infinite art produced in an infinite universe may theoretically produce every conceivable style of art, but people are still going to want their dog painted in the style of Peter Max. And for that, they'd have to know who Peter Max is and they'd have to show his art catalogue to the AI.
This isn't the automobile replacing the horse. It's the automobile industry requiring the never ending theft of horses in order to keep their industry going.
Posted on 2/9/23 at 9:18 pm to Fewer Kilometers
Long arse post ahead.. apologies.
That’s not really true though (or at least shouldn’t be true for a good AI) because the AI is continuously being “trained” by its users. As the users’ stylistic tastes (and, consequently, their decisions) evolve over time, the AI’s stylistic suggestions should evolve as well.
I don’t know anything about Rinna (the specific AI company used by Netflix in this case) but that’s certainly how tools like Midjourney and DALL-E work. They will continue to learn from users’ selections and the styles they produce will learn and change over time accordingly.
Would I expect some divergence between AI- and human-generated art styles over a long enough period of time? Of course. The amount of divergence is largely a function of the quality of the AI.
That being said, the logic of the argument about AI’s being “static” seems a bit weird to me in the first place. Let’s say it’s all true and AI’s are bad because they can’t “change with the times” without being re-trained on new art. Let’s further say that they are bad for being trained on others’ art in the first place. How, then, do humans keep up with the style “of the times” if not through consumption and imitation of art from other humans?
I completely understand the moral argument about AI’s scraping images (particularly without permission) to then imitate those art styles. But going further to say an AI is weak because it can’t change with the times is kind of admitting that human artists do the same thing.
I think at some point we have to acknowledge that there’s a grey area between truly creative “art” and art that is purely a work product. I think most people would agree that there are ethical issues if someone uses an AI to copy the style of a specific architect and sell those plans. But what about an AI that scrapes existing home designs to generate plans for spec houses? Architecture is ostensibly an art, but is generating plans for spec houses (which are notably devoid of creativity) really “stealing” others’ creative work?
It’s admittedly not a perfect analogy, but it’s intended to illustrate that this isn’t a black and white issue. Graphic design is another example. It’s very clearly a form of art, but it’s also very clearly a field where humans quite often plagiarize whatever styles are in-fashion at the time. Yes, there are cases where a brilliant designer comes up with an iconic logo. But is there really that much ethical difference between the owner of “AAA Storage Solutions” getting his logo from an AI vs. getting it on Fiverr?
In the case of this Netflix show, it appears they used an AI art generator to create the backgrounds/landscapes (rather poorly, at that, from the stills I’ve seen). So while I get that there are some legit moral questions around AI art generators, I’m also not convinced that having an AI fill in textures/colors for trees and hills in conjunction with a working artist who sketched the landscape in the first place is some ethical third rail.
/rant
quote:
Think of it as programmatic advertising that stopped monitoring consumers when you were 16. It's now serving you ads for items it thinks you're probably into at 46, but it has no idea how you and the world's tastes have changed or that cassette tapes and VHS have been replaced by streaming.
That’s not really true though (or at least shouldn’t be true for a good AI) because the AI is continuously being “trained” by its users. As the users’ stylistic tastes (and, consequently, their decisions) evolve over time, the AI’s stylistic suggestions should evolve as well.
I don’t know anything about Rinna (the specific AI company used by Netflix in this case) but that’s certainly how tools like Midjourney and DALL-E work. They will continue to learn from users’ selections and the styles they produce will learn and change over time accordingly.
Would I expect some divergence between AI- and human-generated art styles over a long enough period of time? Of course. The amount of divergence is largely a function of the quality of the AI.
That being said, the logic of the argument about AI’s being “static” seems a bit weird to me in the first place. Let’s say it’s all true and AI’s are bad because they can’t “change with the times” without being re-trained on new art. Let’s further say that they are bad for being trained on others’ art in the first place. How, then, do humans keep up with the style “of the times” if not through consumption and imitation of art from other humans?
I completely understand the moral argument about AI’s scraping images (particularly without permission) to then imitate those art styles. But going further to say an AI is weak because it can’t change with the times is kind of admitting that human artists do the same thing.
I think at some point we have to acknowledge that there’s a grey area between truly creative “art” and art that is purely a work product. I think most people would agree that there are ethical issues if someone uses an AI to copy the style of a specific architect and sell those plans. But what about an AI that scrapes existing home designs to generate plans for spec houses? Architecture is ostensibly an art, but is generating plans for spec houses (which are notably devoid of creativity) really “stealing” others’ creative work?
It’s admittedly not a perfect analogy, but it’s intended to illustrate that this isn’t a black and white issue. Graphic design is another example. It’s very clearly a form of art, but it’s also very clearly a field where humans quite often plagiarize whatever styles are in-fashion at the time. Yes, there are cases where a brilliant designer comes up with an iconic logo. But is there really that much ethical difference between the owner of “AAA Storage Solutions” getting his logo from an AI vs. getting it on Fiverr?
In the case of this Netflix show, it appears they used an AI art generator to create the backgrounds/landscapes (rather poorly, at that, from the stills I’ve seen). So while I get that there are some legit moral questions around AI art generators, I’m also not convinced that having an AI fill in textures/colors for trees and hills in conjunction with a working artist who sketched the landscape in the first place is some ethical third rail.
/rant
Page 2 of 2
Popular
Back to top
Follow TigerDroppings for LSU Football News