7 ms·

Note that the guy who submitted the images to the contest created hundreds of images, and spent weeks finetuning the prompts and curating the results. Midjourn

by dcastm 4y ago

Note that the guy who submitted the images to the contest created hundreds of images, and spent weeks finetuning the prompts and curating the results. Midjourney/Stable Diffusion/DALLE are doing to Photoshop what it did to traditional drawing methods. But there's still a human in the loop.

hourago 4y ago
> Note that the guy who submitted the images to the contest created hundreds of images, and spent weeks finetuning the prompts and curating the results. Let a monkey write for a million years and choose the best writing. I think that currently AI is better than that, but not so far as to be a misleading comparation.
- danogentili 4y ago
  Why the downvotes? This is the perfect analogy :)
- tromp 4y ago
  If the monkey only has 10% chance to pick each word appropriately, then in a million years of picking one word per second, you expect only 14 consecutive appropriate words. To get just 100 consecutive appropriate words would take the monkey 10^100 seconds, or over 10^83 billion years...
mid-kid 4y ago
This. People are saying an AI made it, but it sure as hell couldn't have without serious time and effort by the author.
Last5Digits 4y ago
There is still a human in the loop for now, but I'd wager a guess that this won't be the case for long. A lot of the "humans will still be needed" arguments implicitly rely on the assumption that the trend of AI progress stops here. The original DALL-E was released just a year ago. In one year, AI image generation went from generating images that looked slightly normal if you squinted really hard, to the beautiful, detailed and creative results we see today. The same speed of progress can be observed in almost all other domains, from code generation to question answering and music synthesis. For many of these problems, it almost feels like the next step up in quality is just two hours of planning and a couple months of training away. Like they're already solved and just waiting for someone to implement the solution. So yeah, humans are still needed right now. But I wouldn't base my hopes for the future on that, because the world of AI capabilities will probably look very different in just one year.
- vintermann 4y ago
  The intention behind the work is not something that can be automated, since intentionality is (by definition!) a teleological notion. No matter how complicated a process is, I'm not obliged to ascribe intentionality to it. Looking at an ice crystal through a microscope I may be awestruck by it's complexity and beauty, but it won't be art, because there's no person making it and having an intention with it (except, I suppose, God, if you allow that). Conversely, even if something IS made deliberately and by a human, I'm not forced to recognize it as intentional. I could likewise argue that the guy making it is just a big blob of molecules following (mostly) deterministic rules, and deny them personhood. It would be a pretty shitty thing to do, but it wouldn't be factually wrong. Or right. As I said, it's a teleological choice to ascribe something the ability to have intentions. But it would not make much sense to ascribe intentionality to both a computer algorithm and the person using it, seeing as the former is a product of the latter.
- thedorkknight 4y ago
  If/when we get to the point that AI is literally seeking out State Faire art contests and initiating it's own submissions without any human intervention whatsoever, we're going to be having an entirely different conversation
Theodores 4y ago
Exactly. However I think he was trying to get this story. I am spending far too long on OpenAI and I see it as next level creative tool. It is not a gimmick like that deep mind Python app that Google had a few years ago (that turned everything into an uncanny dogs head). What I like is the confidence this new technology has. I have seen lots of things go digital over the years and always found an artist backlash. Digital photography, even Photoshop itself. You can create the next Mona Lisa with nothing more than lasers and space dust and an artist will snub you and go back to making their little clay pots. With this new OpenAI the results are compelling and there is no denying it. You also don't need to know any language apart from English. It is not like you need to do something like know how matrices multiply or do trigonometry (yep, that basic stuff we learn at school and never use in real life). With any tech that is 'wow' you have to make the tech not part of the product. Imagine you have a new image viewer that can give you infinite zoom. You use it to show a gallery of blueprints from a museum. Although the new zoom tool enables you to show off these blueprints, you need to just make it utterly understated and not advertised as 'wow infinite zoom!'. You have to keep focused on the content, the blueprints and making the tech transparent to the objective. With OpenAi it is the same deal, we have the gimmick of fooling the experts. But then when used day to day then it is just another tool to tell the story and to craft the art.
antegamisou 4y ago
The only ones supporting DALLE/Stable Diffusion is going to 'kill' human-made art are tasteless nerds that can't differentiate the Guernica from the last Avengers movie poster. Also funny how everyone reduces art to painting or even worse digital illustration. Talk about 'AI killing art' once the computers create something superior (not a matter of 'subjectivity') to Amore e Psiche.
- fleddr 4y ago
  I have very bad news: most people are "tasteless". In fact, an art judge, supposedly with taste, couldn't tell it was AI.
  - antegamisou 4y ago
    > Most people are "tasteless" Thankfully doesn't apply to people who are familiar with the works of Da Vinci, Michelangelo, Picasso, Klimt, Dali.
    - fleddr 4y ago
      We don't disagree here. There's merit in taste. At the same time I'm painting the macro picture where most artistic labour is of a commercial nature. Where output is leading as long as it is good enough to serve the purpose: sales, entertainment, decoration, whichever.

