AI-Generated Images: Generative AI Concerns & Opportunities for Marketers

Ready or not, we’re rapidly heading into a world where generative AI tools like ChatGPT, Dall-E, and others do more of marketers’ day-to-day work. That transition is not much in question. The only question is how well your organization makes that transition.

To make that transition as smoothly as possible, we recommend taking a cautious and realistic approach toward the here-and-now of these technologies, while keeping an optimistic long-term view. That starts with having a solid understanding of how these technologies work, including what they’re capable of and what they’re not.

Having that understanding will help you deploy generative AI in ways that will help your business grow, and either avoid or safeguard against situations where AI leads to costly errors or embarrassing customer experiences. 

Because the issues and use cases around generative AI vary depending on whether you’re outputting text, code, or images, we’re addressing each of those separately. In this post, we’re focusing on generative AI for images. Let’s start by explaining…

How AI-Generated Images Work

Generative AI for images works entirely differently than generative AI for text. For images, generative AI engines like Dall-E and Midjourney work by denoising images that are originally composed of completely random pixels. Through many iterations of denoising, those pixels are changed to increasingly match the requested image description. 

These image-based engines are trained on alt-tagged images available on the internet. It’s the alt tag description that allows the engine to accurately create images based on the description in your prompt.

However, with images, there’s an extra step in the training. In addition to the original image, these engines are trained on multiple variations of each image, with each variation having progressively more noise introduced to it. It’s this training that allows these engines to incrementally create images from an image that’s 100% noise.

Weaknesses & Safeguards for Use

When creating images using a generative AI engine, it’s critical to be aware of several major weaknesses, so you can protect yourself against them.

1. Potential for copyright infringement & plagiarism. The very fact that generative AI engines are trained on a huge number of copyright-protected images from the internet has already led to lawsuits being filed against Stability AI, Midjourney, DeviantArt, and Prisma Labs. More lawsuits will surely come. And while the focus is on technology providers at the moment, it’s inevitable that graphic artistics, photographers, and other copyright holders will eventually sue deep-pocketed users of these tools for copyright infringement as well.

To safeguard against this weakness, don’t prompt generative AI to create images…

• In the style of artists whose works are still under copyright protection. As a reminder, in the US, for works created after January 1, 1978, copyright protection lasts for the life of the author plus an additional 70 years.
• Of particular people or that look similar to specific people—especially celebrities and other public figures—as individuals control the rights to their identity, including their image and likeness.
• That are similar to the style of logos, designs, and other visual elements that are owned by corporations or organizations.

It’s likely that user prompts will be discoverable from generative AI providers in the event of lawsuits against users, so including any of the above in your prompts will be damning evidence for prosecutors. 

Legal risks can be further reduced by having generative AI modify existing images for which you hold the rights. And risks can be minimized, for example, by asking for the removal of elements (e.g., a person) from such an image or changing backgrounds in generic ways (e.g., adding trees).

2. Potential of leaking intellectual property. It’s not always clear what happens to images that are uploaded to generative AI engines. Sometimes they become part of the AI’s knowledge base, which means they inform the answers to other users’ prompts.

To safeguard against this weakness, you shouldn’t upload or input any proprietary information or trade secrets into these AI engines.

3. The quality can be poor & time-savings unimpressive. Currently, generative AI isn’t anywhere close to being able to replace a talented graphic designer for a couple of reasons. 

First, to fully utilize these tools, a background in creative composition or art direction is helpful to get the result you want. Second, even with solid prompts, most AI-generated images require additional manipulation by a designer. Because of that, in most cases, a talented designer can create images much more quickly than generative AI can. For example, our designers find it much quicker to compose photorealistic images in Photoshop than by using generative AI, because of the post-generation editing that’s required.

To safeguard against this weakness, first, learn how to write prompts that will get you the results that you want. Here are some elements you’ll want to consider including in your prompts:

• Photography terms, such as the lens (e.g., fisheye), film speed (e.g., ISO 400, ISO 100), and shutter speed (e.g., 1/125 second, 4 second)
• Art terms, such as style (e.g., realistic, impressionistic) and medium (e.g., watercolor, pastels)
• Specific colors, such as Pantone color numbers

And second, ensure you have solid designers available, either internally or externally through an agency or freelancers.

4. Images can be off-brand. One of the biggest reasons why generative AI isn’t great at creating images for marketing purposes is that most brands have a very particular style when it comes to photography, illustrations, and other graphic elements. It’s a key way that they communicate their brand voice. Currently, most brands will likely find it quite challenging to have generative AI consistently generate images that are on brand for them, and that also have the right layout and position. 

To safeguard against this weakness, consider focusing generative AI on modifying existing images that are on brand and have the desired layout. For example, generative AI could be used to remove certain elements from a photo.

Best Marketing Use Cases

Given all of those weaknesses, how can brands wisely use AI-generated images in ways that play to their strengths? We see several clear use cases.

Brainstorming. Although not as helpful for brainstorming as generative AI is for text, it can still be worthwhile. For example, you can ask for UI concepts for a website and get decent results that can get you thinking about the possibilities and how they might apply to your brand.

Sometimes, you might even get lucky and have generative AI create an image that’s 75% of what you’re looking for. Then, you could embark on some editing to get exactly what you need.

Upscaling. Do you wish one of your existing images was a higher resolution? AI upscalers can bring low-resolution images up to 4K. This is one of the generative AI tools that our designers use routinely via Adobe Photoshop and Lightroom.

Creating textures. It can take artists days to composite a three-dimensional texture that can be used when modeling. Using generative AI to create the components of the 3D texture is a great example of a designer still having large control over the outcome while eliminating a painstaking process.

For example, generative AI can be promoted to create a “micro lens close up of a wood grain with a knot in it.” That’s way faster and cheaper than the alternative, which is to scour the internet for the perfect wood grain picture, take a closeup picture yourself, or buy a pricey pack of the different layers.

Modifying and creating backgrounds. Generative AI is quite good at generating backgrounds, whether it’s city streets or a forest. It can create a lot of variations in a short amount of time so you can choose the best option.

In all of those cases, generative AI is playing only a role in the overall design process. This keeps brands in better control of their brand image, while gaining some time-savings—and also avoiding legal risks. We’re years away from generative AI being used by brands to create fully composited images that require little to no human adjustment.

The Future of AI-Generated Images

Despite all of our skepticism, we’re optimistic about generative AI long term. We’re just in the messy early adolescence of its development.

Given the hefty investments that are being made, we expect dramatic improvements in the years ahead. We also expect copyright issues to be resolved. Progress on those two fronts will open the door to generative AI becoming a part of designers’ day-to-day workflows.

In the meantime, now is the time to experiment with these new technologies—both conservatively for work projects and more adventurously for fun. At the same time, make use of non-generative forms of AI for design, such as screenshot-to-HTML conversion tools and UI color palette generators.

Start slow and keep in mind that change management is hard. However, also remember that the modern office has changed dramatically many times in the last 50 years, each time creating new and interesting roles and opportunities. Now is the time to explore and start to reimagine your workflows and processes, always with the goal of improving the customer experience.

(Editor’s note: None of this post was written by AI. Not a single word.)

—————

Need help exploring how generative AI and other AI tools can improve your digital marketing? Oracle Marketing Consulting has more than 500 of the leading marketing minds ready to help you to achieve more with the leading marketing cloud, including Analytic & Strategic Services and Creative Services teams that can help you use AI tools wisely and safely.

Talk to your Oracle account manager, visit us online, or reach out to us at CXMconsulting_ww@Oracle.com.