Ethical AI Practices: Collaborating with Visual Artists

In light of the seemingly rampant exploitation of artists by AI companies, a number of partners have been asking about our stance on AI and it’s impact on visual artists. We wanted to share some lessons learned, but first share how we learned those lessons. Back in 2018 we developed a project for what, at the time, was cutting edge machine learning technology, image style transfer. While quaint by today’s AI standards, this precedent provides a more inclusive and mutually beneficial approach to creative collaborations.

Contributing artist Laura Wooten stylized in her painting aesthetic

The experience was an interactive pop-up art gallery, made from deployable custom fabricated easels and a bespoke AI model. Visitors were invited to have their photo stylized (by a technique called neural style transfer) into the aesthetic of four amazing local artists — Laura Wooten, Theodore Taylor, Shannon Wright, and Brandon Robertson. This model was trained with each artist’s consent and they were directly compensated for participating. Additionally, the choice to engage and support local-working artists provided an opportunity for them to engage in dialogue over the application of their work, directly and universally.

Some things we did well...

  • Transparent Conversations: Clearly communicated project objectives and terms of use with each artist to ensure they understood the scope of the experience, intricacies of the machine learning process, and how their work was going to be leveraged. Experience Designer’s have an ethical duty to educate and empower artists on the implications (good & bad) of working with technology, this was especially true in 2018 when AI was relatively nascent.

  • Equitable Compensation: “It’s great exposure” is NEVER a form of payment!!! We aligned on fair compensation with the artists right-sized to the use, duration and distribution of the model. *Consider additional licensing for their creative input and expertise as an experience scales.

  • Frequent Attribution: This is a non negotiable, we had a digital plaque that credited the artist and the original artwork anytime the live style transfer was visible. Share the stage and promote artists work!

Some things we could have done better...

  • Collaboration: Tight timelines required us to ask the artists for a subset of preexisting artwork to train the model. In an ideal world, we’d work with them over time to curate (and potentially produce) a body of work that accurately represents their style, fine-tuning it through artist-led iterative feedback on the output generated by the model.

As AI and Art inevitably become more entangled, we have to consciously and ethically navigate, adapt, and change the way we work together and the policies of major tech companies to protect visual artists.

Previous
Previous

What Is Immersive Experience Design?

Next
Next

Something is Missing in XR… People