Well for one, they didnt come out with a Ghibli filter, they just removed restrictions on imitating art styles and using ghibli style just became a trend.
And honestly who cares what Miyazaki thinks? He’s famously stuck in the past, stubborn and kind of elitist.
A lot of people care what Miyazaki thinks because he created over a dozen films that are beloved by millions because of their artistry, is well respected in the anime/manga world, and is generally regarded as a master of his artform. People tend to take your words seriously when they have nearly 50 years of experience and success behind them.
The issue with Rowling is more that she started talking out her rear end about something of which she doesn’t have much, if any, understanding. If she gave you advice on writing YA fiction, it’d be worth something. Miyazaki is a visual artist broadly respected for his art, so his opinions on visual art have some weight. If it was about the cultivation of kumquats, I think I’d ask a farmer.
Almost no one knows how SD works. That’s not the point. He’s not contrasting it against some other GenAI concept to compare training cost. He’s looking at it based on the results. You don’t have to know how to build a CPU to compare benchmarks for ones built by someone else.
Except how AI works is pretty crucial to the entire anti-AI argument.
The amount of people that claim AI just collages together pieces of existing “stolen” art and use that as an argument against AI is ridiculous.
And your CPU example isn’t great since you would be comparing CPUs to other CPUs, it would be more apt to talk about someone who doesn’t know how a computer works to demonise computers in general and advocate doing maths by hand instead.
Actually, yeah, that’s a bad metaphor from me. Comparing benchmarks would be comparing AI models.
He’s not comparing benchmarks. He’s comparing results, so it’s more about maybe error rate than processing speed.
I guess it’s like comparing the result of an approximation versus an explicit computation. GenAI makes an approximation of art. It very quickly spits out something that looks a bit like the intended answer. It even gets you close enough to be totally satisfactory for some purposes, in the same way 3 can be a usable approximation of π for some purposes. However, the picture is not the full purpose of creating art. Art is a form of communication, transmitting something from one mind to another using indirect means because telepathy isn’t available. AI is not trying to communicate anything. It’s just an approximation of something someone could say.
Miyazaki is someone with years of experience in creating art so he understands the ‘language’ of art better than some. He has ‘fluency.’ AI images hit the uncanny valley for artists because they are attuned to the difference between what an art is supposed to look like vs what the imitator approximates. They have the fluency to spot the fake the same way you might be able to spot someone speaking your native language as a mother tongue vs out of a phrasebook. Because he is ‘fluent’ in art, people take his words on art seriously, just as one would generally take a born-and-raised German’s words seriously regarding German grammar.
Art is a form of communication, transmitting something from one mind to another using indirect means because telepathy isn’t available. AI is not trying to communicate anything.
That’s like saying photography isn’t art because a camera isn’t trying to communicate something. Like AI it’s a tool uses by people to convey that idea.
Like when I use AI generation, I have an idea or specific image in my head and I do my best to come up with a prompt that will produce what I want, or more usually, I use photoshop, so piece piece several pieces together and edit it a bit so the match is more accurate. At a fundamental level, if you consciously try to clear your mind of any existing biases regarding AI, then it’s not really any different to photography or photoshop as an artform.
AI images hit the uncanny valley for artists because they are attuned to the difference between what an art is supposed to look like vs what the imitator approximates.
That is a somewhat valid point, but there are AI models for specific tasks, say generating human faces, that co trolled experiments have found that people can’t distinguish between the AI content and the real thing. There is also plenty of traditional art that hits the uncanny valley or simply doesn’t look right, but that doesn’t make it any less art, does it?
Because he is ‘fluent’ in art, people take his words on art seriously, just as one would generally take a born-and-raised German’s words seriously regarding German grammar.
Good analogy, but there’s still a barrier between the type of art miyazaki is fluent in and AI art. Like imagine a British person saying your English is wrong because you’re using American English or AAVE. Would you take them as an expert because and denounce those variations because they are not British English? Or would you consider their ignorance of the other side limiting to their expertise?
And also, I feel I should add, Miyazaki is famously not a fan of digital art. Should we take him as an expert on art and view digital art as less than traditional art? Or should we just roll our eyes at the stubborn old man stuck in his ways?
I am honestly not informed enough to say whether the whole Ghibli trend grew organically, or whether it came from OAI marketing.
And yeah, Myasaki can be a bit of a dick, but I don’t think he’s wrong here. I worked with both LLM and image generation for some personal projects, and you have to coax it a lot to get anything halfway usable, and even then, it isn’t great. Even stuff that’s heralded as exemplary by the corps behind the trend mostly seems kinda shitty.
Also, even if you like GenAI, most of the stuff that OAI is doing right now feels like desperate attempts to keep the hype train rolling, to justify their frankly ludicrous valuation.
