, , ,

Scrolling through social media, you may have noticed that your followers page has begun to look like a list of cartoon characters or Renaissance portrait subjects. This shift is not because your friends and influencers on social media have gone back 500 years in time to sit for portraits by Leonardo Da Vinci, but rather because of AI art generators.

The most popular is Lensa AI, an app and subscription service that generates “Magic Avatars,” or AI-generated images of its users. The process is simple: Users pay a $3.99 fee for the one-week trial, then upload selfies and choose a gender in the app. After half an hour or so, the app will spit out 50 avatars in their likeness, which users can use for their social media profiles or other purposes.

On the back end, however, there’s a lot going on when you upload your selfies to Lensa. The app uses Stable Diffusion, an AI image generator that returns images based on user inputs of either images or text descriptors. Stable Diffusion was trained on “image-text pairs from a broad internet scrape,” according to its owner, Stability AI. That is, Stable Diffusion determines associations between text prompts and images by analyzing images on the internet. When a user feeds the engine an image or text prompt, it collects all its prior associations to generate a new image based on them.

This process can be novel and lead to interesting results when used by the general public, but on a broader scale, it is already majorly shaking up the marketing industry. Lensa and many other AI art generators allow the use of their images for commercial purposes, which means that marketers or businesses may be relying on them to create brand images, logos, or entire marketing campaigns, reflecting key terms mined from user interests. For example, DALL-E, another AI art generator, allows users to repurpose the generated art for “any legal purpose.”

Such emerging tools could provide a major boon for marketing companies. The AI art engines like Stable Diffusion that are trained on data sets of artists’ work, then create artwork at a low cost and for relatively little labor. In some cases, just a quick search on an online engine is all that is required. But artists contend that they may own the images Stable Diffusion is generating, because when an AI engine creates an image after being trained on a particular artist’s work, that artist should own the rights to the new image. In many cases, the artists used to train the AI have not consented to the use of their art in this way, nor are they credited on the database. As a result, many businesses are holding off on using AI- generated images for commercial purposes, in anticipation of an upcoming intellectual property and ownership battle.

Still, the use of other AI-generated content is growing. Marketing firms are beginning to use AI text generators more and more; AI-generated text is composed more efficiently than human writing and can be more effective for search engine optimization. Similar to the image generators, AI text generators scrape the internet for text, like Wikipedia entries, blog posts, and books, and use the patterns to generate new information in response to text prompts. Similarly, researchers are beginning to find that AI code generators, trained on public GitHub repositories, can outperform human programmers in certain conditions.

As this technology is still emerging, its precise uses by companies and the role that humans will play is yet to be seen.

Discussion Questions:

  1. How might AI-generated images affect artists and graphic designers?
  2. Who do you believe owns an AI-generated image?

Sources: Madison Kircher and Callie Holtermann, “How is Everyone Making Those A.I. Selfies?” The New York Times, December 7, 2022; Kelsey Weekman, “People Have Raised Serious Concerns About the AI Art App That’s All Over Your Instagram Feed,” BuzzFeed News, December 8, 2022; Serra Ikiz, “Many Artists Question the Ethics of AI Art. Here is Why!” Parametric Architecture, December 8, 2022; Matthew Hutson, “AI Learns to Write Computer Code in ‘Stunning Advance’,” Science, December 8, 2022; Forbes Expert Panel, “13 Intriguing Ways Agencies Will Be Leveraging AI in 2023 and Beyond,” Forbes, January 3, 2023; Luke Hurst, “AI Writing Is Here, and It’s Worryingly Good. Can Writers and Academia Adapt?,” euronews.next; August 11, 2022; Andrew Hutchinson, “Can You Use AI-Generated Art in Your Digital Marketing and Content Efforts?” SocialMediaToday, September 21, 2022; “Stable Diffusion Public Release,” stability.ai, https://stability.ai/blog/stable-diffusion-public-release.