I find that interesting, even if I abhor its existence. Feels a little like SEO tea-leaf reading. "If I prompt this, I get that." It's like reverse-engineering the model so you can game it into doing something specific, without actually understanding how the model works and what it's actually trained on.
I don't understand why you have to abhor its existence. I play d&d. I had 2 of my players come over last night so that I could help them generate their character photos. It took 4 hours to make both their characters. It only took 4 hours because I can prompt engineer, tweak in photoshop, understand img2img blending/inpainting methods. Someday I wont have to speak to the AI in specific ways to get what I want, but until then, prompt engineering is the winning strategy. You can be a prompt engineer and understand how the model works and what it is trained on. Most people dont want to do that. Most people don't know how microwave works, but they know how to use it to get quick fast food.
Ironic, when you are on a site that seems mainly dedicated to people who use specialized language to direct computers to do repetitive things at scale.
Programming languages are for the most part deterministic tools that programmers generally understand and can debug. For this audience, the interpreter or compiler is not a "black box" that we only interact with by giving it inputs and seeing what it does.
Although some prompt engineers may have detailed understanding of the AI model and its training corpus, to most people using this kind of AI today the model is a black box, and it's up to them to try things.
Sure, that happens in programming too, but rarely at the level of a programming language, and much more likely at the level of a complex system of poorly documented and buggy interacting parts, such as a long-standing piece of computing infrastructure that has been iterated upon for decades, and now contains a mish-mash of components, many of which were added by people no longer around to explain how they work.
I am currently writing a blog post about this. Here’s an excerpt:
“Consider, for instance, the simple existence of the term “prompt engineering,” which describes the practice of iteratively fine-tuning the prompts you submit to ChatGPT (or whatever) to get your desired output. Prompt engineering is a matter of some identification of what the model responds to, and a lot of guesswork. It’s a matter of treating the natural language of the prompt into a formal language, almost into a programming language, manipulating symbols which lose their human meaning and intuitive structure like some sort of abstract association game. You’re working with the worst programming language imaginable.”
My angle is that using ChatGPT is like using a English as programming language, and it’s a terrible programming language. The reason “prompt engineering” is so irksome to me is that is supposed to be this new advanced mode of expertise, when in reality it’s just shitty programming.
“The Python/C++/whatever-code isn’t some obstacle that we are trying to overcome. The code is the interface that we designed to be able to program the computer, it’s what we want and need. It’s objective, explicit, unambiguous, (relatively) static, internally consistent, and robust. English has none of these properties—it’s subjective, meaning is often implicit, and ambiguous, and it’s always changing, contradictions appear; its structure does not hold up to analysis.”
English may be a shitty programming language, but wouldn't that imply that wrangling it to get what you want is--in and of itself--a skill, and not "shitty programming" as you put it?
It's definitely a form of authorship--and authorship (of a book) isn't shitty. Writing well is a craft that people hone--to induce wetware computers (i.e., brains) to "hallucinate" (using a term that seems popular with people who interface with these AI tools) what the author desires them too.
Maybe that's the difference? Programming is when you use a very formally constructed language to induce a computer to execute a desired algorithm precisely. Authorship or artistry is when you use a formally loose language (painting, sculpture, writing) to induce a computer to "hallucinate" a desired meme (in the original sense of the word, "unit of meaning").
Both pursuits are somewhat similar, but the breakpoint is the rigor of the language used, maybe?
There is a term-of-art for using English to write programs: "The English-Likeness Monster." The poster-boy for this terrible UX is--ironically--AppleScript, a programming language from a company that otherwise prides itself on good user experiences.
If you are dealing with a labor shortage, then how can you honor an IP? If you can't honor it, then you shouldn't produce it. That there isn't a law for this is unfortunate. Until something is done, we will keep ending up with 2022 Trigun CGI remakes.
Is there a single thing twitter mobs won’t get pissed about? I feel like twitter replies are not the best way to gauge public sentiment on this. Are the fans also mad that artists aren’t drawing each frame individually anymore?
