So far, I don't see any evidence to convince me that these AI companies have any long-term plan to license any data at all: they'd much rather see changes to copyright law that allow AI training to be considered "fair use." Any deals they have struck so far have been temporary measures to ensure their data models don't turn into synthetic content cannibals too soon, but my gut says they are waiting to see if Trump wins the election because his admin will be easier to convince that they should be allowed to scrape content with impunity.
I certainly am not convinced either. BUT I see things moving forward toward licensing, due to the ongoing spotlight on the issue -- and the growing realization that massive litigation expenses don't help anyone. Too much risk.
Asking the real questions! And bringing relevant numbers. Great post!
Would be interesting to compare to Big Content licensing approach -- the Gettys, Adobes and Shutterstocks who start from catalogue not algorithm. The initial numbers in from Adobe Stock point to 20% chunks taken out of existing royalty pools while "training bonuses" count in fractions of a cent on the dollar...
Johan, I'd like to learn more about Adobe's "solution" - and how it works. Can you lay it out in more detail - or is there a link you can send me? peter@creativemedia.biz
This is how Credtent.org has built its system. Thanks for the data to support this vision.
So far, I don't see any evidence to convince me that these AI companies have any long-term plan to license any data at all: they'd much rather see changes to copyright law that allow AI training to be considered "fair use." Any deals they have struck so far have been temporary measures to ensure their data models don't turn into synthetic content cannibals too soon, but my gut says they are waiting to see if Trump wins the election because his admin will be easier to convince that they should be allowed to scrape content with impunity.
I certainly am not convinced either. BUT I see things moving forward toward licensing, due to the ongoing spotlight on the issue -- and the growing realization that massive litigation expenses don't help anyone. Too much risk.
Asking the real questions! And bringing relevant numbers. Great post!
Would be interesting to compare to Big Content licensing approach -- the Gettys, Adobes and Shutterstocks who start from catalogue not algorithm. The initial numbers in from Adobe Stock point to 20% chunks taken out of existing royalty pools while "training bonuses" count in fractions of a cent on the dollar...
Johan, I'd like to learn more about Adobe's "solution" - and how it works. Can you lay it out in more detail - or is there a link you can send me? peter@creativemedia.biz
Sure thing