Well for one, they didnt come out with a Ghibli filter, they just removed restrictions on imitating art styles and using ghibli style just became a trend.
And honestly who cares what Miyazaki thinks? He’s famously stuck in the past, stubborn and kind of elitist.
brushes off block button
“I need you again old friend.”
A lot of people care what Miyazaki thinks because he created over a dozen films that are beloved by millions because of their artistry, is well respected in the anime/manga world, and is generally regarded as a master of his artform. People tend to take your words seriously when they have nearly 50 years of experience and success behind them.
Yeah and JK Rowling created one the biggest and most popular series of all time.
That doesn’t mean her opinions deserve any sort of special consideration.
The issue with Rowling is more that she started talking out her rear end about something of which she doesn’t have much, if any, understanding. If she gave you advice on writing YA fiction, it’d be worth something. Miyazaki is a visual artist broadly respected for his art, so his opinions on visual art have some weight. If it was about the cultivation of kumquats, I think I’d ask a farmer.
Yeah maybe, but I would be willing to bet a lot of money Miyazaki couldn’t tell you how stable diffusion worked if his life depended on it.
Almost no one knows how SD works. That’s not the point. He’s not contrasting it against some other GenAI concept to compare training cost. He’s looking at it based on the results. You don’t have to know how to build a CPU to compare benchmarks for ones built by someone else.
Except how AI works is pretty crucial to the entire anti-AI argument.
The amount of people that claim AI just collages together pieces of existing “stolen” art and use that as an argument against AI is ridiculous.
And your CPU example isn’t great since you would be comparing CPUs to other CPUs, it would be more apt to talk about someone who doesn’t know how a computer works to demonise computers in general and advocate doing maths by hand instead.
Actually, yeah, that’s a bad metaphor from me. Comparing benchmarks would be comparing AI models.
He’s not comparing benchmarks. He’s comparing results, so it’s more about maybe error rate than processing speed.
I guess it’s like comparing the result of an approximation versus an explicit computation. GenAI makes an approximation of art. It very quickly spits out something that looks a bit like the intended answer. It even gets you close enough to be totally satisfactory for some purposes, in the same way 3 can be a usable approximation of π for some purposes. However, the picture is not the full purpose of creating art. Art is a form of communication, transmitting something from one mind to another using indirect means because telepathy isn’t available. AI is not trying to communicate anything. It’s just an approximation of something someone could say.
Miyazaki is someone with years of experience in creating art so he understands the ‘language’ of art better than some. He has ‘fluency.’ AI images hit the uncanny valley for artists because they are attuned to the difference between what an art is supposed to look like vs what the imitator approximates. They have the fluency to spot the fake the same way you might be able to spot someone speaking your native language as a mother tongue vs out of a phrasebook. Because he is ‘fluent’ in art, people take his words on art seriously, just as one would generally take a born-and-raised German’s words seriously regarding German grammar.
That’s like saying photography isn’t art because a camera isn’t trying to communicate something. Like AI it’s a tool uses by people to convey that idea.
Like when I use AI generation, I have an idea or specific image in my head and I do my best to come up with a prompt that will produce what I want, or more usually, I use photoshop, so piece piece several pieces together and edit it a bit so the match is more accurate. At a fundamental level, if you consciously try to clear your mind of any existing biases regarding AI, then it’s not really any different to photography or photoshop as an artform.
That is a somewhat valid point, but there are AI models for specific tasks, say generating human faces, that co trolled experiments have found that people can’t distinguish between the AI content and the real thing. There is also plenty of traditional art that hits the uncanny valley or simply doesn’t look right, but that doesn’t make it any less art, does it?
Good analogy, but there’s still a barrier between the type of art miyazaki is fluent in and AI art. Like imagine a British person saying your English is wrong because you’re using American English or AAVE. Would you take them as an expert because and denounce those variations because they are not British English? Or would you consider their ignorance of the other side limiting to their expertise?
And also, I feel I should add, Miyazaki is famously not a fan of digital art. Should we take him as an expert on art and view digital art as less than traditional art? Or should we just roll our eyes at the stubborn old man stuck in his ways?
I am honestly not informed enough to say whether the whole Ghibli trend grew organically, or whether it came from OAI marketing.
And yeah, Myasaki can be a bit of a dick, but I don’t think he’s wrong here. I worked with both LLM and image generation for some personal projects, and you have to coax it a lot to get anything halfway usable, and even then, it isn’t great. Even stuff that’s heralded as exemplary by the corps behind the trend mostly seems kinda shitty.
Also, even if you like GenAI, most of the stuff that OAI is doing right now feels like desperate attempts to keep the hype train rolling, to justify their frankly ludicrous valuation.
Goodbye