HOW TO USE CHATGPT IN YOUR JOB SEARCH WITH SEAN MELIS

Kip Boyle:
Hi everybody. Welcome. This is Your Cyber Path. We’re the podcast that helps you get into cybersecurity, get that first cybersecurity job, and also to help you get your next cybersecurity job. Because why? Because hey, we want to increase our careers, we want to enhance our careers, and we want that for you. We very much want that for you. So we are going to talk with you about all kinds of helpful tips, how to do job hunting, how to get that offer from that hiring manager. And we’re going to tell you that from the hiring manager’s perspective, right? Jason and I are going to give you all of our deep dark secrets for how we find and offer jobs to candidates, and we want you to use that to win. Anyway, so I’m Kip. This is Jason Dion with me. Hey, Jason.

Jason Dion:
Hey, kip. Nice to see you again.

Kip Boyle:
Yeah, it’s good to see you. And we’ve got a guest, but before we introduced Sean, I got to tell you that right now in American politics, we’ve got this debt ceiling thing that’s going on, and so everyone’s freaked out because they think, “Well, if the government defaults, the US government defaults, that’s a big deal.” And it reminds me of a couple of times when there was a government shutdown. I was actually on active duty in the Air Force when there was a government shutdown one time, and it was just weird. It was just super, super weird because some of us kept getting paid and couldn’t report to duty because only essential people could report to duty. And I don’t know, it was really bizarre and I didn’t know what to think of it. That was the only time I’ve ever been directly affected by something like this. But you were in the Navy, did that ever happen to you?

Jason Dion:
Yeah, multiple times because it’s-

Kip Boyle:
 Oh, really?

Jason Dion:
… been getting worse and worse over the years, and this keeps happening. Even the last couple of years we had a couple of those where there was a threat of the shutdown and they signed it at like 11:59 at night or whatever and prevented the shutdown. But because they weren’t sure leading up to that shutdown, we had to start sending everybody home. We had to prepare for the fact that only essential people could come in, and that was like 70% of our staff wasn’t allowed to come into work. And then depending on which contract you were on, some could come in and some couldn’t. And whether you were a government civilian or a military or a contractor, it was a real mess. And it always really aggravated me to be quite honest, not just as an employee of the US government, but also as a taxpayer because the amount of time and money we spent preparing for this pending government shutdown that may have happened was a lot of time and effort.

And then ultimately it didn’t happen because they signed the paper in time. I’ve also been affected where they’ve had it, where they didn’t sign it in time and we were told, you’re still going to come into work and we’ll figure it out later. And really what I’ve seen is a lot of times when they do this, they say, “Oh, you’re not a central stay home.” And then two weeks later they bring them back in, they go, “Oh, you know what? We’re going to pay for the two weeks you’re home anyway because it’s not your fault we shut down.”

 And so we still spent the money, but yet these people sat at home for two weeks not getting paid, and it was a big disruption of their family when all we had to do was sign a piece of paper. But again, Congress sometimes doesn’t do the right thing and they try to hold out. And to me, I just think it’s a political football that we as Americans shouldn’t be playing with, but we continue to bring this up time and time again. And it seems like every year, every two years, this is coming up. And honestly, as a taxpayer, I’m just frustrated and wish that politicians could get their act together.

Kip Boyle:
I am too. And so it was interesting to hear your perspective. It doesn’t sound like it changed all that much from what I experienced. I was in the military a little bit before you, but what I know, we have a lot of people who listen to this show from around the world. It’s not just the United States. And so I just thought it would be interesting for people to hear, “Yeah, Americans think this sucks too. It’s not just you guys.”

But on that note, let’s talk to our guests. So it’s Sean Melis. Now, Sean’s here from Australia, and speaking of people from outside the United States, and I am so excited to have Sean here because I took his Udemy course recently because I wanted to know how am I supposed to prompt ChatGPT to get the most out of the time that I spend with my over eager digital intern, and I love the course. And so I reached out to him and I said, man, we’ve got to have you on the podcast because I just think you’re very good at describing what’s going on, and you actually have a background in artificial intelligence. And so you’re not just some guy trying to make money off of the latest trend. You actually have been living this. And Jason, I know you’ve been using ChatGPT more than me, right?

