Generative AI can be a powerful tool in the hands of a creator. It can expand creativity by working as a partner in brainstorming, ideation, and execution, and it has the ability to help creators produce their best work quicker.
But since the technology climbed into public consciousness, creators have, understandably, had their guard up. Uncertainty abounds about its potential impact on creators’ livelihoods. There’s been little transparency around how companies are training their AI — and whose content they’re using.
Adobe has set out to solve for this uncertainty with Adobe Firefly, its family of creative generative AI models built to bring out the best in human creativity — while granting creators the transparency and choice they deserve. Designed to be safe for commercial use, Adobe aims to make Firefly the most creator-friendly generative AI solution on the market, delivering on the promise of generative AI responsibly.
Generative AI that creators can trust
A creative identity is an amalgamation of many things — the creator’s vision, their experiences, their personality, the content they’ve consumed, and so much more. It’s a unique style and expression that’s honed over years of practice.
A concern with generative AI is whether it will make it easier for bad actors to mimic a creator’s style without credit. So, Adobe is taking seriously its commitment to constructing a better environment for creators. The company only trains Firefly on content it has the permission and rights to use, such as licensed material from Adobe Stock and public domain content — never customer content. Unlike many other companies, Adobe doesn’t train Firefly with content mined from the web without permission.
In addition, Firefly is designed so that it does not generate content that infringes on copyright or intellectual property rights. Adobe deploys safeguards at each step — prior to training, during generation, when the engine is prompted, and when the engine creates an output.
When contributors add their work to Stock, they do so with transparency that it may be used to train Firefly. And Adobe has provided bonus payments to Stock Contributors whose work was used in training Firefly.
Putting human creativity on super drive
Creators know that their boundless imagination sometimes outpaces the time available to bring their ideas to life. Integrated seamlessly in popular Adobe apps that creators know and love like Adobe Photoshop, Lightroom, Illustrator, Premier Pro, and Adobe Express, Firefly aims to close that gap with generative AI features that speed up and expand the creative process.
For instance, creators can make quick edits or create variations of their existing content — while maintaining brand standards. In Photoshop, the Firefly-powered Generative Expand feature helps creators broaden the frame of a photograph, while Generative Extend in Premiere Pro can add frames to videos for a smoother overall edit. In Lightroom, Generative Remove allows photographers — from hobbyists to pros — to easily and quickly remove unwanted objects and distractions even on complex backgrounds. In Adobe Express, creators can use Firefly-powered Generate Image and Text Effects to generate unique videos or create standout social posts and stories.
These tools all help maintain the integrity of the creative process while freeing creators from tedious and time-consuming tasks so that they can efficiently summon their best work. Firefly generative AI models are powerful and, critically in the age of AI, developed responsibly.
Protecting creators and driving trust
Adobe’s efforts, however, don’t stop at the level of its own AI. The company is an advocate for responsible innovation and creator rights across industries, even within legislation.
Adobe is a founding member of the Content Authenticity Initiative, a massive collaboration of some of the most recognizable names in media, technology, and civil society that aims to restore trust and transparency in the digital ecosystem. In an age in which deepfakes can deceive even the trained eye, the Initiative’s Content Credentials — which act as a “nutrition label” for digital content — have emerged as the global standard for content provenance, providing a peek into the content’s origin and creation process. The tamper-evident metadata can be attached to digital content to show information like who made it and how it was created — including whether AI played a role. This information not only gives creators recognition for their work, but also protects them from anyone looking to repurpose their work without giving proper credit.
Content Credentials are already supported in Adobe apps like Photoshop, Lightroom, and Adobe Express. They’re also gaining significant momentum across the industry, with implementation spreading to, among other companies, Google, Meta, LinkedIn, TikTok, and OpenAI.
Beyond technical solutions, Adobe believes that there’s also work to do at the policy level. With some generative AI, it takes just a few clicks for a bad actor to impersonate a specific creator in the marketplace. It’s crucial that we institute regulations and protections to help guard creators’ rights. Adobe is advocating for a federal anti-impersonation right that can help protect creators’ style and work in the age of AI.
By putting creators at the center, Adobe is setting an example that the entire industry can follow. To learn more about Adobe’s approach to generative AI with Firefly, visit Adobe’s website.
The post How AI can empower creators while respecting their rights appeared first on Vox.
