21/11/2024
Our most recent issue of interactions magazine just came out. I am very excited about a lot of the content, and especially about the featured cover article by Matt Jones, Dani Kalarikalayil Raju, Jen Pearson, Thomas Reitmaier, Simon Robinson, and Arka Majhi.
Their article entitled "Beyond ‘Slumming It’—AI and the Real Lives of Global South Communities" addresses how GenAI imagery misrepresents everyday life in troubling ways. The authors share insights from a series of workshops during early 2024 in urban, peri-urban, and rural Indian and South African communities. They share with us the sheer dismay community members feel when they see how GenAI imagery represents their homes, thriving communities full of life and culture.
Usually in the background, we wanted to share with you the multifaceted process of collaborating with the ACM art department in every issue of Interactions, and especially on this cover of interactions magazine. Here is the story of the 2024 Interactions Nov-Dec cover:
Thinking through what the article emphasizes, we worked with ACM's art department to reflect image distortion on our Interactions cover. The attached image shows the image iterations we explored, and here is the voice of our art department collaborators:
"Once the decision to go with the gestures photographers use to frame up shots without the camera, the next step was to generate the hand gesture/orientation needed. It’s commonly known that AI Generated hands suffer from odd anatomical anomalies.
The prompts for the AI image generator was a repetition of prompts for hands, all framed in slightly different ways. This took many repetitions, and much refinement. Below are some of the prompts used to generate the images that were good enough to download:
* two hands framing an imaginary composition
* two hands frame composition crop
* hands frame crop composition
* female hands forming a frame in blank for insert text or design isolated on white background.
* hands frame picture visualize crop
After exploration, the prompts we, the art department, used were fairly similar, because we wanted consistent enough results to fit the brief. Ultimately, we chose the images that were good enough (or not too odd) to be reworked in Photoshop. Photoshop helped with the merging of two sets of generated hands. The rest of the image was a collage of stock image and AI-generated picture of Dharavi, one of the locations featured in the article."
First, we, the Interactions editorial team, want to foreground and applaud the art department at ACM who work with us iteratively for every issue of interactions magazine.
Second, we want to emphasize that GenAI is not a simple prompt to content interaction. It is an iterative and collaborative interaction that involves a conversation chain of refinement and a tool-chain of different products and surfaces to come to a reasonable end result–in this instance, a magazine cover.
Third and finally, we want to thank ACM's art department for collaborating with us on researching the edges of GenAI for this issue. With their help and with the contributions of our contributing authors, we can all reaffirm that GenAI is a work in progress. This is very much a human-AI collaboration engagement in all senses, from data to prompting to fine-tuning with a complicated tool-chain to a satisfactory end result.
Critical exploration is crucial if we are going to understand how to make improvements. One of the best aspects of this exploration is to showcase how visual distortion from GenAI is so much more obvious than textual distortion. A good reminder to be critically engaged in *all* forms of GenAI produced content.
A huge THANK YOU to the ACM - Association for Computing Machinery's art department for being research partners in this work, and for being design partners in all our publications.
