Artificial Intelligence has been around for decades, but recent innovations alongside wide accessibility has led to a small, yet noticeable paradigm shift in the way information is sought and managed. Most notably, open-access language models such as ChatGPT have proved to be quite the pandora’s box in regard to education. To both educators and students, it can serve as either a great informational assistant, or a crutch that lets the mind grow lazy and rot.
At first, I realized ChatGPT was an excellent conversationalist. It has an impressive ability to parse information and return coherent results (which in turn, is limited by its relative “narrow sightedness”). On top of that, it could aptly handle odd and abstract questions much better than some humans could. Unfortunately, its nature as a language model–as opposed to a scientific model–means that in cases of STEM disciplines, such as chemistry or mathematics, it returns unreliable results. For this reason, most people are out of luck when it comes to relying on GPT for help with their math homework. But I had found myself in a middle ground. There were still many unanswered questions about my discipline far too niche to ask Google, but which weren’t related to homework; that is, I didn’t need the mathematical or scientific precision that GPT, at the time, lacked.
For that reason, I decided to use GPT to do what it could do best: answer generalist questions with the knowledge of a field expert. The more I asked, the more I realized that it could give exactly the results I was looking for and answering my questions in the exact way I needed them, again, arguably better than a human. I could ask it about specific degrees or for help finding a career that aligns with my interests and goals; GPT had proved itself an excellent–and passionate–advisor.
Continuing, there was one computational aspect GPT could do, even as a generalist, (somewhat) inherent to its own structure: coding! No different than how it could flawlessly translate across languages (while still keeping the proper context and meaning), it could also assemble code well. Of course, it was prone to errors, but it could get the right idea. If brute forcing Google oddly specific questions for what type of operator I needed for a specific task failed, I would reliably turn to GPT and use its amazing ability to always understand my nonsense.
I would like to conclude this essay by saying that while GPT may serve as a great assistant, it should not be used to entirely write code (or essays). As an educator (aspiring and somewhat in practice), I strongly believe that one’s mind should be outsourced as little as possible. Sure, having a “useless” assignment written for you may seem like a great idea, but once this habit becomes cemented in your routine, you’ll be left wondering why it’s so hard to come up with and express unique ideas.