by dcastm 4y ago

hourago 4y ago

> Note that the guy who submitted the images to the contest created hundreds of images, and spent weeks finetuning the prompts and curating the results. Let a monkey write for a million years and choose the best writing. I think that currently AI is better than that, but not so far as to be a misleading comparation.

danogentili 4y ago
Why the downvotes? This is the perfect analogy :)
tromp 4y ago
If the monkey only has 10% chance to pick each word appropriately, then in a million years of picking one word per second, you expect only 14 consecutive appropriate words. To get just 100 consecutive appropriate words would take the monkey 10^100 seconds, or over 10^83 billion years...

mid-kid 4y ago

This. People are saying an AI made it, but it sure as hell couldn't have without serious time and effort by the author.

Last5Digits 4y ago

There is still a human in the loop for now, but I'd wager a guess that this won't be the case for long. A lot of the "humans will still be needed" arguments implicitly rely on the assumption that the trend of AI progress stops here. The original DALL-E was released just a year ago. In one year, AI image generation went from generating images that looked slightly normal if you squinted really hard, to the beautiful, detailed and creative results we see today. The same speed of progress can be observed in almost all other domains, from code generation to question answering and music synthesis. For many of these problems, it almost feels like the next step up in quality is just two hours of planning and a couple months of training away. Like they're already solved and just waiting for someone to implement the solution. So yeah, humans are still needed right now. But I wouldn't base my hopes for the future on that, because the world of AI capabilities will probably look very different in just one year.

vintermann 4y ago
The intention behind the work is not something that can be automated, since intentionality is (by definition!) a teleological notion. No matter how complicated a process is, I'm not obliged to ascribe intentionality to it. Looking at an ice crystal through a microscope I may be awestruck by it's complexity and beauty, but it won't be art, because there's no person making it and having an intention with it (except, I suppose, God, if you allow that). Conversely, even if something IS made deliberately and by a human, I'm not forced to recognize it as intentional. I could likewise argue that the guy making it is just a big blob of molecules following (mostly) deterministic rules, and deny them personhood. It would be a pretty shitty thing to do, but it wouldn't be factually wrong. Or right. As I said, it's a teleological choice to ascribe something the ability to have intentions. But it would not make much sense to ascribe intentionality to both a computer algorithm and the person using it, seeing as the former is a product of the latter.
thedorkknight 4y ago
If/when we get to the point that AI is literally seeking out State Faire art contests and initiating it's own submissions without any human intervention whatsoever, we're going to be having an entirely different conversation

Theodores 4y ago

Exactly. However I think he was trying to get this story. I am spending far too long on OpenAI and I see it as next level creative tool. It is not a gimmick like that deep mind Python app that Google had a few years ago (that turned everything into an uncanny dogs head). What I like is the confidence this new technology has. I have seen lots of things go digital over the years and always found an artist backlash. Digital photography, even Photoshop itself. You can create the next Mona Lisa with nothing more than lasers and space dust and an artist will snub you and go back to making their little clay pots. With this new OpenAI the results are compelling and there is no denying it. You also don't need to know any language apart from English. It is not like you need to do something like know how matrices multiply or do trigonometry (yep, that basic stuff we learn at school and never use in real life). With any tech that is 'wow' you have to make the tech not part of the product. Imagine you have a new image viewer that can give you infinite zoom. You use it to show a gallery of blueprints from a museum. Although the new zoom tool enables you to show off these blueprints, you need to just make it utterly understated and not advertised as 'wow infinite zoom!'. You have to keep focused on the content, the blueprints and making the tech transparent to the objective. With OpenAi it is the same deal, we have the gimmick of fooling the experts. But then when used day to day then it is just another tool to tell the story and to craft the art.

antegamisou 4y ago

The only ones supporting DALLE/Stable Diffusion is going to 'kill' human-made art are tasteless nerds that can't differentiate the Guernica from the last Avengers movie poster. Also funny how everyone reduces art to painting or even worse digital illustration. Talk about 'AI killing art' once the computers create something superior (not a matter of 'subjectivity') to Amore e Psiche.

fleddr 4y ago
I have very bad news: most people are "tasteless". In fact, an art judge, supposedly with taste, couldn't tell it was AI.
- antegamisou 4y ago
  > Most people are "tasteless" Thankfully doesn't apply to people who are familiar with the works of Da Vinci, Michelangelo, Picasso, Klimt, Dali.
  - fleddr 4y ago
    We don't disagree here. There's merit in taste. At the same time I'm painting the macro picture where most artistic labour is of a commercial nature. Where output is leading as long as it is good enough to serve the purpose: sales, entertainment, decoration, whichever.