Hi, just a quick note on Firefly –– you DO know that none of the training material was appropriately licensed, right? Existing terms were silently expanded to cover this novel use, and no opt-out is being offered. It's just one in the long, long line of misappropriations by tech companies.
In the case of Firefly, that's not correct. Adobe says it trained its models on Adobe Stock and Creative Cloud content, which under the terms of use going back to at least 2018 is covered under the terms of service. And you can opt out for future images. Meanwhile, Adobe is working on a method of compensating people whose images were used to feed to models. I wrote about this in January: https://photoai.substack.com/p/adobe-using-your-photos-to-train
That said, I suspect the "compensation" will be something like "we've added X months to your plan" or "we've amortized X amount from your account to apply a discount of [some cents] each month you continue to be a subscriber" or something like that. But Adobe could surprise me. And this doesn't get all the other companies off the hook. But I do give Adobe credit for trying to do this in a more fair, hopefully-equitable way.
You’re misinformed. The January brouhaha was not about generative AI. Well, the uproar was, but the opt-out wasn’t. Creative Cloud data is not being used for Firefly. Firefly is sourced separately, on Adobe Stock, and without an opt-out. Check the Firefly FAQ.
Perhaps you can explain more when you say "none of the training material was appropriately licensed." The Firefly FAQ says:
"The first Firefly model is trained on Adobe Stock images, openly licensed content and public domain content, where copyright has expired."
Granted, that's pretty sparse.
The terms of use for Adobe Stock also indicate that Adobe has the right to use material as it sees fit:
"4.2 Licenses to Your Content. Solely for the purposes of operating or improving the Services and Software, you grant us a non-exclusive, worldwide, royalty-free sublicensable, license, to use, reproduce, publicly display, distribute, modify, create derivative works based on, publicly perform, and translate the Content."
That's the current license; I didn't look up any prior versions, but I suspect something like that has been in place since the beginning.
True, there's no option to opt-out of GenAI model training other than to pull one's material from Adobe Stock. But that's always an option.
I'm not trying to be an apologist for what the tech companies are doing, but that's where we are right now.
Exactly. ”Solely for the purposes or operating or improving the Services and Software” Image generators are very clearly an entirely novel and separate service, and one that significantly undercuts Stock contributors.
Further, no ”openly licensed” content I know of permits commercial use without credit, except CC-0.
Which is likely why their Senate testimony used much more cautious wording.
Hi, just a quick note on Firefly –– you DO know that none of the training material was appropriately licensed, right? Existing terms were silently expanded to cover this novel use, and no opt-out is being offered. It's just one in the long, long line of misappropriations by tech companies.
In the case of Firefly, that's not correct. Adobe says it trained its models on Adobe Stock and Creative Cloud content, which under the terms of use going back to at least 2018 is covered under the terms of service. And you can opt out for future images. Meanwhile, Adobe is working on a method of compensating people whose images were used to feed to models. I wrote about this in January: https://photoai.substack.com/p/adobe-using-your-photos-to-train
That said, I suspect the "compensation" will be something like "we've added X months to your plan" or "we've amortized X amount from your account to apply a discount of [some cents] each month you continue to be a subscriber" or something like that. But Adobe could surprise me. And this doesn't get all the other companies off the hook. But I do give Adobe credit for trying to do this in a more fair, hopefully-equitable way.
You’re misinformed. The January brouhaha was not about generative AI. Well, the uproar was, but the opt-out wasn’t. Creative Cloud data is not being used for Firefly. Firefly is sourced separately, on Adobe Stock, and without an opt-out. Check the Firefly FAQ.
Perhaps you can explain more when you say "none of the training material was appropriately licensed." The Firefly FAQ says:
"The first Firefly model is trained on Adobe Stock images, openly licensed content and public domain content, where copyright has expired."
Granted, that's pretty sparse.
The terms of use for Adobe Stock also indicate that Adobe has the right to use material as it sees fit:
"4.2 Licenses to Your Content. Solely for the purposes of operating or improving the Services and Software, you grant us a non-exclusive, worldwide, royalty-free sublicensable, license, to use, reproduce, publicly display, distribute, modify, create derivative works based on, publicly perform, and translate the Content."
That's the current license; I didn't look up any prior versions, but I suspect something like that has been in place since the beginning.
True, there's no option to opt-out of GenAI model training other than to pull one's material from Adobe Stock. But that's always an option.
I'm not trying to be an apologist for what the tech companies are doing, but that's where we are right now.
Exactly. ”Solely for the purposes or operating or improving the Services and Software” Image generators are very clearly an entirely novel and separate service, and one that significantly undercuts Stock contributors.
Further, no ”openly licensed” content I know of permits commercial use without credit, except CC-0.
Which is likely why their Senate testimony used much more cautious wording.