pretty good overview! i personally think ais "creativity" comes from the datasets and the "voting" of all the users who use for example midjourney. Because lets be clear machinelearning-programs like these are simply put a valuesystem finetuned over time and these values come from taste of the viewers or creators. the question of ownership should be pretty clear , its not licensable because the data used isnt owned by the user or the company. for example stable diffusion, midjourney and dall-e are all based on the laion5b dataset which is a dataset consisting of a load of private data. every user of stablediffusion/midjourney/dall-e can ask themselfs why should these programs know who "greg rutkowski","craig mullins" or a celebrity is if it isnt using data from websites they shouldnt use the data from without asking for permission? these tools data sets should be made via using bought data from the owners of the data or from license free data (and i think even then the owners should be asked beforehand because it isnt the same scope of use anymore which would have been possible before). and anyone whom thinks artists are just crybabies its the same as using "reference" no it isnt its more like a photocollage adding noise-filter after that denoise-filter nothing more. this doesnt make it creative or ok to use someone elses work at all. i personally think the puplishers should held the most liability for this whole licensedebacle and after that the people whom try to make money of of it xD. to your question i think computers cant be really creative. the programs outputs we interpret as creative are based on human creativity and the correction work of other humans. take away the datasets from these programs and any image the user takes from some site and you will be left with some pretty awful results because if this wouldnt be the case why the use of such a big dataset? thanks for sharing !
Thanks for sharing your thoughts! I agree with you on all points, especially about computers being creative. It’s such a nuanced concept, but the way these models draw ‘inspiration’ from existing artwork is not the same way humans are inspired by other humans to make art.
@@janpancake im happy to share my thoughts ^^ . i havent researched as much as i should have probably but i think when even other CEOs of AI-companies think an AI-model should be highly evaluated before being released/puplished we as customers/bystanders can be suspicious. Yeah its really different. have a nice day :) !
Yayyy Janice is back!!!!
Well done
Nice overview. Thanks
pretty good overview!
i personally think ais "creativity" comes from the datasets and the "voting" of all the users who use for example midjourney. Because lets be clear machinelearning-programs like these are simply put a valuesystem finetuned over time and these values come from taste of the viewers or creators.
the question of ownership should be pretty clear , its not licensable because the data used isnt owned by the user or the company. for example stable diffusion, midjourney and dall-e are all based on the laion5b dataset which is a dataset consisting of a load of private data.
every user of stablediffusion/midjourney/dall-e can ask themselfs why should these programs know who "greg rutkowski","craig mullins" or a celebrity is if it isnt using data from websites they shouldnt use the data from without asking for permission?
these tools data sets should be made via using bought data from the owners of the data or from license free data
(and i think even then the owners should be asked beforehand because it isnt the same scope of use anymore which would have been possible before).
and anyone whom thinks artists are just crybabies its the same as using "reference" no it isnt its more like a photocollage adding noise-filter after that denoise-filter nothing more. this doesnt make it creative or ok to use someone elses work at all.
i personally think the puplishers should held the most liability for this whole licensedebacle and after that the people whom try to make money of of it xD.
to your question i think computers cant be really creative. the programs outputs we interpret as creative are based on human creativity and the correction work of other humans. take away the datasets from these programs and any image the user takes from some site and you will be left with some pretty awful results because if this wouldnt be the case why the use of such a big dataset?
thanks for sharing !
Thanks for sharing your thoughts! I agree with you on all points, especially about computers being creative. It’s such a nuanced concept, but the way these models draw ‘inspiration’ from existing artwork is not the same way humans are inspired by other humans to make art.
@@janpancake im happy to share my thoughts ^^ . i havent researched as much as i should have probably but i think when even other CEOs of AI-companies think an AI-model should be highly evaluated before being released/puplished we as customers/bystanders can be suspicious.
Yeah its really different.
have a nice day :) !