I recently spent some time with an LLM doing some brainstorming. Some people have observed that our current crop of AI models don’t so much create as they copy and remix. Let me just start out by saying that I’m not great at everything, and if there’s a way for me to ask the, I dunno, few hundred million humans who are better than I am for some advice, then that’s good, right? Well, maybe.
The problem I was attacking was, “What should I name this new app I’m working on?” I’m terrible at coming up with names. It’s just not what I spend time on, and so I’ve not developed the skills or discipline for thinking about the problem. So I fired up Ollama and Chital and gave it a whirl. I described the problem I wanted to solve and the application I wrote to solve it, and asked for some ideas.
I treated the conversation as an improv exercise. Rather than saying, “No,” to particular suggestions, I said, “Yes, but,” and, “What if…” to steer the suggestions around to things that I found more attractive or interesting. The suggestions themselves were not things I would have come up with by myself, and they were obviously remixes or echoes of things that have come before, but then, my problem and my solution are hardly novel: they’re just my take on a situation experienced by nearly everyone.
So that was pretty interesting. Ultimately, I arrived at a name that’s better than anything I’d come up with before, so I’m going to call that good.
Following on that process, I decided to take a stab at an application icon. I’d already mocked up something (because I hate the default “no icon” icon of in-development apps) but I know just enough about graphic design and human factors to know that, A) the icon was Not Good (and why), and B) I’m not skilled enough to come up with a better one.
So I installed Fooocus and asked my text-based LLM to come up with some prompts for Fooocus to use to generate some icon images. I asked the Ollama LLM how I could show it the generated images and then talk about them, to refine the prompt; I was surprised to discover that it could fire off HTTP requests and scan the images if I put them on a server. So that’s…interesting.
I think I ran out of patience for the process before coming up with a good icon. I might try that process again. If this application ever gets to the point where I’m going to release it, I still think I’m going to hire someone like IconFactory to design one.
And now, a few days later, I’ve not done any more LLM stuff. I stepped away from the tools and the application and just did other important things in my life, and I’ve got a bit of unease about that whole process. The LLM was just so relentlessly positive and enthusiastic. It would spit out some suggestions in response to my prompt and it would end with something like, “I’m really excited to see how this comes out!” or, “This is really great!” If a human told me those things, I’d think those expressions were signaling something about their physical state — not just that the idea or topic was of interest, but that there’d been some hormones released and their pulse was quickening, irises dilating, breathing altering, that sort of subtle response we all have to things that are really engaging. But that’s not what was going on, obviously. The AI didn’t care how long it took me to respond; it didn’t wait a minute for me to reply and then, when I didn’t, keep thinking about the situation and come up with more ideas, or think critically about what it had already said. It wasn’t actually engaged in the discussion. So the constant interjections of, “That’s a great idea,” and, “I’m really excited by,” or, “That’s a really interesting point,” were just lies. They were meaningless (to the AI) strings of tokens. And that’s all that the rest of it was, too.
So. The stuff that’s good? I think it’s accidental, and for the most part it arose because I was thinking critically and creatively and using the tool to guide my thoughts. But the effect the tool had on me? I’m not at all comfortable with it. It feels a lot like what I imagine it would be like to be surrounded by people who, no matter what bullshit crazy idiocy I spouted, told me I was brilliant. It’s never a good idea to surround yourself with sycophants.