Laziness, Impatience, Hubris, and AI

I’ve been feeling more and more like I’m letting AI do my thinking for me. Starting to use Cursor to write code was the most recent development that really made it clear.

I had such a clusterfuck in my head the other night trying to make sense of what I wanted to code. I felt drained, but not drained enough to stop because I had AI to help me. I prompted Cursor with these half-thought-out prompts, and expected it to infer what I wanted to do. I’ve become so lazy, I did not even reply with what it got wrong. I tried to will it telepathically when I said things like “this is wrong” or “that’s not what I meant” in response to its mistakes, with no initial elaboration on my part. But I had enough energy to give these nudges, and so that’s what I did. I kept trying to gesture to it that I wanted something, like in Spirited Away when No-Face is trying to say “take my money” with little grunts. Take my money. Take my brain. It’s indicative of a frightening form of faith in it.

What is driving me is an addiction to my sense of agency, and the desire to have maximal agency at maximal efficiency, measured by my level of effort and productive output. I want to know that these grunts of mine are making some sort of difference downstream. Grunting into a void is not compelling. Maximizing productivity at the expense of agency is not compelling. I don’t just want to watch the show, I want to direct the show with my thought. At least that’s what my chat history with an AI would have you believe.

I’ve felt this way for a long time, increasingly uncomfortable with the extent to which I and those around me delegate cognition to something or someone else. I remember, before ChatGPT, I used to plead with people not to google things when something was on the tip of their tongue and they wanted to give up. I still similarly caution people about treating AI responses as gospel. Now I see that the biggest danger of AI is not in it telling us falsehoods, it’s in the atrophying of our skills and of our desire to exercise critical thought, patience, and memory.

In my case, this has been fueled by the perception of agency and productivity and the increasingly effective role AI has played. However, the extent to which I appear productive to myself is often an illusion; I have an initial burst of output, only to find out there is bug I am not equipped to address because I have no recall or understanding of the codebase. That recall and understanding would have been cemented by the struggle of doing it myself, had I gone through with it. AI, even with its limited context windows, frequently possesses more context than I do.

Despite this, it’s become increasingly compelling to use AI, especially as a software engineer. So far, though, I have found myself without the discipline to to intervene with what I now consider self-destructive behavior.

I’ve thought of having days where I don’t use AI. Think of how far AI has come for me to say something like that. That says something huge about my relationship to AI; I’m not thinking of having a cheat day on Saturdays, where I treat myself to AI, I’m thinking of having a meatless Monday where I allow myself a break. And I’ve been hesitant to take that break because I know that, in terms of real output, I may be hampering myself in the short term. But perhaps I’m helping myself in the long term by allowing myself to keep my faculties regularly exercised instead of being totally complicit in my own cognitive decline.

I know I must sound like opponents of the printing press, concerned that the mass availability of the book would hamper people’s ability and will to memorize information. Even though people have been complaining about technology in similar ways for millennia, I think that the concern of being able to maintain an ability to reason about the world in which one operates is well founded. Yes, delegating, whether writing things down to remember later or asking an AI to draft an email, lightens or changes one’s burden to some extent.

Relying on anything external to do the thinking for oneself can erode one’s ability to do it effectively. But to delegate or not is a choice that can be made in each moment. Being aware of when you are making these choices, and not mindlessly forming a habit that reinforces itself through learned helplessness, is critical to avoiding the trap. So that’s what I’m going to do — remain mindful of my appreciation for my craft and cognition and make one small decision at a time.