Yes there is a local vs global optima conversation. OTOH this is happening because of historically aggressive behavior by aggregation startups monetizing other people’s work.
What is the popular view on this? Investor bias aside 😉
its popular for artists interested in protecting their work for sure. i think the misconception is that ai companies are specifically trying to replicate their styles. they're not even trying to do that - if you're popular enough that you have a distinctively recognizable style then your style will be in pop culture and unavoidably in the training data anyway, in enough non nightshaded versions and detail level that someone will be able to produce work that looks like yours. but its not because someone is out to replace you/discredit your work, its because your name has successfully become associated with a distinct style that someone wants.
ultimately this will be popular for an "old guard" of artists who think they're sticking a big middle finger at ai companies using these things; and equally there will be a "new guard" of artists that are more ai native.
Prior to LLMs -> mimicking someone's style in art/literature was harder and required serious domain skill to execute successfully. Post LLM world, this does not require domain skill anymore, rather the ability to learn emerging APIs and good prompt engineering.
It is a strange thing, as a society we are simultaneously arguing that reusing content (plagiarism) is venal when done on an individual scale (see Claudine Gay and Neri Oxman) but when done via a platform at a larger scale that it is inevitable. IMO, a generation of creatives are in effect collateral damage for the emerging economic model powered by LLMs. No one wants to just come out and say it to their faces :)
Y'all have one of the best pods/YT channel AFAIK to track this space. Thank you so much ! and keep up the good work :-)
Sweet recap. Would love to hear from y’all on the nightshade conversations (https://futurism.com/the-byte/nightshade-poison-ai-models)
Yes there is a local vs global optima conversation. OTOH this is happening because of historically aggressive behavior by aggregation startups monetizing other people’s work.
What is the popular view on this? Investor bias aside 😉
its popular for artists interested in protecting their work for sure. i think the misconception is that ai companies are specifically trying to replicate their styles. they're not even trying to do that - if you're popular enough that you have a distinctively recognizable style then your style will be in pop culture and unavoidably in the training data anyway, in enough non nightshaded versions and detail level that someone will be able to produce work that looks like yours. but its not because someone is out to replace you/discredit your work, its because your name has successfully become associated with a distinct style that someone wants.
ultimately this will be popular for an "old guard" of artists who think they're sticking a big middle finger at ai companies using these things; and equally there will be a "new guard" of artists that are more ai native.
Prior to LLMs -> mimicking someone's style in art/literature was harder and required serious domain skill to execute successfully. Post LLM world, this does not require domain skill anymore, rather the ability to learn emerging APIs and good prompt engineering.
It is a strange thing, as a society we are simultaneously arguing that reusing content (plagiarism) is venal when done on an individual scale (see Claudine Gay and Neri Oxman) but when done via a platform at a larger scale that it is inevitable. IMO, a generation of creatives are in effect collateral damage for the emerging economic model powered by LLMs. No one wants to just come out and say it to their faces :)
Y'all have one of the best pods/YT channel AFAIK to track this space. Thank you so much ! and keep up the good work :-)
thanks very very much for the kind words! gets lonely sometimes :)