Our artificial intelligence innovation is reviewed and researched: Grace Yee from Adobe

Our artificial intelligence innovation is reviewed and researched: Grace Yee from Adobe

Introduction

The company realizes that it should lead in making that tone in which these AI models are developed. Or as Grace Yee, the Senior Director of Ethical Innovation (AI Ethics and Accessibility) at Adobe says "our AI ethics framework to ensure ethical considerations are central to our technology development".

In a conversation with HT, Yee goes on to detail how AI ethics have been developed with company-wide efforts involving many teams as Firefly finds deeper integration across different apps, as well as her take on any evolution of these ethics as generative AI and the larger space continues to evolve.

"Over the past five years, my team has translated our AI ethics principles from theoretical concepts into actionable guidelines that sit within our engineering and product development practices. These principles form the foundation of our approach,"

Key Insights from Grace Yee

AI is playing that kind of end kind of pivotal role across Adobe's really wide portfolio of software, meant for enterprise and creators and consumers, right from desktop to mobile use cases. The scale and the speed with which it's going with development perhaps reflected easily at the company's MAX 2024 keynote. Or, for that matter, the "over 100 new features," as they're now being called, across Photoshop, Lightroom, Express, Illustrator, and more. Beneath much of that new functionality lies the Firefly AI model suite, which includes the rather tight distraction removal generative AI features for photo editing and Generative Extend for video timeline fills and edits.

Scaling in Firefly

We at Adobe developed three principles: accountability, responsibility, and transparency.

Accountability: Ensuring we have proper mechanisms in place to receive and respond to concerns.

Responsibility : Ensuring that our AI innovation undergoes evaluation and careful diligence.

Transparency : Ensuring that our customers know how AI is integrated into our products.

The principles we developed for classic AI are the same for generative AI.

What has changed is the embedding of AI features into our wide range of products.

We use the same process and approach, just at a larger scale, understanding how AI technologies power our products.

It is important to understand our customers and how they are using these AI features.

Challenges Faced by Firefly

We started integrating our Firefly model into various Adobe applications, such as Photoshop and Illustrator. We understand the guardrails that are and need to be in place for these integrations. Different teams are working with the model, doing more with it, and we are assisting them with that innovation. Using Firefly in specific ways may not require deep evaluation because the necessary guardrails are already in place. Our focus is on the additional changes being made, which may require less intensive evaluation than building a new Firefly model.

Balancing AI ethics principles

We trained Firefly on our Adobe Stock content and public domain content for which copyrights have expired. I understand it's an incredibly small dataset compared to what others have used. But I think that's where the magic of Adobe kicks in. What our teams are able to do is really leverage some of that magic that we've baked into our previous features and really make the output amazing.

Vision on AI and ethics

One of the great things is to work with product teams we are working with help early on in the process via a risk discovery process. Sure enough, change and tweak as necessary will be made; it's an era of processes, more learning and seeing things that are going to change which we will have to make. That will depend on the kind of technologies that we will actually develop, and eventually, how we integrate it into our products and features in the future.