My AI Compromise

8-19-26

I've had conflicting feelings about AI from day one, probably like a lot of people.

On the one hand, there's no denying its utility. I mentioned this in my last blog post. I distinctly remember the first time I tried an AI coding assistant and how mad I felt that it was good. After all the time I'd put into learning to code the "classical" way, how dare this machine be better than me at writing code!

On the other hand, I don't love the negative impacts that AI is having on society and the planet. Part of me is holding out hope that those impacts will lessen over time, but that may just be a cope. I guess time will tell.

Last month, I had all but decided that I was taking myself out of the AI game. I've had the attitude since the beginning that I should know how to do things myself, and that AI was robbing me of important lessons. But an acquaintance of mine, someone who's a lot more knowledgeable, suggested a compromise that I've been testing out over the past few weeks, and I've been surprised to find that it's pretty much resolved that inner conflict.

It was a suggestion so simple that I wish I had come up with it myself: let the AI write the code, and as I review it, ask it to explain anything I don't understand. By doing so, I can work at AI-powered speeds, but still fulfill my desire to understand what it is I'm making.

My current project involves learning Databricks, and this new workflow has proven to be exactly what I needed. I started out the project determined to do everything myself, but I quickly realized that I wasn't really wrapping my head around the conecpts necessary to create a good data warehouse. As I've introduced AI into the project, I've learned a lot about SQL, data management, and the medallion structure. It's been a bit like drinking from a firehose, but I'm doing my best to keep up. Being a late adopter of AI, I have a lot of ground to make up. But better something than nothing, and better late than never.

I remain skeptical of most applications of AI beyond writing code, and that probably won't change. I don't intend to ever let AI generate a blog post for me, and I still thing AI "art" is a tragedy (this is coming from an art school graduate, so no surprise there).

Now as I'm wrapping up my MS in Data Science, where (understandably) I was not allowed to use AI for homework assignments, I'm curious to see how it helps me develop my data science acumen. I still have so much to learn, and I'm a little scared to let AI review my old projects for improvements, but I can only get better from here.