Jason Dion:
Yes. Yeah, right. I’ve been using it a lot. And yeah, I mean as we bring on Sean here, I just want to give a quick introduction to who Sean is so people get a background for him. Sean is somebody we met as Kip said through his Udemy course. And Kip bought the course just because he thought it looked good, and then as he took it, he is like, “Hey Jason, I think we should bring Sean on.” And so we reached out and we talked with Sean. And Sean is a seasoned developer. He’s a seasoned multimodal designer with 15 years of experience. He’s skilled in lots of different professional areas including social media, marketing, web design, video production, community management, technology consulting, and of course AI and conversational AI like ChatGPT. He’s the founder of a successful startup known as Bot Hello, which you can find at bothello.io.

And he’s developed different chatbots for renowned brands like Maybelline New York, Lancome, Anytime Fitness and Mercedes-Benz, as well as for Australia’s largest music festivals, health food brands, and AFL teams. Prior to his entrepreneurial journey, he used to work as an AI consultant over at Deloitte, which is a huge consulting company. And many of our listeners would love to work at Deloitte one day, I’m sure, because they do a large cybersecurity presence as well. And Sean is really a dedicated lifelong learner, and he loves exploring AI and emerging technologies. And that’s why he started playing with ChatGPT when it first came out, got really good with it and then made his course on it. And so we invited Sean to connect with us here and discuss AI and ChatGPT, and that’s what we’re going to be doing here today. So I want to welcome Sean Melis to the show. Sean, welcome.

Sean Melis:
Yeah, thank you so much for having me. Great to be on.

Kip Boyle:
Did we miss anything in your bio? You wanted to mention?

Sean Melis:
You pretty much nailed it. I ran the business for, or have been running the business for five years now. Just the caveat that I’ll make is the first three years, the business was a chatbot development platform. So we were helping companies onboard their own clients, like companies like agencies and people who had other clients and helping them build chatbots for themselves. But we pivoted the business into more of an agency slash consultancy because of some of the challenges during COVID and some of the updates to Facebook Messenger and the API. So we’re now more consultancy and we work with other leading SaaS products, customer support platforms like Intercoms, Zendesk, HubSpot and so on, and leverage their conversational AI tools, automation suites and implement them on behalf of our clients.

Jason Dion:
That’s awesome. And I know we got introduced to you through your course, which is on Udemy. So for those who are interested, you can find it on Udemy. It’s called ChatGPT 101, Supercharge Your Work and Life. And if you want to find the direct link to that, the easiest way is to go to Sean’s website or his LinkedIn profile, excuse me, linkedin.com/in/seanmelis, which is spelled S-E-A-N-M-E-L-I-S. And we’ll have that in the show notes as well. If you click on that link, you’ll get it at the lowest price and you’ll be able to take the course. But before we do that, we definitely want to talk a little bit about ChatGPT and how you can use this in your job hunting search and how you can use it on the job. So let’s start out by, Sean, can you tell us, if you’re explaining ChatGPT to somebody who knows nothing about it and it’s a layperson, how would you explain ChatGPT, what it does and what it can be used for?

Sean Melis:
Yeah, sure thing. So I think most people have used a chat bot on a website. The thing that pops up in the bottom, you can send it a message often responds with something that isn’t particularly helpful is how that chatbot has been trained is through keywords. It’s looking at keywords that are relevant in the question and then will respond with an answer in a library of answers have been pre-written and they candor responses every single time. So they’re templated responses. They don’t evolve, they don’t change, they don’t augment based on the conversation and they rarely remember context. If you ask a chat bot the same question and after four or five responses, it’s not going to recall that you’ve asked that question or provide a context related to that question. It’s just going to respond again with its order response kind of like an if then statement. It’s all rule based.

The difference with ChatGPT is it’s leveraging a very large training set. Instead of these canned responses and templates, it’s canned responses is the entirety of the internet. It’s got all sorts of training corpuses related to Wikipedia, blog posts websites. It uses something called the Common Crawl, which is a publicly available database of information. So as much data as OpenAI, which is the company who created ChatGPT could feed this model, it is leveraging that information to generate its answers. So when you ask it a question, it’s going to that database trying to understand if there’s anything in that database that is relevant based on input off the question. And it determines what the relevancy is through something called attention where it’s looking at keywords within that input or the prompt or the message that you send to ChatGPT and kind of pairing that with which reference points in the database is going to be most relevant.

So it’s really a chatbot that has a lot more context, a lot more information, a lot more knowledge to be able to help you with any topic, any question, any use case. And that’s essentially talking to some kind of super intelligence. It’s as if there’s this alien intelligence that’s arrived and we all have access to its knowledge and it has access to everyone else’s knowledge throughout the entirety of human history. And that is phenomenal thing to be able to engage with. And the use cases are coming out in droves every single day as to how it can help personal and professional life. But the easiest way I like to explain it is it’s a super intelligent conversational tool, just like a chatbot you’ve used on a website.

