This is the crux of it for me. I won’t get into the artistic, ethical and environmental impacts here because that’s a huge topic of its own.
I’ve used AI tools for projects in the past with mixed results. I’m not trying to be some arbiter or gate keeper of the “Right way” to do things but in my experience it’s best used at the very beginning stage of something to get jumping off points into subjects I might have less knowledge on. A fancy interactive version of JGL (just google it.)
Then I have a good list of subjects to read real technical documentation on, and search for first hand sources I trust. That last step is being lost from what I’ve seen.
I work in an RnD engineering field and being able to quickly spin up some stupid file management powershell script has genuinely saved time and effort towards very specific needs. I’ve learned enough about powershell and python to know how to steer it towards what I want and it’s faster than me having to remember the syntax and specifics each time as someone who doesn’t need to interface with code often. I know enough to skim the structure and make sure it’s doing what I want for these simple applications.
I don’t want to make this a generational thing but I’ve seen a few younger coworkers who are actual software engineers go to AI as the first step to solving a complex problem before they even understand it
Without going into specifics, I interface with a very specific and bespoke piece of industrial testing equipment. It was having some issues and I’m generally a good trouble shooter from my years of breaking shit on my own.
Software is not my wheelhouse or responsibility but I was there with a pen and pad, writing down the architecture of the setup as I learned about it and poked at the edges. Narrowing down the problem to one specific region of software and even getting it to a point I could reproduce the issue reliably, which nobody had attempted before I got there.
That software guy I mentioned said “he talked to chat about it and it said his code is fine, so the problem must be mechanical.”
He hadn’t even attempted to understand the structure of the process as a whole before brining his small snippet of code to AI for a pat on the back. The problem was of course something in software that I ultimately guided him towards with real poking and prodding the conditions of the system in meat space.
I lost a lot of faith in the future that day.
Those guys were always around but I feel like there was a stronger filter when you had to at least partially know how to walk the walk.
