Generative AI tools such as Midjourney, DALL·E and Stable Diffusion have the potential to () the creative class as much .
How tools like Midjourney might impact the creative class
My opinion is that tools like Midjourney, even as they exist today, will make it more challenging for illustrators and graphic artists to survive. Successful illustrators will likely continue to be in demand, but illustrators who are just getting started or were already struggling to make a living might decide to give up on illustration as a profession and change careers. And that’s something that the creators of these tools should take into consideration.
Where illustrators might’ve used quick pencil sketches to consider different directions on an assignment, they could also use Midjourney to explore different ideas and directions during the sketching rough draft phase, before spending the time to produce a final illustration.
Many people trained as print/graphic designers in the ’90s had to redefine themselves as UI/UX/visual designers in the ’00s, or find entirely new career paths. The same thing will likely happen to professional illustrators and graphic artists. They will need to adapt to survive in a world where anybody with a $10 subscription can generate any imagery imaginable in seconds. Considering how quickly generative AI is improving (see Midjourney’s timeline from v1 to v4), it feels like another significant market shift is underway.
What the creative class can do in response
Illustrators could invest further in original styles and techniques that AI can’t produce (it’s comically bad at drawing hands, for example, and text). And perhaps by working together in solidarity to demand that these AI tools not allow the use of their names as keywords in image prompts or their work being used in training data. From what I’ve seen, using a specific artist’s name - be it a painter, filmmaker, or illustrator - will cause the AI to faithfully apply that artists style to the image. If that approach is taken, it may only temporarily work - it seems like simply using these AI tools creates ever more training data.
Midjourney could take steps to address the valid concerns of artists by creating revenue sharing agreements ala Spotify, or allowing artists to blacklist keywords. That feels a bit like blackmail though.
Finally, the regulatory environment could catch up. It’s very difficult for a human to draw the line around what’s a derivative artwork by an artist, but perhaps an AI could be trained to do so? For example, an artist could “own” any AI-derived artwork if it’s found to be 90% derived from their style, and be owed royalties on use of that image, or expect payment for it as they would for original commissions.