Adobe Continues AI Push By Keeping a Close Eye on User Behavior

Adobe wants to watch its users a little closer so it can recommend how to properly change the brightness in Photoshop using AI.

Mikaela Shannon on Unsplash

Sign up to uncover the latest in emerging technology.

Adobe wants to watch its users a little more closely. 

The company is seeking to patent a system to generate “personalized in-application recommendations” by watching user behavior and determining their intent. Essentially, this system monitors your work in Adobe’s apps and gives you pointers using AI. 

Adobe’s system first monitors user’s “behavior signals,” which may be derived from a number of different user actions as it attempts to get a “holistic view of the in-application behavior” to understand user context. For instance, the system can monitor behavioral signals such as the skill of the user, the properties of the document (i.e. the photo, video, etc.) itself, or the edits they are trying to make to the document. 

From there, Adobe’s system would make recommendations in one of a few ways, either pulling them from a predetermined “curated list” based on previous user activity, or by feeding these behavior signals to a machine learning model. This model may also take into account historical activity. 

Finally, the user’s interface is updated to reflect those recommendations, popping up as an option in the toolbar on the side of a user’s screen. As users’ behaviors change, their recommendations change “in real time,” Adobe noted.

For example, if a user is trying to edit a photo in Photoshop, Adobe’s systems might monitor what kind of photo it is and what actions a user has taken, and recommend adjusting certain brightness or shadow settings, or to remove a background object from a photo. 

Photo via the U.S. Patent and Trademark Office.

The company has filed a number of patents to improve its systems using AI, and has sought to patent improved machine learning to identify and predict content in documents, a “visually guided” language model, and a tool that uses computer vision for “diversity auditing” (which could come with unintended consequences). 

It also lines up with Adobe’s broader AI push. The company recently launched Firefly, an AI image generator and competitor to Midjourney and DALL-E, which it claims is “designed to be safe for commercial use. ” And at the Adobe Max conference this week, the company also reportedly plans to unveil a new powerful AI-based editing tool that allows users to to alter images with no editing experience. 

Amid the AI hype, continuing to file for AI-related patents and embed AI across its suite of products could be key to holding onto dominance in the field and remaining the standard among anyone that works with a visual digital medium. This is especially poignant as Canva, its main competitor after Adobe acquired Figma, beefs up its own AI tools – and offers at least some of them for free. 

However, like any AI-based system, the bias problem is ever-present. If this has any sort of feedback mechanism, or takes in user experiences to adjust the AI model, it could replicate users’ same biases to others.