Adobe embraces generative AI, but at what cost to creatives?

This is an editorial from Photofocus publisher Richard Harrington.

Many of the images used in this article are screen captures from Adobe’s keynote address at their MAX conference. You can watch the original presentation here to see Adobe’s presentation on Generative AI.


The reality of generative AI has created quite a controversy among photographers. In this case, AI creates images using verbal prompts and other input sources. For some, this is seen as an opportunity; for others, it can create many new problems.

What’s the problem with generative AI?

For example, we’re already seeing significant issues related to generative AI and copyright. Designers and artists who use generative AI to create content may end up with works that cannot be copyrighted or fully protected. This can also cause business issues for the companies that use their services.

At the Adobe MAX keynote, Adobe has come down on the side that generative AI is a good thing. They affirm that it can help with many things. Adobe included several examples in their keynote.

Adobe shares key statistics about the pressures creatives face. The need for fresh content and time pressures are reasons given to support the use of generative AI.
  • Moving shadows in a photo
  • Changing the time of day with computer-generated skies and lighting
  • Creating new designs and content based on verbal prompts
  • Changing seasons from spring to winter in a photo
  • Creating new photos with people in them, people created with AI

You can watch the keynote here.

Why generative AI makes me uncomfortable

As I looked at these examples, I cringed. I understand the need to enhance a photo, the pressure to replace a sky, etc. But this newly generated content takes the work of others and makes new things without paying the creators, locations or models who serve as sources.

Photoshop has a Portrait Generator tool that’s in limited beta to create new models and portraits.

For transparency, I say this having worked on the product team for Luminar (I frequently felt uncomfortable with Sky Replacement, though). Currently, I work on Radiant Photo, an AI-powered image editor that can fix color and tone problems with AI. I also help the Mylio Photos team use AI to help surface and discover your most important pictures.

To be clear, I really like AI, especially when it can take someone’s work and help them improve it. I think Adobe’s AI-powered masks are fantastic, letting users enhance precisely and fix their content. Adobe clearly gets this with many of its offerings, and we are seeing companies like Radiant, Skylum, Mylio, ON1 and Capture One embracing useful AI with recent releases and announcements.

Adobe and other companies feel pressure from industry and stockholders to embrace generative AI.
The company said it is focused on doing it in a way that benefits creative professionals.

But I genuinely hate generative AI with a passion. The new examples Adobe showed were a whole new level of technical power and ethical creepiness. Watching a photo transform from spring to winter or a shot from daytime to a computer-generated sky felt lazy and untruthful.

Is this evolution, or something new?

Let’s discuss Skylum’s Luminar as an example. First, it had a Sky Enhancer powered by AI. That was awesome and accurately selects and enhances what’s there. Then, there was Sky Replacement, which lets you take a photo and match it to the scene with color grading and gentle light changes. People screamed about how awful this was. But at least it used photos made by named artists who could license their skies as stock images to others and had their names preserved in their work.

Replacing a sky and relighting a scene with Luminar. Luminar was the first tool to automate the process of sky replacement with AI and has had their work copied by other software companies.

At the MAX keynote, I see Adobe taking Sky Replacement even further, generating new skies using AI and creating all-new skies for the photo. Useful, sure. But very creepy. Even worse is Landscape Mixer in Photoshop, which lets you combine two images to completely change what’s there.

This generated AI content is powered by Adobe Sensei, which sources much of its learning from the Adobe Stock Library. Adobe showed this a few years back as they generated new content for landscape photos by drawing from the stock library to create “new” pixels. But this new content is clearly based on source photography. This made me uneasy then, and today’s latest tech made me feel even worse.

We saw this a few years ago with Photoshop’s Neural Filters. It had the ability to modify a photo and add a smile with teeth to an image that didn’t have one. Except, the AI ​​doesn’t know if that person has braces, a chipped tooth or drinks too much coffee like me.

Adobe’s vision for generative AI and content authenticity

Adobe President David Wadhwani insists, “We believe that generative AI should benefit creators, not replace them.”

Adobe President David Wadhwani addresses Adobe MAX in Los Angeles.

That statement is a good thing. Also, on the positive side, Adobe is doing some very noble work with the Content Authenticity Initiative, being able to focus on providing proper authentication and tracking changes. Adobe also announced partnerships with Leica and Nikon, who are building content authenticity into camera capture. This ensures that people can attach themselves to their intellectual property at the time of capture.

Adobe also demonstrated helpful technology to ensure their images are properly licensed. For example, it shows how tools like Illustrator and InDesign can ensure that content is clear for use with payment and license. It is clear that much of Adobe understands and values ​​the creative output and the creatives that buy its products.

The bottom line? I feel like Adobe is a company with many excellent initiatives. But it is also a company trying to figure out how to balance cool with authentic. It wants to help creative people deal with the pressures to create better and faster with an fluid sense of ownership.

As a creative, I think we need to join this discussion. I believe that generative AI is exceptionally problematic. While Adobe can take steps to ensure that things are done correctly, there are too many risks and gray areas here that can fundamentally hurt artists like photographers, as well as the people we put into our photos.

Can Adobe get this right?

When Adobe launched Creative Cloud, they had an advisory board. ThinkTAP COO Dave Moser and I served on this board for two years with 18 other well-versed creative pros. We helped ensure that the needs and goals of creatives were heard as Adobe entered the new world of cloud computing and cloud storage. Adobe listened, and I believe it got almost all things right.

What’s clear, though, is that Adobe needs a new group of artists and photographers to join the discussion here. Those who understand AI as well as copyright should be part of it. Adobe has the power to lead here. Otherwise, they will just emerge as the greatest driver of a new area of ​​technology that is becoming unethical and uncertain with the capability of doing genuine harm.

Adobe, your move. Give in to what’s “cool” and buzz-worthy, or take a stand for intellectual property and image rights.

Source link

Leave a Comment