Jason Dion:
Yeah, I think that’s a really good point. And a lot of people I’ve seen when they first use ChatGPT, they think it’s Google and they go and then they go, what is two plus two or what is the capital of America, or something like that. And it’s like that is such a slim piece of what it can do.

Sean Melis:
Almost insulting.

Jason Dion:
It can answer that stuff, but it’s like it’s really, really not a really good use of it, right?

Sean Melis:
Yeah.

Jason Dion:
And the reason why it’s called ChatGPT is because the G stands for generative, it can create new things, not just creating things based on its own knowledge. It will actually go a little bit further and you could do things like I’ve seen people using it to write poetry and make a song and make a rap and all this stuff that is what we normally consider creative work by humans. And it can do this. And you could say, “Take Star Wars and rewrite it in Shakespearean [inaudible],” and it can do that for you. It has that ability, but sometimes the answers you get kind of don’t make a lot of sense.

And I’ve noticed this, one of my friends calls it ChatGPT is a probabilistic word salad. And what he means by that is that it just kind of… Sometimes you get some random stuff and it makes its best guess, but it is after all a computer and it doesn’t actually know the intonation and what you’re asking underneath. Have you come across a lot of that where you get some answers and you’re just like, what the heck is that? That was even close to what I was asking.

Sean Melis:
Yeah, absolutely. The hallucinations are a common sort of problem, particularly for enterprise being able to trust that information and the accuracy of it. As you mentioned, it doesn’t understand the context, it doesn’t connect two different ideas together. It is simply a very advanced auto-completion tool, just like the auto complete on your iPhone. It’s using the same underlying algorithm and technology called a transformer where it’s predicting the next word in each sequence and then it’s actually predicting the next word based on the previous words in addition to input, which is crazy to think about because it generates pages of responses to be able to predict all of those different words. But across that process there’s got to be some holes in the way in which it’s referencing its database and putting it all together. I’m sure that’ll be improved over time. Yeah, it’s just about the size of the training set and also something called RLHF, reinforced learning human feedback.

So it’s the plus, sorry, the thumbs up and the thumbs down in all the responses of ChatGPT that OpenAI is encouraging the users to rate based on how helpful that answer was. And that was quite a controversial implementation of that particular tool because it means that the users themselves are steering the AI’s sort of intelligence and knowledge and contextual understanding of information. And that could be bad if people are thumbs up something that shouldn’t be thumbs upped. So that’s how kind of they’re getting away with, sorry, not getting away, but solving the hallucination problem by getting the feedback from the users to say, yeah, this wasn’t helpful.

Jason Dion:
And I think a good point on that is if you’re using this in your job and I use ChatGPT a lot in my current job, I would recommend that you use it as a starting point and not a finished product. And what I mean by that is if I’m writing a script for a new course, like Kip and I just finished doing a course for Udemy called Security Essentials, and I use ChatGPT to write a lot of the scripts I used. But that being said, that is not the final thing you saw in the course or heard from me in the course because I use that as my rough draft. And then I went in and I verified it for accuracy, I verified, and in my case, I already know the information is accurate because this is security essentials and I’m a security expert, so it’s easier for me to do that.

But if I was asking it to do something in a field that I don’t know about, like medicine or chemistry or biology, I would need to get an expert in there to verify it’s right or I’m going to have to Google it all and make sure it’s all right. Because what I have found is that, as you said, Sean, that the hallucinations happen where it sounds a hundred percent confident, but it really doesn’t know if it’s right or not. And it will say things that I know for a fact are dead wrong and I have to go in there and change those. And so if you’re using this in your work product, definitely take it with a grain of salt. It’s very useful and really good at summarizing data, but when it creates new stuff, there is a lot of room for errors.

Kip Boyle:
That’s why I call it my over eager intern, my overconfident intern. At this point anyway, I know it’s going to get better, right, Sean?

Jason Dion:
Yep.

Sean Melis:
Yeah, absolutely.

Kip Boyle:
Yeah, it’s going to get better.

Sean Melis:
Yeah, you go.

Kip Boyle:
Another thing that I think about just to call back to what we were talking about a moment ago, but I want to ask you Sean, if this is reasonable way for me to think about it. Since ChatGPT is algorithmically driven and it’s doing sort of a probability analysis on which word it should show me next. I mean I just think of it as it’s just statistically driven. It’s just using advanced statistics to decide how it’s going to talk back to me. And the fact that it does it as fast as it does is that’s what blows my mind is that this is ones and zeros and it’s multitasking because having chats with millions of people at the same time about millions of different subjects, is this a reasonable way to think about it?

