Making Exercise Nine: Generation

AI generated text and images are platforms that have great potential for meaningful critical making projects—especially at their current stage of imperfection. Projects using these tools will likely tend more toward the genre of glitch art than fully polished projects, but then again, we have valued process over outcome throughout these patterns. Quirks and emergence can contribute to the overall work, and we embrace these much like Marshall McLuhan’s famous cover typo of “massage” instead of “message.” As you engage in generation, consider:

• Probe the conversational limitations. Try starting vague, or working with different ways of phrasing and expressing ideas to see where meaning is conveyed and where it is lost. These conversations are in part an opportunity to see what is included in the underlying archives (training data)—can you tell what might be missing?
• Explore a range of expressive outputs. Machine learning models are particularly powerful for assistance in programming and development work, as tools like GitHub Copilot suggest: in time, it’s likely these tools will be an inescapable part of certain workflows. Think about where the AI assists your making, and where it is limiting.
• Contextualize the experience. How does the system you are co-creating with reflect the headlines, fears, and assumptions that both generations of science fiction writers and current journalists have centered in their visions of AI futures?

The work of generation is in many ways a conversation, and it is that conversation you’ll be experimenting with and documenting this week. A few examples of conversational approaches one might take to an AI generated tool include:

• Display the results of a conversation or “interview” with the AI about your research topic, asking it to answer using the perspective of a specific person, theory, or ideology. Evaluate the results and reflect on their implications.
• Ask the AI text or image generator to create/refute misinformation about your research topic. Evaluate the results and reflect on their implications.
• Use an AI art generator to create a visual or artistic representation of aspect of research that you conducted. Evaluate the results and reflect on their implications.
• Try co-creating a website or digital art project with the AI, developing code and deploying it to test the results. Evaluate the results and reflect on their implications—which aspects benefitted from AI? Which aspects were impeded by AI?
• Use both text and art generators back and forth to generate a representation—or to garble a representation—of a research topic or theory. Evaluate the results and reflect on their implications.
• Generate another version of a work you created using a different digital tool by prompting an AI tool of your choice. Compare and contrast the output with your work, and reflect on the implications.

We will be trying some specific methods for using generative AI as a coding assistant while engaging in critical making over the next few weeks, but this week, the goal is to build familiarity and understanding while critiquing the experience. Keep in mind the readings in both Design Justice and Your Computer is On Fire, and make specific connections to the problems raised by these tools using those lenses. I recommend using your UCF credentials and Microsoft CoPilot for this project: just log in with your NID at CoPilot.