I wish there was some context for declaration of outrage. Are there tens of fans or ten thousands of fans upset about this? What percentages of the viewing market? It’s a bit difficult to raise empathy for three stoned Tweeters on a couch in Reseda vs thousands of unemployed artists.
Some Twitter users do not seem to realize that the video attached to the tweet is the entire anime, and are quote-tweeting the video that they're urging people to pirate.
I point that out because maybe some of the outrage is due to misunderstanding/language barrier, since it seems out of proportion for an experimental short? Maybe it was in poor taste for Netflix's tweet to emphasize the "labor shortage", their article [1] is more about experimenting with new ways for animators to work with AI tools.
Is there actually a labor shortage? If this claim is qualified with "there's a shortage of labor willing to work for a pittance" then there's no labor shortage. I'm more disgusted with a business choosing profits over people, especially when we all know that Netflix budgets are generally very substantive. If the cost of getting the talent you need to produce the work is too high, turning to generative art is a band-aid for a real systemic problem with how Netflix is producing content. AI should be a tool: if the art can't be produced without it, you're really just scraping by. And that's a big red flag for Netflix's business, imo.
There's an irony to me here - one of Anime's most distinctive artistic features (lower framerate, repeating frames/backgrounds) was specifically a way to increase content by reducing animator's workload.
Huh. Backgrounds have long been semi-automated. I'm sure it's putting people out of work, but it's a long-running issue.
I'm not an animator, and I'm sure I don't understand all of the issues. I could see this being terrible for new animators, who expect to get low-visibility work to get experience. Art of any kind has always been a precarious way to make a living, and it would be depressing if studios and audiences settled for something lesser to avoid paying them.
But is this a case where it frees them up to do more work? If fewer animators are involved, does that mean the same animators can produce more films? Studios are desperate for content, which is part of the "labor shortage".
Perhaps it would be an opportunity for more niche animation -- to tell more personal stories with less wide appeal because the overhead is less? There's a ton of such material on Netflix; it seems to be their bread and butter.
I don't want to get all techno-utopia, especially in a domain I don't know. My guesses are not as good as their experience. But I am genuinely curious.
I don't really see the issue, this is how it has always worked. If someone or something can do your job good enough (not necessarily better) but for a fraction of the cost, you will eventually be forced to face that reality.
It doesn't have to always work that way. There's examples of groups of people that optimize for things besides profit. Like the Amish, for example.
Not saying that this needs to apply in any specific example. I'm just trying to advocate for the idea that we can imagine different social constructs as being possible. I like to think that a society can potentially decide to do anything that society agrees they want to do, and that could include valuing people over profits.
Remember watching cartoons as a kid then growing up and realizing where the cost cutting happened based on lack of animated frames, and reused animation cycles?
The kids of the future will get to spot AI deformations, extra digits, missing limbs, non-Euclidean geometry!
40 comments
[ 2.5 ms ] story [ 82.2 ms ] thread"You mean there's a shortage of people willing to work brutal hours for subsistence wages."
Charlie Conservative:
"If your job can be replaced by AI, you deserve to starve, you aren't creating value for investors."
Bertha the Bot-Writer:
"I can replace both of you with generative AI that writes highly viral, low-effort, simplistic and partisan social media posts and comments."
"The new artist jobs will be the people who direct the AIs creating the art."
Gross.
Just add a second button that is "inject happy drugs" and you get a really good scifi novel that we shouldn't want to live in.
Ironic, when you are on a site that seems mainly dedicated to people who use specialized language to direct computers to do repetitive things at scale.
Programming languages are for the most part deterministic tools that programmers generally understand and can debug. For this audience, the interpreter or compiler is not a "black box" that we only interact with by giving it inputs and seeing what it does.
Although some prompt engineers may have detailed understanding of the AI model and its training corpus, to most people using this kind of AI today the model is a black box, and it's up to them to try things.
