Job interview. Tell me, how can you tell when the person on the other end is sitting with ChatGPT? And I don’t know, maybe a friend types up the questions, he sees the answers and tries to use them on the fly without really understanding them. I can’t say for sure that this was the case with the dude I recently interviewed, but it looked very similar.
During the process, I figured out that you can partly combat this by asking questions that don’t have a simple answer, and through rapid-fire questions which ChatGPT can quickly respond to, but if the person isn’t familiar with the topic, a quick LLM response won’t really help them, and there’s no time to read a long response. Meaning, ChatGPT will still answer correctly, but to use ChatGPT’s response, the person needs to carefully read through the whole block of text in the answer. And conversely, you shouldn’t ask questions that ChatGPT can immediately answer.
But generally, it’s quite a task to come up with such questions.
For example, instead of asking “How does SOAP differ from REST”, it’s better to ask “for which task might REST API not be the best choice”. ChatGPT starts giving a smart, bulleted answer, which you can’t simply read off the screen without understanding it.
But I’m really curious, what have HR departments come up with to combat tips from LLMs? After all, you could quite well train an LLM to display something you are generally familiar with, but recalling it yourself would take longer and with mistakes.
At the end of the interview, I requested a Live Coding – shuffle the array, where I planned to change the task setting after receiving the first version of the code. Well, you get it, there wasn’t even a first version beyond the shuffle() method header.
Kira Kuzmenko
Hi, I'm Rauf Aliev. 