Sean Melis:
Yeah, it is. But I think for even today’s AI researchers and scientists, they’re still mind blown at the speed, and it’s doing this across millions of users simultaneously all day every day with very little issues regarding latency or speed. It’s always fast there a couple of moments where it needs to be refreshed or restarted, but as a whole, the user experience has been phenomenal for a product that was only released a couple of months ago.

Jason Dion:
Yeah, I think that’s really interesting. The other thing that I thought was really cool about this whole chat thing is the fact that they made it in… I mean we’ve been using AI for years. I mean it’s been at least 10 years that we’ve had AI out there doing a lot of this stuff in the background. This is one of the first real AI tools that people got to have their hands on and interact with directly. And the fact that they made it into this chat interface allowed it to continue on with the conversation. So when I think about Google, I can Google what is the capital of America and then if I ask what is the capital of Australia and what is the capital of another country, I can do all that individually each time. But Google doesn’t remember what I asked before. With ChatGPT, I can give it a prompt, that’s what we call it, prompt engineering as Sean’s going to talk about.

And I could say, “Tell me the capital of the following country: United States,” enter, it would give me the answer, “Washington DC.”

“Australia,” okay, here’s the answer.

“UK,” here’s the answer. And it will remember things in that chat thread from earlier on just as I was talking with Kip for an hour. And then I go back to something we talked about an hour ago at the beginning of the podcast. He’ll remember that. Well, ChatGPT does too. And so you can ask it to do the same operations over and over. You can ask it to use the same methodology. You could say, “Hey, here’s the way I like to write practice exam questions.” If Jason writes him a certain way, “Now write me three new ones in Jason’s style.” And it can do that because it remembers what you gave it. And that’s one of the other things I think is really cool with this ChatGPT thing and the way that they use this as a chat interface and this conversational AI as opposed to just being a search engine or a database recovery tool.

Sean Melis:
Yeah, I think that was the real holy crap moment for people was the context that it remembered everything you said and you come back to it and continue to have this conversation in such a human-like way. And that certainly was, for me, context is very challenging in traditional chat bots, remembering all of the inputs. You use something called entities and tents and it’s a very manual process for collating those and capturing those in the conversation itself. But this is almost like it’s a real person remembering context and then you can build on top of that and almost have this domain expert living within that chat thread forever, which is super cool.

Jason Dion:
So the next thing we want to talk about is now we’ve kind of gotten a basics of what is ChatGPT and what are some of the things you can do with it. Our audience is people who are in the cybersecurity industry or trying to break into the cybersecurity industry. So one of the things we thought would be interesting to talk about is how can I use this as a job hunting tool or to craft my resume or other things like that. What are some things you think are good use cases for using something like ChatGPT as you’re on the search for job hunting or even once you’re in your career?

Sean Melis:
Yeah, absolutely. Very powerful for this. Because it’s such a great copywriter, it’s certainly helpful for rewriting key points within your resume. If you’ve got a resume with summary restatement and a few bullet points and you want to adjust the writing style so that it’s more professional or in a particular tone, it’s great for that. But I think the powerful part, particularly for job hunters who need to apply for things three or four times a day and scale that operation up to increase their odds is the personalization component. So you can go to the job ad, copy and paste all the requirements of the ad, paste it into ChatGPT with a prompt that says, “Act as my career coach,” or, “Help me to find a particular job. I’ll be applying for X num of jobs and I need assistance with crafting a personalized resume to those jobs. Below I’ll be pasting the job description of a job for this company. This is the title. And below that, I’ll be asking you to adjust the dot points of my resume to that job description to pull out the key things within my resume that are relevant to the job description,” and retailor those dot points, remembering the context of description.

Jason Dion:
That’s so cool.

Sean Melis:
Yeah.

Jason Dion:
Yeah, I want to point out though-

Kip Boyle:
We tell people to do that by hand now.

Sean Melis:
Yep.

Jason Dion:
Yeah, yeah. And the thing I want to point out is if you notice, dear listeners, what Sean just said. The way he just spoke to us is how you would write it in ChatGPT, you would tell it, “I want you to act as my career coach. You’re going to help me rewrite my resume, here’s my resume, colon, paste it in, here’s the job description, colon, paste it in. Now rewrite my resume for this particular job,” hit enter.

