Ubuntu-AI: New Collaborations from African Heritage Communities at the Intersections of Art, Design, and Artificial Intelligence

When Stamps Professors Ron Eglash and Audrey Bennett were invited by leading artificial intelligence developers to help democratize AI, they didn’t stop with the original prompt.

Building on two decades of developing computer-assisted learning tools based on ​“heritage algorithms”, and recent efforts to help artists in Africa return some of the value generated by their creations back to their own communities, Eglash and Bennett went to work on their own set of research questions for the OpenAI Foundation, the research giant behind ChatGPT.

“What they asked for was democratic deliberations on AI policy — what kinds of questions it should refuse, for example. We proposed instead that we examine how to democratize the economics of AI. Ideally we would like to see the actual ownership of the AI to go to the grassroots, to the folks who are doing most of the hard work in society. And much to their credit, they agreed to let us spend the research money on that instead.”

As a result, OpenAI awarded Eglash and Bennett a grant in 2023 to develop a platform for exploring how African artists and designers could co-develop AI tools, retain control over how their data and images are used in AI model training, and build an online community around democratic and shared ownership of the platform itself. 

Now, with additional funding from a recent Office of the Vice President for Research grant, the researchers and their team, which includes Wayne State Professor Kwame Robinson and U‑M School of Information Ph.D. candidate Micheal Nayebare, are examining how this platform might connect to their work with Detroit artisans. ​“The NSF funded our research on how digital fabrication tools might benefit small worker-owned shops in Detroit,” Eglash explained. ​“With so many of them located in Detroit’s Black communities, it seemed like a unique opportunity to explore how AI and platform-based technologies might create new collaborations between African heritage communities on both sides of the Atlantic Ocean.”

Addressing the ​“Double Bind”

With the recent boom of platforms like ChatGPT, generative AI has become a powerful and popular tool capable of producing creative work in any style or genre. But it does so without giving any credit — or compensation — to its sources. At the same time, as AI becomes a global hub for human knowledge, its algorithm can only produce work based on information it’s been fed. This situation creates what Eglash and Bennett call a ​“double bind” for Black creators who, historically, have been exploited for their intellectual property, but who also stand to be further alienated by this new technology if their works aren’t included in it.

To address this, the platform they developed, Ubuntu-AI, was designed with the principles of generative justice: those who create value should have it returned to them, without extractive exploitation or alienation. Micheal Nayebare, who is originally from Uganda, coded the platform, building on research interviews he had conducted with artisans in Detroit about how AI might benefit their own work. With the help of local project managers in Africa working with art and design collectives there, Eglash and Bennett amassed 400 contributors from Nigeria, Ghana, Uganda, South Africa, Kenya, Namibia and Cote d’Ivoire, who have registered their images and texts through the platform. Visitors can click on any image they see and negotiate a price to either buy the product or license the image for use in AI model training. And the artists have already tried out a prototype for the first of the AI tools they helped develop.

But Eglash noted that the idea of shipping products overseas was not ideal. ​“It has a huge carbon footprint, and much of your profit is lost to shipping costs. That was when we started to think about how collaborations with the Detroit fabricators might make a lot of sense: ​’designed in Africa, made in Detroit’ would be a great product slogan. And if we are creating a space for discussions between two African heritage communities, all sorts of artistic and cultural collaborations, explorations of AI, reconceptualizations of Black design ecologies and other themes might be possible.”

Making Connections

For their latest project, Bennett and Eglash will lead teams of undergraduate research assistants in Ann Arbor and artisans in Africa and Detroit to further the work started with Ubuntu-AI and their previous Artisanal Futures project, and also explore additional research questions related to flipping the script on AI to make it better serve Black artists.

Six U‑M undergraduate students working in areas including digital media, heritage arts, and fabrication techniques will work with six Detroit artisans familiar with digital fabrication technologies, including 3D printing, laser cutting, and other techniques. After initial training in AI from UMSI graduate and Wayne State University Assistant Professor Kwame Robinson, the U.S. groups will partner with African artists to develop products using neural style transfer algorithms. These algorithms will aid the teams in producing new creations that adopt designs — with credit and payment to the artists — licensed through Ubuntu-AI. The groups will also use AI to conduct research on which products might be most successful in the market. The final phase involves fabrication, and then marketing via the Artisanal Futures platform.

For one early test run, researchers used a licensed artwork from Ubuntu-AI and a photo of one of their project partners in Detroit to produce an original portrait of him in the rich, colorful style of the artist, which was based in traditional Ethiopian artwork.

From Information to Iteration

Along the way, the team hopes to explore several questions related to AI and design work. Bennett is particularly interested in the collaborative interactions between design professionals, artisans, and AI.

“What is the role of the design expert with formal training in the design process?” Bennet asked. ​“How does AI inform the collaborative design of the outcome or artifact? What impact does AI make on the collaborative design process and outcome? How are design decisions negotiated within the collaborative design process between design scholars, community artisans, and AI?”

While current AI imaging tools can produce effective and impressive work, Bennett said in her own experience, they aren’t able to replicate the same results using the same prompts or produce new iterations or versions of an image once it’s been produced. 

In addition to market research and design transfer and licensing, the team hopes to employ AI to help spark ideas about, say, what might be created from a specific set of available materials — such as in upcycling projects—or suggest a replacement material for those that are scarce or expensive. 

Covering the Last Mile

By overseeing the process — and AI’s role — from start to finish, Eglash said the team hopes to help cover what Nayebare refers to as ​“the last mile” for AI. Nayebare is writing his dissertation on how AI can benefit worker-owned production, and has already helped install digital fabrication tools in Detroit. His interviews with artisans on both sides of the Atlantic examines which aspects of their work they would like remain strictly human, which to have machines involved with, and in what ways. ​“The goal,” he said, ​“is to allow them to focus more on doing what they love, while making it more financially and environmentally sustainable. It’s great having all these images on the screen, and it’s great having all these makers, but you’ve got to connect that last mile.”