The less I've thought about something, the easier it is for AI to give me what I "want"
I've been working on a million different software projects lately, and I've noticed that the less I actually have opinions about something, the easier it is to get AI to do what I "want." But conversely, the more I care what it actually looks like, the harder it is to get AI to align with me.
Below are iterations of designs for a kids' savings app where I want to visualize progress toward a goal. Currently the site just shows a number, and I wanted more. First, Claude decided that filling a glass was the best metaphor.
Then the glass got bubbly
Then we had a monstrosity of an AI vector art piggy bank (sorry, Andrej, I think we're still not there yet on the pelican test).
Then we got a stack of coins, which is along the right path, but $86 coins is still a bit too big a chunk for most kids to feel good about progress.
Finally, we got the design I landed on.
If I didn't care, didn't have an opinion, I could have been done days ago. Sure filling a glass makes no sense in the context of saving money, but it's visual, it shows progress, so why not? At some points during the endless cycles of design and redesign, I had to answer for myself: "Do I care enough about this to keep going, or is this good enough?" I could have stopped with the filling glass, and if I didn't care, that's exactly what I would have done.
The less I care about how something looks, though, the easier it is to just let Claude take the wheel. That's why it's so easy to make an 80-90% complete prototype, and why it's so easy to take AI's information as truth in areas you don't understand. If I don't have an opinion (or the knowledge to assess the accuracy of AI's output), I'm going to take whatever it gives me.
This is part of why AI art (and music) is so polarizing. People who don't care think "oh, that looks like something I recognize as art," and they appreciate it for its art-ishness. People who have spent years studying, practicing, and learning about art instead see it as a knock-off of what art is. To them, it is shaped like art, but it's not art. (that, plus the fact that AI art is trained and profits from the uncompensated, copyrighted public work of thousands to millions of authors and artists).
Ultimately, we all have things we care about and things we don't. We all make decisions about where to spend more time and where to say "that's good enough." This isn't new or unique to AI, just something I've noticed in my recent work.