And within 15 seconds you now have a new resume written for you that is based on all those keywords it’s finding in there, which means if the AI sees it, when you apply for it, the AI systems that are scanning your resume are also going to catch it and that’s going to help you get through some of those HR filters. So it’s really critical when you do this.

And the nice thing is once you set this up like this and you have that chat string, you can say, “Do it again, but for this job, do it again, but for this job.” And you can in 10 minutes have 20 or 30 different resumes created for 20 or 30 different jobs that you need to apply for. And so it’s a really great time saver that way. And again, you can read through it, make sure it makes sense to you. The other thing I really like using it for resumes, this I don’t know about you guys, but I struggle a lot writing about myself and saying all the wonderful great things I’ve done or I tend to be very boisterous and talk a lot or type out way too much. Go figure, right Kip? I know Kip’s giving me a look right now. “No really? Jason, you talk too much.”

“Yes I do.” And so if I have, a lot of us struggle, we’re like, “Hey, I was at this job for five years and I have three paragraphs of information on it.” Give that to ChatGPT and say, “What are the two or three bullet points I should create from this paragraph?” And it will do that for you. And so it’s really good at summarizing and helping you in that type of stuff. So I think resume writing is a really good use of ChatGPT. What are some other good suggestions you think we [inaudible].

Kip Boyle:
Can I ask a question before we go on? I just have a question. So one of the things we tell our folks is we say, look, we want you to write your resume a certain way. And what we are essentially telling people is write it like a newspaper article where you’re the top of the page, you’re going to say who you are, what you’re looking for, and then as you go down the page you’re going to unveil more and more details. And the idea is to hook a hiring manager so that they can’t help but to read the next thing, the next thing, the next thing. So we do that. Now can ChatGPT emulate the style of the resume the way we are suggesting that it be written and, Sean, how would you tell ChatGPT to do that?

Sean Melis:
Yeah, great question. So you would pretty much explain what you just explained to us. When I create a resume, I recommend that it’s structured in a certain way, kind of like a newsletter, sorry, a newspaper where the key information is up the top. I want to restructure my resume in this way. Also, below I’ll paste the summary, the dot points information from the resume. Please reconfigure the structure of the resume to follow this structure. So you’re almost prefacing how the ChatGPT career coach should rewrite the resume. And you can do that in any way. You can include an example if you’d like and paste in an ideal example from a blog post or somewhere that you find online. But in terms of how it creates that output for you, just provide the context to how it should be structured and what sort of format should follow.

Kip Boyle:
It shouldn’t be that easy, Sean.

Jason Dion:
But it is. Yeah. One of the things I do-

Kip Boyle:
That’s why it’s magical.

Jason Dion:
One of the things I do a lot with that, Kip, to solve that problem is I would take, for instance, our template that we give in the irresistible course on Udemy. And if you take that template, you can say, “I want Jack act as a career coach. This is the format I want you to use for my resume,” in between the two sets of three stars. And I’ll put star, star, star enter, paste in our sample template, star, star, star. And then I’ll say, “Now between the dollar signs, you’ll find my resume,” dollar, dollar, dollar, “Here’s my resume.” And then that with dollar, dollar, dollar. And I’ll say, “Now here’s the job posting in between dash, dash,” and I’ll put that in there. So now they have the template of what I want to look like, my current resume and the job posting.

And then I’ll say, “Now rewrite my resume in that format and it will spit that out.” And then because I just started that thread, I only have to do that the first time. Then the second time I could say, “Now here’s a second job post, write me a new resume.” And it will then spit that out and you can keep that conversation going.

So it already has the information, you want it, here’s my template, here’s my resume for all my facts and figures. And like I tell my military folks when they’re getting out, you might have a seven or eight page resume because you need all those details because it has to be cleared by the clearance person first. But you’re not going to put all that on your resume every single time. But you can give that to ChatGPT and now it can pull what’s relevant for this particular job out of that seven pages to get you that one to two page resume that you want. So having that conversation is really, really helpful.

Kip Boyle:
Yeah. Okay, that’s enough about resumes. We have time and constraints. What’s next?

Sean Melis:
A big one I think for people is interview preparation.

Kip Boyle:
Yes.

Sean Melis:
Utilizing ChatGPT to act as the potential interviewer. And you can do that after you’ve drafted this personalized resume because it’ll have context of the job description, how you are framing, basically selling yourself for this particular role. And then say, “Based on the above and what you know of the job description, pretend to be an interviewer from this company and ask me questions that would be asked in the interview. But after you send me the first question, stop. I’ll respond verbally and then when I type next, send me the next one.” So you’re almost creating this mini application within ChatGPT where you’re able to have this conversation with, obviously it’s not going to pick up the verbal response if you’re on your phone or you have a Chrome extension, you can input audio into the ChatGPT

Kip Boyle:
I was wondering about that.