Sure, that happens in programming too, but rarely at the level of a programming language, and much more likely at the level of a complex system of poorly documented and buggy interacting parts, such as a long-standing piece of computing infrastructure that has been iterated upon for decades, and now contains a mish-mash of components, many of which were added by people no longer around to explain how they work.
“Consider, for instance, the simple existence of the term “prompt engineering,” which describes the practice of iteratively fine-tuning the prompts you submit to ChatGPT (or whatever) to get your desired output. Prompt engineering is a matter of some identification of what the model responds to, and a lot of guesswork. It’s a matter of treating the natural language of the prompt into a formal language, almost into a programming language, manipulating symbols which lose their human meaning and intuitive structure like some sort of abstract association game. You’re working with the worst programming language imaginable.”
My angle is that using ChatGPT is like using a English as programming language, and it’s a terrible programming language. The reason “prompt engineering” is so irksome to me is that is supposed to be this new advanced mode of expertise, when in reality it’s just shitty programming.
“The Python/C++/whatever-code isn’t some obstacle that we are trying to overcome. The code is the interface that we designed to be able to program the computer, it’s what we want and need. It’s objective, explicit, unambiguous, (relatively) static, internally consistent, and robust. English has none of these properties—it’s subjective, meaning is often implicit, and ambiguous, and it’s always changing, contradictions appear; its structure does not hold up to analysis.”
It's definitely a form of authorship--and authorship (of a book) isn't shitty. Writing well is a craft that people hone--to induce wetware computers (i.e., brains) to "hallucinate" (using a term that seems popular with people who interface with these AI tools) what the author desires them too.
Maybe that's the difference? Programming is when you use a very formally constructed language to induce a computer to execute a desired algorithm precisely. Authorship or artistry is when you use a formally loose language (painting, sculpture, writing) to induce a computer to "hallucinate" a desired meme (in the original sense of the word, "unit of meaning").
Both pursuits are somewhat similar, but the breakpoint is the rigor of the language used, maybe?
https://deprogrammaticaipsum.com/the-english-likeness-monste...
People directing AI to create things are not artists.
At best they are art commissioners
What about the Sistine Chapel? Jeff Koons?
http://artandobject.com/news/artist-workshops-their-history-...
Almost the dream technocrats were predicting in the 60s, the only difference is that it's the complete opposite.
As a society our use of social media will one day be looked back on the way we look back on the Romans using lead as a sweetener.
...
I point that out because maybe some of the outrage is due to misunderstanding/language barrier, since it seems out of proportion for an experimental short? Maybe it was in poor taste for Netflix's tweet to emphasize the "labor shortage", their article [1] is more about experimenting with new ways for animators to work with AI tools.
[1] https://about.netflix.com/ja/news/the-dog-and-the-boy
They will choose profits over their own suvival - they sanctionbusted to sell to nazis, choose shirt term profit over long-term growth, etc.
I'm not an animator, and I'm sure I don't understand all of the issues. I could see this being terrible for new animators, who expect to get low-visibility work to get experience. Art of any kind has always been a precarious way to make a living, and it would be depressing if studios and audiences settled for something lesser to avoid paying them.
But is this a case where it frees them up to do more work? If fewer animators are involved, does that mean the same animators can produce more films? Studios are desperate for content, which is part of the "labor shortage".
Perhaps it would be an opportunity for more niche animation -- to tell more personal stories with less wide appeal because the overhead is less? There's a ton of such material on Netflix; it seems to be their bread and butter.
I don't want to get all techno-utopia, especially in a domain I don't know. My guesses are not as good as their experience. But I am genuinely curious.
Not saying that this needs to apply in any specific example. I'm just trying to advocate for the idea that we can imagine different social constructs as being possible. I like to think that a society can potentially decide to do anything that society agrees they want to do, and that could include valuing people over profits.
The kids of the future will get to spot AI deformations, extra digits, missing limbs, non-Euclidean geometry!