AI as Art the Importance of a Human Touch

Generative AI tools like Midjourney and DALL-E have made AI image generation accessible for free by anyone with an account. This has resulted in an estimated 34 million pieces of AI-generated content being produced every day. This widespread accessibility has prompted fears of vanishing jobs and, unfortunately, given AI-generated art a reputation that isn't exactly warranted – cliched, biased, and effortless to create.

At Pinkston, we believe AI-generated content can be a legitimate tool for producing art – provided it has the human touch or clearly identifies when it doesn’t.

First, we believe there’s a difference between an AI image hastily constructed with a few word prompts; and an image created with a thoughtfully constructed word prompt, that is then subjected to thoughtful, intentional, and often meticulous human edits. Even the act of compiling a prompt is an exercise of human creativity. There’s also the science behind prompts – “prompt engineering” – understanding what makes a prompt functionally more effective.

Experimental designer Dev Valladares, for example, uses AI to enhance his capabilities as an artist. He used a program called Pixel Art Diffusion for this image in The Baffler. Pixel Art Diffusion creates highly detailed images constructed from millions of pixels and allows designers to adjust various parameters of the algorithm to create their own unique artworks.

https://valladares.dev/the-baffler

Second, AI can be a promising design solution for clients who envision artistic scenarios that don’t exist in the physical world, or who don’t have the timeline for large-scale campaigns. Design Army was one of the early adopters of AI-as-Art. In early 2023, DC-based Georgetown Opticians enlisted them to create a campaign for its new store opening (to put the timeline in perspective, ChatGPT had only been launched a few months earlier, on November 30, 2022).

They had a one-month deadline – far short of the six months required to design, scout, shoot, and edit an in-person campaign. So the studio used AI to create surrealist images and videos that combined past and future and called it “Adventures in A-Eye”. The result was featured in PRINT Magazine, AdAge, AdWeek, and more. Photo manipulation, animation, sound design, and typography were all done by the human design team.

Stock photographers are facing an evolving landscape where generative AI offers clients fast, customizable image solutions, potentially threatening traditional jobs. Despite these pressures, companies like Shutterstock and Getty Images continue to balance AI adoption with a robust demand for authentic imagery, which underscores the enduring value of human creativity. This shift highlights the importance of transparency—not just in disclosing AI use, but also in exploring how AI and human creativity might coexist to meet new and changing market demands. The unresolved tension between traditional photography and synthetic alternatives underscores the critical role of the human touch in defining the future of art and imagery.

1. Consumers are “turning against ‘the AI look,’” according to The Wall Street Journal – distinctions like flawless skin and shiny colors. Plus, AI has gotten so good that even AI tools that detect AI-generated content can be tricked. This introduces a significant challenge as generative data can suffer from issues like data poisoning and model corruption. Such corruption degrades the quality and reliability of AI outputs. As the utility of generative AI evolves, it must be trained on authentic content, not merely on synthetically generated or manipulated datasets. 

2. Transparency matters. This doesn’t necessarily mean marking the final product “AI-generated”, but a client should always understand how AI was involved in the creation of their content.

3. Intellectual property matters. Last year, Getty Images sued Stability AI, a company that creates images based on text queries for using photographs from Getty’s catalog, without permission, to build its software. While users can’t know the details of how every algorithm was created, we don’t stand behind art that is “inspired by” or “influenced by” other artists’ work.