Sean Melis:
Yeah. And maybe have it actually read the response that you’ve given verbally and come back with a follow up question. I haven’t tried that yet, but I’m sure that’s possible. Essentially just creating this back and forth of questions and answers. And then you can write after maybe six or seven of them, come up with 10 more unique out of the box questions that may not be asked that will challenge my thinking in this space. Really push yourself and cover whole sort of bases.

Jason Dion:
Yeah, I’ve done that and played with that a little bit myself. And just like you said, Sean, if you’re doing it on your iPhone or I’m on a Mac, I can use dictation where I can hit the function key twice and it will take whatever I’m saying and convert it to text instead of me typing it out. So I will do that with ChatGPT and say, “Ask me a question, I’ll then respond and then grade my response and tell me how I could have improved.” And so it’ll say, “Kip, why are you looking for a new job here? Why do you want to work here?” And I’d say, “Oh, I want to work here because you’re a great company. You’re a Fortune 500,” blah, blah, blah. And they’ll say, “Well that’s good, but you should have talked more about what we can bring to us instead of what we can bring to you.” And they’ll give you that feedback.

Now again, it’s kind of the 80% solution. It’s not perfect, it’s not as good if you were sitting down with Kip and getting one-on-one coaching or somebody who does this for a living. But it is pretty good. And the other thing I would tell you is the results you get out of a ChatGPT are only as good as the prompts you give it, which is why prompt engineering is so important. And that’s one of the things I love about your course, Sean, is you include over 700 different prompts for different things, whether it’s how to work on the job, how to summarize email, how to build Excel formulas, how to do interviews, how to do resumes, all that stuff is part of your course that people can dive into and get a lot of great work out of there.

But being able to do that and having that two-way conversation because it is a simulated assistant and it actually is much more intelligent than the old chatbots of the previous days. What other things would you use in your job hunting or job resumes, interviews, negotiations? What are the things can you think of Sean?

Sean Melis:
Yeah, I’m keen to get your feedback on this one as well in the context of cybersecurity, but I think using it as a brainstorming partner for potential roles, someone may not know all of the different roles that exist within the cybersecurity industry. You could say that in response to the resume as well. Here’s my resume list out 30 different roles related to the skillsets that I have within the industry of cybersecurity, just to give you a broader understanding of different pathways you can go down. People may not know that there are these 20 other roles that exist that are perfect for their skillset or they just need to do one or two courses and then they’ll be ready for that. Yeah. Is there a very broad sort of range of roles that people may not know that exist and would this be a useful way to find that out?

Kip Boyle:
Oh, yeah.

Jason Dion:
Yeah, definitely. I know earlier on in our podcast, I think it’s episode 33, I always forget which number it is, but I think it’s 33, your cyberp.com/33 is all the jobs around the SOC. And we spent an hour going through and talking about five minutes each for each of the 30 different jobs or so going around the SOC, what is a SOC analyst, what is a pen tester? What is a SOC director? What is an instant responder, blah blah blah. But you can do the same thing in ChatGPT and just go, “Tell me in 500 words or less, what is the job of a instant responder?” Because you might see a job description that uses some title you’ve never heard of and you can grab that and throw it in ChatGPT and get it explained to you.

The other thing I really like is that you can ask it to respond to you at a certain level. So I’ve had some things I was doing recently in a business deal and they were talking a lot of accounting stuff and MBA level stuff that I just didn’t understand because I’m not an MBA. And so I started taking some of the questions when I was getting questions saying, “Hey, we want you to give us this thing.”

And I’m like, “What is that?” I went into ChatGPT, I said, “Explain this to me as if I was a fifth grader.”

And it would come back and say, “Oh, well this concept means this, this and this.” Like, “Oh, now I know what they want. I call that this thing. They call it that thing,” and that helps that language as well.

There are some cons of using ChatGPT and I already kind of mentioned, we talked a little bit about the hysteria that it has where it makes up things with hallucinations, the fact that it thinks it’s right when it’s wrong. So one of the other big cons that I found with ChatGPT is the relevancy or the currency of the information. Last time I checked, I think they were only current up through 2021. Is that right Sean? Or is it newer now?

Sean Melis:
Yeah, September, 2021 is kind of where [inaudible] ends.

