The dilemma of using ChatGPT in the interview process

There are three ways I've used ChatGPT while interviewing for Senior/Staff iOS Developer roles at various companies:

1. Writing the job description of my current role at Secret Atomics
2. Writing the branding statement and skills assessment of my marketing plan for LHH
3. Writing the unit tests for a take home technical assignment

In the case of 1), the assigment was literally to take 5 job descriptions and boil them down into something that sounded like the job I wanted to have. This is literally what the resume branding advisor told me to do, so what better way to do that than to ask the machine to do it first. ChatGPT did not produce a concise result, so it was left to me to take a long description and make a short one that sounded more like the experience and background that I bring to the table. IMO, this is fair game for resume writing, and even though the description sounds a bit all over the place, it works for now.

In the case of 2), all I really did was ask ChatGPT to take my resume and turn it into a branding statement. Again, probably better than doing it myself, since it has the experience of millions (?) of english teachers and marketing professors built in, and is better at branding than I claim to be. For the skills assessment, I just asked it to give me a list of the skills for the job roles that I have done in the past at the level I want to be. For good measure, I asked it to give me the skills at a particular company (Apple/Google). I also asked it to give me a culture statement for Apple in a single sentence, and what it gave me works just fine. IMO, again this is fair game. It's unclear to me why I need to repeat what is already known to everyone who is a recruiter in my marketing plan, but I suppose its just to pick out the stuff that I identify with. ChatGPT was a good resource for coming up with a bag of words to choose from as fast as possible.

The third case is a bit more sticky. My take home test asked for unit tests. I haven't written a single unit test in 5 years as a UX Engineer. I've had various hiring managers tell me they are both important as well as a waste of money and time. Each company decides what to do with unit tests according to their time, budget, and willingness to do other types of QA to mitigate the risk. My manager at Homer Learning tried to instill the benefits of unit tests on us, and she was right! But they are also a pain to write, and they sometimes can double the development time of a single feature.

So in this case, I first submitted my take home test without any unit tests. The guidance from the recruiter was to take half a day to write the entire test, and I got through 90% of the requirements in that time. Rather than "cheat", (according to hacker rank you aren't supposed to use other people's code), I simply left them out. However, it bothered me that I didn't finish in time, and I did want to see what those unit tests would look like if I wrote them. I turned to ChatGPT again, gave it the functions that I wrote myself, and asked it to write my unit tests. It did not give me a complete answer, but it did give me enough to start with to get through the testing in less than an hour. I don't know for sure, but I think writing them myself would have taken 3-4x as long. And in this case, I would advocate that everyone everywhere who is writing code should run their code through ChatGPT or some other protected LLM to get unit tests out. It solves the dilemma of taking the extra time. It does not solve the dilemma of using them in a take home interview test, but that is something I left up to the recruiters to decide. The day after I submitted my deficient repo, I sent another copy with the unit tests and a disclaimer of what I used. It's up to the company to decide if I was smart or if I cheated.

For those who are interested, here are some links to show the outcomes: NASASearchAPI Unit Tests

Ultimately, I would say that ChatGPT is a tool that is useful the way that StackOverflow is, or an encyclopedia, or lets say it's a StackOverflow for everything in the encyclopedia. There's some garbage, for sure, and it's very chatty... but common sense and the ability to edit it down to somethign that makes sense makes it worthwhile. One major downside is obviously the data collection OpenAI is doing, and I understand why Apple and some governments have banned its use. However, I would say that open source models will eventually solve the data vacuum problem, and until then I'm happy to live in a closed ecosystem.