Write more to think more

Lately, I’ve been wanting to write more and more. Each day I try to create a pleasant situation of waking up early in the morning, have a few silent hours all to myself, and getting my creative juices flowing. Unfortunately, I still somehow can’t get myself to think about and then decide what to write.

Though I know there may be a few different ways to mitigate and deal with this, it got me thinking about a bigger danger that is lurking around that I find worrying.

I am talking about losing my natural creativity due to excessive LLM use.

I learned recently that I’ve been a ChatGPT user for over 1,000 days. There is no question that ChatGPT is a software I’ve used the most, of all softwares, ever. How I use it and why is unfortunately beyond the scope of this thought experiment, but the point is, if GPT was a real person, it would easily be the one I talk to the most.

9 out of 10 times when I have a challenge at work or at home that requires a bit of deep thinking before action, I’ve noticed my first instinct now is to speak/write to ChatGPT. Granted, I am more aware than to take every LLM response at face value and try to do my due diligence by cross-referencing different models’ responses to the same question, I still can’t help but see that I’m actually becoming lazy to think deeply about things.

It’s a little bit similar to how back in the day when we wanted to know about something in nature, or the world, we’d go to the library and pick up an encyclopedia or a related book. Then suddenly one day I realized that I could just google it and there would be something I’d find that was directly relevant, and definitely faster. Google completely changed the way we approached looking at the outside world. In some ways, it brought the marginal cost of ‘search’ to zero. I wonder what aspect of ourselves did we lose in that transition.

In similar ways, I think ChatGPT is bringing the marginal cost of ‘thinking’ to zero, but it seems we’re losing the ‘what to think about and how’ also with it.

Is that the price to pay?

I know there is some research beginning to come out on the cognitive costs of heavy LLM usage. I plan on spending some time making sense of what’s going on there but for now, I am trying to write more to force myself to think more.