ChatGPT blew up in my first year of grad school. Here’s how it’s going for me.
My first experience with ChatGPT was in November of 2022. My boyfriend and I were playing around with Stable Diffusion, making weird images of Joshua trees and people holding bees, and we decided to try using ChatGPT to come up with prompts. At the time, generative AI was sort of a “game” for us. Him and I competed to see who could produce the weirdest image possible — the winner got to make their image the other’s Discord profile picture.
I think January was when ChatGPT became more than a game for me: it was in the news, my professor mentioned it in class, and I started really to try to understand what it could (and couldn’t) do. It seemed potentially disruptive to the academic norm, but it didn’t worry me yet. I think this is because I hadn’t used it too much at that point.
February was the month of playing around with it. At the time, I was (impossibly) trying to learn both R and Python for different projects I was working on and mostly used it as a learning tool. It temporarily veiled the difficulty of programming, making me feel superficially brilliant… God-like. In my class on urban infrastructure, one of our writing assignments required us to use it to answer critical thinking questions and to reflect on its use. I’d helped design the assignment earlier in the month, but by the time I completed it for myself, I was blown away with how much ChatGPT had improved in a matter of weeks. Inspired by all of this, plus some discussion with my classmates about their frustrations with no clear “rules” on when we can and can’t use it (especially in the context of group projects!), I decided to begin creating an adaptive management plan for my university department.
I think March is where things became serious. I interviewed a few professors about their thoughts on ChatGPT and had long informal conversations with my boyfriend, brother, coworkers, and anyone else who would listen on everything from consciousness to ethics to business opportunities that could come from ChatGPT’s use. Everyone I spoke to was pretty excited about it. We shared use cases, experiences, articles, as well as tips, hopes, and dreams.
By April, it felt like ChatGPT had consumed every aspect of my life. Here’s what an average Wednesday looked like: I started my day by writing a proposal for summer researched focused on large language model (LLM) applications for decision-making in infrastructure. I then attended a webinar that my university hosted on various aspects of generative AI and academia. I spent most of the afternoon sending emails to members of a discussion group about LLMs that convened for the first time the day previously and reading AI 2041. A significant chunk of my nightly call with my boyfriend that day focused on how human bias plays into LLM training, after which I went to what was supposed to be a game night with my friends but quickly turned into discussing and testing out how well ChatGPT understood our own research questions and what it might mean for our dissertations. I wish I could say that I then went home and had ChatGPT write me a bedtime story, but thankfully, my nighttime routine remained sacred enough.
By the end of the semester, I was a ChatGPT Plus user. I’d spent hours playing around with GPT-3, 3.5 and 4, trying to force the models to seem to believe that I had seen a dinosaur, or have them write me a new recipe, or come up with ideas for how to conduct literature reviews in different ways. I’d finished and distributed my adaptive management plan to various professors and my department head. I’d been invited to participate in a discussion group surrounding all things LLMs in academia, where we talked use cases, ethics, successes, and frustrations with its use and regulation (or lack-of) both within and outside of academia.
It’s now the end of May. I’ve procrastinated writing this, partly because I spent a lot of my “free time” this month helping a professor test a course to teach students how to effectively use ChatGPT, and partly because I didn’t want to cave and use ChatGPT to help me speed up this reflective task. While ChatGPT has consumed my life over the past few months, it’s been interesting to see how has or has not altered the lives of those around me. I remember a peer of mine shying away from using it when I suggested he use it to troubleshoot a programming error in R that I wasn’t familiar with, saying he wasn’t familiar with “ChatGBD” and didn’t know how to use it. On the other hand, my grandfather has been helping me stay up to date on the most interesting ChatGPT related articles ever since I showed interest in it. This has been immensely helpful, since there’s a lot of articles (like this one!) to plow through nowadays. It feels like everybody is writing about it, and everything, faster and more coherently than ever — probably thanks to these LLMs. I’m anxious to see what changes LLMs may catalyze in academia, and I hope this is not the only thing I write on this topic.
Until next time,
