If GPT Comes for My Job

Generative “artificial intelligence” (AI) seems to have replaced “crypto” and non-fungible tokens (NFTs) as the topic du jour. I burned out on this topic long before chatGPT launched and shifted the conversation from image generation to large language models (LLMs), so I’ll keep this brief. It surprised me today when I thought to myself, “if these large language models change software development and design the way some folks seem to think they will, that might be the change that finally gets me interested in becoming a manager.”

A number of folks have been writing about how these technologies can change the way we write software. Dave Rupert wrote about GitHub’s copilot and how it shifted their work from writing code to reviewing it. Thommas Semmler touched on AI in their great piece on the craft vs industry of web development. Robin Rendle wrote about their conflicting feelings over the hype. Just today, Ethan Marcotte wrote about how the de-skilling we’re seeing with AI is not new and we need to start thinking about what we want to do about that.

Reading all of this got me thinking: if writing software and designing user interfaces becomes more about coaching some neural network to produce software and doing some light editing on the results, then I’ll probably do something else. I’m just not excited about using it. I have no desire to use a neural network to write plugins for me (YouTube). I don’t mind writing my own tests. Some of this may just be because I’m an old fuddy-duddy. I still code in (Neo)Vim, and I sort of have some code completion set up (finally) using Language Server Protocol, but it’s not working great, and I don’t rely on it much.

All of which is just to say that I barely rely on existing tools to automate my work beyond Vim’s macros and some shell scripts, so I’m just not excited about a neural network programming for me. And if the industry does move in that direction before I retire, there’s a good chance I’ll start leaning more on my background in Human-Computer Interaction (HCI) and go back to focusing on user research.

Or maybe that shift would be the thing that got me more interested in management than being an individual contributor. That was kind of a surprising thought for me, since I’ve long said I don’t have any interest in moving into management. But maybe this is kind of the natural course for a lot of people who move into management. Maybe you reach a point where you’ve either lost interest in or become tired of keeping up with the changes in your industry, so you move down a layer.

I’ve long wondered if I’ll reach a point where I’m no longer interested in the latest trends and tools in software development, and then my skills will just slowly lose relevance. As I’m approaching my 40^th birthday¹ I can’t help but wonder if that’s what’s happening here.