unnamed. Credit: DALL·E 3 x Tor

As we, humans, are increasingly collaborating with AI, new questions about creative patterns arise. A student of human and AI collaboration and an art buff in my spare time, I wanted to use DALL-E 3 to realize an idea for a photo. Here’s what the human and DALL-E 3 collaboration was like:

1. Human: think of how to communicate your point of view (POV) and values.
2. Prompt DALL-E 3 with some context supporting that POV. E.g. How can you help being born in one place of the world and not in another? How can you possibly not imagine being in the shoes of other people who by coincidence were born in another region of the world?
3. Prompt DALL-E 3 with your goal and instructions to create. E.g. Based on this POV, create a photo of a group of human beings standing in the waterfront by the Opera House, waiting to get approval and access to enter the country. The people express hope, uncertainty, confusion, vulnerability, humility. Leverage contrast between the people and place — between where they are and where they want to go.
4. Human: Evaluate the outcome of DALL-E 3 and come up with a new prompt to close the gap between the actual AI outcome and your idea of what that outcome should be — your goal or intention. E.g. Add some more expression of belonging to the faces — the people are lost and want to belong if they are allowed.
5. Prompt DALL-E 3 with modifications.
6. Human: Repeat until you have the desired outcome.

As you might notice, the core idea here is feedback: taking action on the gap between AI’s outcome and your intent. Feedback is the mechanism that brings humans and AI together as both parties are trying to get their work done. Feedback is the intervention point of humans when they want to improve AI. And AI relies on that feedback to evolve.