With a flood of AI generated content reaching our social feeds and inboxes, it’s more important than ever that we know who, or what, is behind what we see, watch and hear.
However, there isn’t a reliable way to find out who created that cat meme or validate a photo of the recent floods, nor is it easy for creators to assign ownership to what they make.
And while creators can ‘burn in’ their credentials to photos when using Adobe’s Photoshop and Lightroom, and videos made with Premiere Pro coming soon, a wider adoption is needed.
It’s prefect timing then, that at AdobeMAX 2024 conference in Miami, Adobe opened its credentials system up to anyone who wants to tag ownership of their photos, images, videos and music. Also, users don’t need to be paying for Adobe’s creative tools or have a Creative Cloud subscription to add credentials, so it’s free to all.
To assign a to an artwork file, creators simply need to access an online form and attach their name, identity and social media handles. The attribution will create ‘nutrition label’ style report, not just about who made the content, it will also indicate if generative AI has been used.
New ways to protect your ownership
There are also two new concepts coming to content attributions: the first is the ability for creators to tell compass that they don’t want their artwork to be used to train generative AI. This means that Adobe won’t train new Firefly generative AI models on anything but artwork that they have been given permission to use.
It’s still very much the ‘Wild West’ out there, however, as many AI companies have trained their models on artists work without their permission, and may not pay attention to the Content Attribution Initiative. Hopefully change in time.
The second concept will help prevent content credentials from being removed from existing creations. Adobe’s Andy Parsons, Senior Director of the Content Authenticity Initiative, says that: “one of the things I’m most excited about is something we call durable content credentials.”
And as an information consumer, I want nothing more than to be able to see content credentials everywhere.
Andy Parsons, the Senior Director of the Content Authenticity Initiative at Adobe
This uses open source technology to prevent copying of original works or deleting the credentials. Parson’s explains: “…what happens if I just take a screenshot, or I take a photo of a high resolution photo and pretend that it’s a photograph, won’t the metadata be destroyed? The answer has been, in general, yes, but with durable content credentials, we combine some concepts around watermarking and fingerprinting with the cryptographic certainty of a content credential.
Protecting past work
An existing problem is that many creators have 100s of gigabytes of legacy data. Thankfully, Adobe has a free app that lets you apply content credentials, en mass, at the push of a button.
In addition, creators can use the app to tell companies that they don’t want their legacy work to be used to train generative AI systems.
Bad actors are still out there
Despite applying content credentials, there are still going to be bad actors who anonymously create AI-generated misinformation, or pass off existing artwork as their own.
Thankfully, the easier it is for creators to apply attribution to their works, the move obvious it will become when people don’t. We may also see social platforms requiring attribution for things to be posted, as well as government legislation to protect the integrity of public content.
Parson says that: “So it’s not really our concern to catch bad actors, and instead to empower good actors to say, you know, this is authentic, and I am motivated to prove that it’s authentic.”
Ultimately Adobe’s Content Authenticity Initiative will continue to build trust, as it shows how the assets were created, and by who.
Valens Quinn attended the AdobeMAX conference in Miami as a guest of Adobe Australia
The post The AI wild west: How Adobe is protecting us from content misuse appeared first on GadgetGuy.
0 comments:
Post a Comment