When I first started my B.S. degree in Computer Science, I would have never imagined the capabilities that ChatGPT provides being right at your fingertips. Yet now, it seems like the first thing a programmer might do on a new project or technology is ask ChatGPT – “Assume you are a software engineer. Write me a small ‘Hello World’ script using XYZ.”
At this point in my educational career, I’ve tried my best to steer away from ChatGPT, mainly due to the plagiarism or originality complications that can result. Yet I might have been missing out – while ChatGPT can try and answer your homework questions, one of the most pivotal features it provides is a creativity and brainstorming companion.
ChatGPT seems to work best when you define strict limitations, or the scope of your request is small. My general rule for using ChatGPT is ‘Trust, but Verify’.
For instance, I gave it a small prompt as follows: Tell a short funny story about 3 friends. Anytime you write “ra”, “re”, “ri”, “ro” or “ru”, replace it with “reh”.
The response: Once upon a time, three pals—Trehvor, Stehve, and Gehorge—decided to go on a whimsical adventu-reh…
I’m not fully sure what happened, although it converted Steve to Stehve, and George to Gehorge.
Imagine a similar mistake happening but you ask it for a simple bash script to aid your project – the ramifications could be dire. And hence, my ‘Trust, but Verify’.
In this capstone project, I’ve taken that statement to heart but tried to increase my usage of ChatGPT to aid in more menial tasks, which has turned out quite successful. In this project, I went in knowing almost no Java, along with the libraries Spring and jOOQ that would help build the foundation of our project management website. I’ve been able to consistently use ChatGPT as a valuable starting point for these libraries. From there, I usually take a trek on Google using the results I get, but it usually points me in the right direction, even if its suggestions are wrong.
Additionally, ChatGPT has been incredibly useful in mocking data for our database. While we do envision generating an official source of mock data in use for our website, ChatGPT can fill in the gaps, and all it needs is our database schema. I’ve found this incredibly useful in reducing the time it takes to spool up mock data in SQL, which can sometimes be frustrating due to a typo or inappropriate/invalid value.
There are times when ChatGPT is incredibly confident and forceful in an incorrect answer, which I believe they refer to as ‘hallucinations’. But for simple tasks, it can reduce the initial time it takes to begin. Remember, always be sure to Trust, but Verify.
