Show HN: Practice Job Interview with ChatGPT API

[marcelc63]

Hi HN!

ChatGPT API was just released and I'm eager to hack something with it.

So I created PracticeInterview.co to help people practice for their next job interview.

ChatGPT acts as the interviewer and you can choose to interview for various roles such as Finance Manager, Social Media Manager, and Software Engineer.

I think this is a cool implementation of ChatGPT and the interview sessions so far has been pretty coherent.

Please give it a try! It's free, no logins required, and your interview session is saved so you can comeback to it later.

Since it's an MVP, I would welcome all feedback. Thanks HN!

[metadat]

I clicked "Backend developer".

> You: I'm ready

> Interviewer: Great! Here's your question: Can you explain the difference between a stack and a queue data structure?

> You: Definitely.

> Interviewer: Okay, great! A stack and a queue are both abstract data types that can be used to store and manipulate collections of data. However, they differ in their approach to managing their elements. A stack is a Last-In-First-Out (LIFO) data structure that allows elements to be added and removed from the top of the stack. The topmost element in the stack is also known as the "head" of the stack. When an element is added to the stack, it becomes the new head. When an element is removed from the stack, the head is removed first. On the other hand, a queue is a First-In-First-Out (FIFO) data structure that allows elements to be added to the back of the queue and removed from the front of the queue. The front of the queue is also known as the "head" of the queue, while the back of the queue is known as the "tail". When an element is added to the queue, it becomes the new tail. When an element is removed from the queue, the head is removed first. In summary, the main difference between a stack and a queue is the order in which elements are added and removed. A stack follows the LIFO order, while a queue follows the FIFO order.

> You: Cool

And then it just keeps prattling on about it, without ever actually testing my knowledge.

And even if I start off saying "No, not really", it spits out the answer just the same.

How does this help me prepare? (:

In real life, the interviewer probes the depths of your knowledge.

I do like the UI, though. Nice job there, it's easy and intuitive.

Edit: @marcelc63: No need to apologize. I had fun trying it out and wish you all the best. Don't give up! Feel free to ping me when you're ready, metadat.luvs.u@big-gmail

[marcelc63]

Hi! Thanks for trying it out.

Yes I do admit the interview is not perfect yet. More prompt engineering to optimize. I think with a few more optimization we can get ChatGPT to conduct a more quality interview session.

Sorry you didn't find value in the mvp yet, my teammate and I was taking ChatGPT API for a spin and share it as it. In hindsight we could've done more optimization before posting here.

If it's ok, I'll let you know after we optimize the prompt for you to try it again.

Thanks for the feedback, we appreciate it a lot

[halvo]

To be fair, “definitely” would not be a good answer to that question in a real interview :)

[metadat]

It's a direct answer to a fairly broad and vague qualifying question, intended to be a response to kick off a dialog and proceed to discuss what the real question is.

Interviews aren't and shouldn't be about regurgitating rote memorization.

[gmanis]

This is quite nice for practice even if you know its ChatGPT. I tried for a product manager role https://practiceinterview.co/interview/80?key=a3f9cef7-5bbf-... and it was quite an interesting chat.

[marcelc63]

Wow, reading your interview session feels like a knowledge base in itself. Thanks for sharing!

Perhaps these mock interview can be a resource in itself later on?

We're looking to further optimize the prompt and definitely to cross check the facts that ChatGPT spit out (since its notorious to state false information as fact).

Hope you enjoy the MVP

[gmanis]

Yeah, these interviews can be a knowledge base. You could have links to past interviews by category for people to read through. But do give people an option if they want their interviews publicly listed or not. Great for practice and clarity of thought.

Edit: Just read my own chat again. I did the "interview" quite hurriedly. I can't believe how far we have come with these tools. I think a lay person who is not aware of this technology would not be able to discern that the interviewer is not a human and that's saying something.

[marcelc63]

Thats a great idea. I'm thinking there's a lot we can do with the idea. Especially through this layoffs season, might be a valuable tool for many.

Re your edit: I know right! I've built a few thing with GPT-3 before but the ChatGPT API somehow gives a totally different result. Especially in regards to tone.

Slap Whisper API to this, you can practice with your voice and not only via typing.

[james-revisoai]

you could link the knowledge base to ways to answer questions with context or create a fun gamified way to study for interviews, like leetcode but for general interviews

we are on the mission of AI-generated lessons for most content at revision.ai if you are interested in hearing more - you could upload that interview session as a PDF easily

[derwiki]

Pretty neat!

I tried the "fullstack" and it was good enough to get me to type real responses. It was a little frustrating when it asked for ways to improve database performance, and then told me my answer was "a good start" but that I neglected many non-database performance aspects. https://practiceinterview.co/interview/352?key=b1bcabb4-3293...

[marcelc63]

Yes, for now the prompt we seeded was giving context to ChatGPT about their role as an interviewer and area of expertise.

Something we want to try next is to seed several real interview questions, perhaps it'll lead to a more quality interview.

Thanks for trying the MVP, we appreciate the feedback

[james-revisoai]

I think it's great, surprisingly good at separating what I enter to return a response.

however, it has the mindset of an exam marker, not an interview. Let me elaborate - I listed 4 things as possible ways to optimise a machine learning model

The feedback for "How would you optimise a machine learning model" is "You: optimise the parameters; look at bias needs; can it be put into production? run test runs on small subset of the data first

Interviewer:

Your answer is a good start, but it could be more specific. Here are some tips to improve your answer: - When you say "optimize the parameters", make sure to clarify which parameters you mean. Are you referring to hyperparameters or model parameters? How would you go about tuning them? - Regarding bias, it's important to check for both bias and variance. How would you assess whether your model is suffering from high variance or bias? What steps would you take to address these issues? - When you say "put into production", this implies that you have a specific deployment strategy in mind. Can you elaborate on what that strategy might look like? How would you ensure that your model is scalable, reliable, and maintainable? - Running tests on a small subset of the data is a good idea, but how would you choose which subset to use? What metrics would you use to evaluate your model's performance? How would you know when you've achieved the desired level of performance?"

The last point is important, the point on subsets is the kind of question I would ask to test train/test/validation set and cross validation knowledge without prompting those terms.

But the first points are not necessarily a good idea. If a candidate went into so much detail on one point, they're not hitting the mark on succinctly and clearly communicating how they'd optimise a model. That should be for follow up questions.

That said, despite that the response is not interview like (let's say, an interview would just follow up with one question, not many), if I was warming up my skills for a job this would be a godsend, because it mentions all kinds of things that would catch me out in an interview... not sure I would actually learn them though, more learn the appearance of knowing them, which is an ironic reflection on ChatGPT itself...

[whoisobed]

Thanks for the feedback! This is great insight! We will modify it to ask probing questions incrementally :)

[lrpe]

Took two inputs for my session to devolve into the thing answering its own question and asking me if I would like to know more.

Then I asked it if I got the job and got the "As an AI language model bla bla".

It was at least entertaining.

[whoisobed]

Thanks for the feedback! We are actively modifying it to ask probing questions only and only answer clarifying statements to asked questions

[efangs]

fun tool, but tbh I found it frustrating right out of the gate. It felt like being stuck in an infinite loop. https://practiceinterview.co/interview/330?key=36a0893b-6a2b...

[marcelc63]

Thanks for trying it out.

Yes, currently it needs a severe promot optimization to increase the quality of the interview session.

Our idea is to seed several real interview questions, perhaps this will improve the quality of the interview.

If it's ok, I'll let you know after we make the update for you to give it another try.

Thank you for the feedback