How is Adobe setting ethical standards for Adobe Firefly?

[EasterlyArt]

AI machine learning is all built off image databases. This means that for any image generation to happen, the systems in place need to be trained on a set of images. That can easily be abused if images are sourced without permission, whole likenesses are easily replicated, or any other non-adobe owned materials are used.

Adobe could help set a great standard here with sourcing images ethically, and for any images provided for the Adobe Firefly database for the AI to be trained on, the owners of said images would be compensated accordindly. Adobe could also do their part to stop any image generation prompts from using copyrighted name or the names of exitsing properites as to prevent anyone from abusing Adobe Firefly. 

That said, I am curious as to what Adobe is actually doing with Adobe Firefly.

AI machine learning image generating systems are amazing, but also have a boat ton of ethical issues that need to be addressed. We need a company like Adobe to help set a standard that can help governments set standards and practices that protect people first and not for profit companies. 

It would be nice to get some statements from Adobe addressing this so we know they're not just kissing up to "the mouse" for their profit driven motives and doing their part to actually protect us as creatives in all forms. 

[CMass]

Thank you for your questions. You can read more about our ethical approach here.

[Kaitlyn28583976c12d]

My understanding is that artists who contribute to Adobe Stock are not able to opt out of their work being used in Firefly's database. Additionally, these artists aren't compensated for their work's use in Firefly's database. This model feels exploitative to say the least. Am I misunderstanding the relationship between Adobe, Adobe Stock, creators, and Firefly? 

[EasterlyArt]

If these users weren't at least offered an email giving the option by saying "Hey, would you like to have your work included in this new thing we're making with AI, and yes we'll pay you for it" seems pretty suspect. Though, maybe that isn't the case and more clarity would be appreciated. 

That said, I did find this thanks to the link CMass posted:

Respecting creators’ choice and control

Creators want control over whether their work is used to train generative AI or not. For some, they want their content out of AI. For others, they are happy to see it used in the training data to help this new technology grow, especially if they can retain attribution for their work. Using provenance technology, creators can attach “Do Not Train” credentials that travel with their content wherever it goes. With industry adoption, this will help prevent web crawlers from using works with “Do Not Train” credentials as part of a dataset. Together, along with exploratory efforts to compensate creators for their contributions, we can build generative AI that both empowers creators and enhances their experiences. 

Though how much of this will be used industry wide is a bigger question and if that will be protected by governments and not just Adobe. 

Also worth bringing up this video looking at the ethics from an outside perspective. Potentially more videos that I could post, but I'll start with this. Also, if you're unable to see the video, just look up "The Problem With Adobe Firefly Is BIGGER Than You Think!" by Theoretically Media. 
https://tinyurl.com/4d8nt6rf 

Ultimately this conversation has only just begun when it comes to AI generated images, video, audio, music, etc. as Firefly is just the beginning. Hopefully the topic of ethics doesn't get stomped out and can remain open as we want to be sure that artists and other creatives who've made a living off original work don't get phased out in a large way by technology that, albeit pretty amazing, is essentially putting people out of work so that companies can has excuses not to pay a living wage to employees. 

[KatarinaJovanovic]

Hey James, 

Model is not trained on whole database of the Adobe Stock, but openly licensed content and public domain content, where copyright has expired. Model is not trained on any trade marked data.

For the contributors who were contributed to training there is a compensatory model 

https://helpx.adobe.com/stock/contributor/help/firefly-faq-for-adobe-stock-contributors.html

[EasterlyArt]

My biggest concern with "openly licensed content" is that databases for that may actually have a fair amount of copyright material that were not approved by the license holders. A quick search for something like "comic books" brought up a number of licensed works by publishers like Marvel and DC. I'm pretty sure the lawyers didn't sign off on that. 

Additionally, for the sake of asking Adobe proper, how do they define "openly licensed content" and what do they see as "public domain content" as well? I'm sure the ethics committee has chatted about this, but considering that in the video talking about ethics they mention talking about transparency, this would be a great opportunity to be transparent. 

Again, this technology is so new, and even with the best intentions, Adobe is still a gigantic for profit entity that may be more swayed by larger investors. Maybe it would be an extreme example to say "the mouse" may have influence over them, or other hyper weathly donors, but I just want to be sure that creative professionals doing the best they can to make ends meet aren't put out of work for the benefit of entities that are larger than anyone could fight against alone. 

Would the ethics committee/team for Adobe be open to discussing how governments can help protect people from the misuse of machine learning AI in all forms and what can be done to encourage building platforms that enhance what already exists rather than completely replace? 

[EasterlyArt]

I feel like with the recent email about Adobe Illustrator image generator, this just seems to be getting worse. 

Sure, it will be super easy for anyone with a subscription to adobe to bring an idea to life, but this may cause more damage for artists in the long run and cheapen the overall value of art is it's just being auto-generated rather than created. 

Auto-genterated vs Created by Humans 

This is be the line in the sand moving forward with a lot of the creative community, and I don't feel that Adobe is doing a good job in advocate for artists first. They're simply using them as stepping stones to make it easy for any inhumane business owner to cast aside those who need to be able to pay their bills in favor of something that will allow them to cheaply create an "orange astronaut cat" and sell t-shirts or promote their next big campaign. 

The notion of whether Adobe is doing things ethically seems pointless to debate if the rush for AI generated imagery is taking the forefront. Just feels like we're on the fast track for Idiocracy rather than innovation. 

Thanks Adobe... 

[EasterlyArt]

As a followup to all of this, @CMass, will Adobe pull any imagery from BeHance? 

Also, how do we know what platforms Adobe uses for "openly licensed content"? Do they check to see if those are not using copyrighted or lisenced works? 

Again, AI is moving WAY too fast with not enough legislation in place to protect artists. It's really scary to think companies could be snagging up as much material to claim as "original" right now for their businesses, but built off the hard work of amazing creative people who are now being cheapened for the benefit of those in power. 

I'm all for moving technology forward, but not if it means people will lose the chance to make a living, pay bills, and have money for food. My livelihood depends on my ability to create, and I'd hate to think someone sleezy manager type will convince their boss into using AI to do everything for them.