Jason Dion:
Yeah. So yeah, ChatGPT doesn’t know anything about the world after September, 2021. So if I ask who is the prime minister or who is the king of England right now, it would go, “There is no king of England, it was Queen Elizabeth.” Well she’s passed away and now the king is Charles, right? But it doesn’t know that because it’s world knowledge stopped two years ago, essentially 18 months ago. There are other AI tools out there like Bard that are connected to the internet and they’re more current, but I still have found that ChatGPT does a better job than Bard in most things. One of the big differences though is that Bard can take input from the internet. So you can actually say, “Consider the following job posting, paste in the url,” and Bard can go out and read that and then respond to you.

ChatGPT can’t. It’s a sandboxed environment. So that is one of the limitations of ChatGPT when you’re using it. And so if it’s writing your resume, it may be great to write a resume in 2021 style, but if people are doing it differently in 2023, which does happen, then it’s not going to be hitting those things. So if you’re asking, we use a lot for SEO optimization as well on my social media team, that’s one of the things I always warn them is the way that Google and YouTube and Facebook looked at algorithms in 2021 is not the same way that we do it today in 2023. So ChatGPT is using 2021 algorithms, not 2023 algorithms, and that can give you an improper result. Are there other limitations you’ve come across, Sean, that people should be aware of?

Sean Melis:
Well, I think the accuracy of the outputs is really the major limitation and the drawback for a lot of larger companies who want to implement this. The way they’re getting around that is having a very strict sort of auditing and verification and validation process where it’s just part of this workflow. You input your prompts, you get your output and then you spend however long you need to validate that response through Google searches, looking at relevant websites, blog posts, journal articles, make sure that all of the information that’s being presented there is accurate and up to date. And even with that process, you’re still saving 60 to 70% of that existing sort of research phase. It’s giving you the structure to understand what kind of items need to be included in that output, and it’s your job to just make sure that those items are date and accurate.

Jason Dion:
Yeah, one of the other limitations I’ve come across is length of your input and output. So again, I’ve been trying to use it for drafting scripts, writing practice exam questions and things like that. And so if I give it something that is five pages long, it will choke on that and it goes, “This is too long, I can’t do it.” So I have to break it up into smaller pieces or feed that information in chunks and say, “I’m going to give you five pages of information, one page at a time. Here’s the first page,” paste.

It says, “Okay, give me the second page.” I give it the second page. And I’ve come up with some workarounds to do that. But that is one of the things is length can be an issue and it will go, “That’s too long. Try again with a smaller prompt.” Or when you try to ask for things on the output, for me, I have a certain length of video that I like, and for most of my videos they’re between five and seven minutes.

So for me that means I need 500 to a thousand words on a topic. So if I was saying, “Hey, ChatGPT I want you to write me a thousand word script on how to do job hunting on LinkedIn,” or whatever it is, it can do that. But when I get the response, sometimes I’ll look at them like that looks pretty short, and I copy and paste it over to Word and it’s like 300 words and I’m like, “No, I told you to do a thousand words. Try again.”

It’s like, “Oh, okay. And then it’ll give me 700 words.”

I’m like, “No, I need a thousand words. Do it again. Add an example, add an analogy, whatever.”

So sometimes you have to coax it along. And as Kip said over zealous intern, it’s kind of like if I had a high school intern working for me, I said, go off and find, write me a script. They might come back with something that is close but not perfect and I still have to check it. I still have to make it expanded to the way I want, add the right examples. Or sometimes the examples they give are stupid or not relatable, and so I have to have them change those. And so those are some of the limitations I’ve found when using ChatGPT.

And when it comes to job hunting, you can’t use ChatGPT and say, “Go find me five cybersecurity jobs,” because it doesn’t know anything after 2021. The jobs it finds are going to be outdated by-

Kip Boyle:
It can’t even search the internet by itself, can it?

Jason Dion:
Exactly. But it does know everything that was on the internet as of 2021. So it does know all the jobs that were there. And so if it says, “Hey, is there a penetration job available in Orlando, Florida?” It’ll go, “Yes, there was. Here it was. It was with Lockheed Martin and they were offering it for this position, blah, blah, blah.”

And it thinks that’s current, but that’s a three-year-old job that nobody’s hiring for. So you need to be aware of that. And so it’s useful for research development, helping you with the resume, helping with your interviews, but if you’re looking for current, current, current information, ChatGPTs not going to be able to help you there.

Kip Boyle:
Right. Okay. Can you think of any other-

Sean Melis:
Very shortly, I think they’ve released the web sort of crawling plugin. It hasn’t shown up on mine yet, but I know there are some countries around the world that have access to this and that will solve that ability to access more frequent and recent information.

