On Feeding ChatGPT My Dreams

Adam here. I’ve talked on the podcast about writing down my dreams when I wake up and I have shared my dream journal. That’s risky behavior: dreams are notoriously boring.

I learned that Fellini drew his dreams each morning. I’ve envied that project, but I felt I lacked the drafting skills to do it myself.

Over the past week, I’ve started pasting one-line summaries of my old dreams into ChatGPT and asking for a dreamlike photo. AI art might be more notoriously boring than dreams—and yet I’ve combined the two.

I was hesitant to share this with anyone. But then I saw Pulp’s new video, where Jarvis Cocker experiments with AI to reflect on AI itself. I felt like I had permission to experiment.

I can’t explain why I’ve been doing this. I am fascinated by the mystery of dreams and the possibilities of their patterns. Maybe I want to find a way to do what Fellini did. Maybe the AI will inspire me. (The truth: start lousy, get better.) I even ask AI text agents what they dreamt about last night. Maybe feeding my dreams to ChatGPT is my bizarre idea of generosity.

I’m dumping my subconscious into artificial intelligence—something I’m sure I would’ve called the act of a maniac just a year ago.

Anyway, here they are. My favorite one is first. The rest: I don’t know what I think.