Would ChatGPT take our software engineering jobs?

A common question asked by many tech managers, designers, recruiters, and others not involved in software engineering is whether ChatGPT will take our software engineering jobs. I asked ChatGPT-4 the same question and here’s the response I received.

I distinctly recall my first experience with ChatGPT and how it amazed me with its ability to understand my questions and provide well-thought-out answers. I knew that this innovative technology was powerful and shared the same sentiment as those around me: “This is a little scary.”

Since discovering ChatGPT, I have relied less on Google and barely visited “Stack Overflow” — a website every software engineer checks daily to find answers to software bug questions. In the past ten years, I only had to ask three questions on Stack Overflow because everything else have been already addressed. Now, I hardly open the site at all.

ChatGPT can easily solve Google and Apple interview questions — problems that I might struggle to address at the time of writing this article. The technical capabilities of this technology are astounding. So, naturally, I purchased a subscription to use ChatGPT-4. It became an integral part of my daily routine, providing invaluable assistance while working. Initially, I asked for help with simple tasks like creating a Python class or a Bash script. I also inquired about straightforward matters that I would typically Google, such as how to change all file extensions in a directory from *.txt to *.md. Everything was going smoothly until it wasn’t.

I encountered difficulties with a software library I was trying to use. After delving into the documentation and source code, I was on the verge of giving up but decided to consult ChatGPT. To my surprise, it produced code using parts of the library that didn’t exist in reality — it had simply made them up. As I was already familiar with the library, I recognized the error immediately. ChatGPT has the capability to answer google interview algorithm questions, but is that what we build and deploy to production?