I’ve always been a little hesitant to use AI to help with my coding. I don’t think my avoidance has necessarily been to my benefit. Partially it’s my fear of being replaced by robots, partially it’s me being a bit of a luddite, but I never wanted to give it a chance. I didn’t want to feel like anything I did was propped up on the work of something that wasn’t me – but in a lot of ways, using something like ChatGPT isn’t that much different than just looking something up on StackOverflow, right?
Most of the time, when I go to StackOverflow to look something up, the answer I find is pretty general, only tangentially related to what I am looking for, needs to be tweaked for my specific use case, or is so condescending that I feel like an idiot for even trying. ChatGPT, on the other hand, is similarly good for finding pretty general answers or answers that need to be tweaked for my specific use case. It’s like ChatGPT but faster and with significantly less condescension. As long as you’re keeping your usage of ChatGPT (or similar) to smaller cases, I think it can be really helpful.
For my Cap Stone project this term, I was struggling to get a specific library to compile with our project. I thought I had tried everything – several hours in and I was ready to start banging my head against my keyboard with little to no progress. A group member suggest I try using ChatGPT, that it might help point me in the right direction. I told ChatGPT I wanted help parsing an error message and copy/pasted the several-paragraph error message in. The answer it gave me was pretty general, but got to the heart of the problem I was struggling with. 15 minutes after throwing the problem at ChatGPT, I had a solution and could start working on what I actually wanted to do.
It’s cases like this that I think are the best for Large Language Model (LLM) AIs – you shouldn’t be using them to write your code for you (because it will usually be bad or wrong), but you should be using them to do things that are too tedious humans to be good at. Could I learn how to figure out exactly was going wrong with my compilation issues based on the error messages? Probably. Would it be a waste of my time? Almost definitely, especially if it meant taking time away from the work I needed to be doing.
In the future, I think I’ll continue to use ChatGPT and other LLMs for similar purposes. It won’t make me a worse coder or software engineer – it will make me a more efficient software engineer. Especially for dealing with tasks that, let’s be frank, humans aren’t well suited to.
Leave a Reply