You are viewing a single comment's thread. View all
-2
el_hoovy on scored.co
1 month ago-2 points(+0/-0/-2Score on mirror)1 child
> modern comfyui workflows beat the top digital artists.
as someone who uses the modern comfyui workflows to make covers for his singles, an artist remains indispensable when you have an extremely specific artistic vision, like for the video game i'm developing with two artists.
i mean, i'll freely admit to having used AI just to mass-generate character design concepts, but for every 50 generations you get one good design, and for every 20 good designs only one gets picked out to be implemented, which still has to be rethought and redone by the artist in actual implementation.
> This is reddit-tier ludditism.
it's practical knowledge acquired via observation. AI can only chance upon originality by smashing together everything else it's seen. this is obviously inferior to actually creating something original as a human, which even if it's not 500% alien and original can still be an original natural and cohesive development of other art. randomly putting a victorian hat on a medieval farmer holding an m16 because the prompt contained all of those words is something humans can already do with photoshop, and even the latest LLM won't give you a good story to pair with that anachronistic weirdo.
> We exist in a flood-tide of mental filth that was created by jews doing just this same thing ... Someone on our side has to do it [...]
jews don't mass produce AI slop masquerading as high art like the OP seems to imply is possible (look, we can make sci fi movies!). jews have billion dollar studios with big name actors. the material the AI had to consume to produce things like OP was first produced by (Whites employed by) jews who did it the manual way, and it still is to this day even if to a lesser extent.
besides, my point isn't that we shouldn't counter spam with spam. we should. but the jew does it via subterfuge, astroturfing, etc, not via outsourcing Hollywood to machine learning. we need to build actual communities of people who work together to achieve things like that, we can't do it from our basements and win "meme wars" no matter how advanced AI gets. eventually you need to offer people an alternative to all the filth - you know, like the whole thing we supposedly stand for?
> AI can only chance upon originality by smashing together everything else it's seen.
This is false, as this is the definition of overfitting. Models are intentionally not overtrained so that they learn features and relationships, not raw memorization (called overfitting, and it leads to what you describe, called regurgitation).
The ability to generate without regurgitation is a concept roughly called 'generalization', basically how well a model trained on a number of scenarios will perform on new scenarios it has never seen. This isn't new tech btw, researchers have been studying how well models generalize since the 50s.
Or, in short, well-trained models are literally generating, and literally creating. Models that don't are considered bad models.
> jews don't mass produce AI slop masquerading as high art
This can't be known to be true. Julian Assange's 2017 'proof of life' video famously had weirdness (like his lapel morphing into his shirt), which was claimed to be compression artifacts, but we now know is classic artifacting of movie-generating models.
> eventually you need to offer people an alternative to all the filth
Understandable, but I think the solution here is also AI. There is no way we would have the resources or structure to combat this filth otherwise (without a proper DOTR or actual government cleanout).
as someone who uses the modern comfyui workflows to make covers for his singles, an artist remains indispensable when you have an extremely specific artistic vision, like for the video game i'm developing with two artists.
i mean, i'll freely admit to having used AI just to mass-generate character design concepts, but for every 50 generations you get one good design, and for every 20 good designs only one gets picked out to be implemented, which still has to be rethought and redone by the artist in actual implementation.
> This is reddit-tier ludditism.
it's practical knowledge acquired via observation. AI can only chance upon originality by smashing together everything else it's seen. this is obviously inferior to actually creating something original as a human, which even if it's not 500% alien and original can still be an original natural and cohesive development of other art. randomly putting a victorian hat on a medieval farmer holding an m16 because the prompt contained all of those words is something humans can already do with photoshop, and even the latest LLM won't give you a good story to pair with that anachronistic weirdo.
> We exist in a flood-tide of mental filth that was created by jews doing just this same thing ... Someone on our side has to do it [...]
jews don't mass produce AI slop masquerading as high art like the OP seems to imply is possible (look, we can make sci fi movies!). jews have billion dollar studios with big name actors. the material the AI had to consume to produce things like OP was first produced by (Whites employed by) jews who did it the manual way, and it still is to this day even if to a lesser extent.
besides, my point isn't that we shouldn't counter spam with spam. we should. but the jew does it via subterfuge, astroturfing, etc, not via outsourcing Hollywood to machine learning. we need to build actual communities of people who work together to achieve things like that, we can't do it from our basements and win "meme wars" no matter how advanced AI gets. eventually you need to offer people an alternative to all the filth - you know, like the whole thing we supposedly stand for?
This is false, as this is the definition of overfitting. Models are intentionally not overtrained so that they learn features and relationships, not raw memorization (called overfitting, and it leads to what you describe, called regurgitation).
The ability to generate without regurgitation is a concept roughly called 'generalization', basically how well a model trained on a number of scenarios will perform on new scenarios it has never seen. This isn't new tech btw, researchers have been studying how well models generalize since the 50s.
Or, in short, well-trained models are literally generating, and literally creating. Models that don't are considered bad models.
> jews don't mass produce AI slop masquerading as high art
This can't be known to be true. Julian Assange's 2017 'proof of life' video famously had weirdness (like his lapel morphing into his shirt), which was claimed to be compression artifacts, but we now know is classic artifacting of movie-generating models.
> eventually you need to offer people an alternative to all the filth
Understandable, but I think the solution here is also AI. There is no way we would have the resources or structure to combat this filth otherwise (without a proper DOTR or actual government cleanout).