Kip Boyle:
That’s the kind of thing I’m waiting for because I often have to write about big things that just happened in cybersecurity, but I can’t ask ChatGPT to help me because it’s anything about something big that happens. And I also need to interpret it for the context that we work in. And I have not yet found a way to prompt it to get it to do that. I don’t know that it can create really genuine insight, genuine human insight. I don’t know. Sean, what do you think?

Sean Melis:
Yeah, I think it depends on, again, the prompt and whether you wanted to almost imagine genuine human insight. Give it the frame of what it needs to embody the way in which it approaches this response or problem. You can have it act as a particular person throughout history or a combination of authors or domain experts from a field. It’s that sort of persona that it embodies prior to engaging in the response that’s going to give you that best possible output. It’s really framing how it’s going to engage. And that’s important if you’re wanting it to have insight from a particular point of view.

Kip Boyle:
So once again, all roads lead to prompt engineering.

Jason Dion:
And the other half of that though, Kip, that if you want to be technical and get really into the weeds, and if you’re a programmer, you can do this is OpenAI does have an API that you can call and use. It’s a paid service and they charge you based on the prompts and the tokens you’re using. So if I use 500 words or a thousand words, a thousand costs more than 500 and that kind of thing. And we actually use this at Dion training. My CTO has built in our own ability to query ChatGPT from inside of our own systems using the AI and the API. And you also have the ability to put custom models on that as well. So instead of it just being trained on the general internet, you can then say, “Okay, here’s a bunch of Jason’s scripts that he’s written in the past. Now you understand how Jason writes.”

Or, “Here’s how Jason’s written practice exam questions. Now you understand the format that we like to use. Go and write me 50 more questions. Go and write me 10 more questions, go and write me another script.” And you can do things like that.

The other nice thing about using an api, if you’re a coder, and you can do this very simply with something like JavaScript or Python, if you have a basic coding knowledge is let’s say I wanted to write 30 resumes for 30 jobs. Well, I can use Python to crawl the website and grab those job postings and then I can feed that in through this prompt and say, “Now consider this job posting,” it automatically paste the text in, gets the resume back, and then shoots that out to me as a PDF. And I can actually automate the entire thing where I can say, “I’m looking for a new job as a CSO. Go out and find me any job on a company that’s at least on the Fortune 500 that’s open, go find those positions, take my resume, rewrite it, give me it, and then tell me where to submit my resume.” And it can do all that in the background for you.

And for us, we do a lot of that so we don’t have to copy and paste in every objective from a CompTIA exam. There’s 36 objectives and 50 different sub bullets on each objective. And so we’ve written some scripts where we can tell ChatGPT to go and create a baseline of questions for us. And again, that’s our first draft. And then we go and have human experts like myself and Reid and Jamari go through and rewrite those for what we’re going to use in our final course.

And then the last thing I would say about ChatGPT as we start wrapping this up is that remember that it is being the one that’s doing this generation. And so if you take its content and try to pass it off as your own, you can’t claim copyright on that. I’ve seen people writing books with it. I’ve seen people creating images with it, creating PowerPoints with it, writing poetry and songs with it. That’s all great and you can copyright the music, but you can’t copyright the lyrics if ChatGPT wrote them. And so using it as a basis and a first draft is okay, but if you make significant modifications to it like we do, now we can copyright that content because we now own it because we’ve changed it significantly.

So keep those in mind. And this is an area that is still evolving rapidly. There’s a lot of cases going on. I know right now the Senate and the House of Representatives has been having hearings on AI and trying to figure out what this is going to look like here in the US under copyright law and intellectual protection and privacy rights and all of that. And so this is an evolving field, but it is a great tool that you can use right now for a lot of the things you’re doing at work or as you’re trying to get a job. That being said, I’ll pass over to Sean for any final comments and then I’ll pass it to you, Kip, and then back to me to close out.

Kip Boyle:
Great.

Sean Melis:
Yeah, no. Thank you so much. I always encourage people to just experiment with it, even ask it what prompts will be helpful for this particular task, treat it as the super intelligence as it is. It can help with anything, any level of complexity, any problem, any challenge, any position you have. It can take on any person, any identity, any expert. Just consider that the power of that in your hands and how it can help you across your personal and professional life. Because it’ll also save you a lot of money with having to go to lawyer’s, accountants, just to get that initial sort of advice and feedback on things which are typically gated behind very costly service professionals. This is a good way to get that initial education before engaging someone. So I’ve been using it a lot for those type of use cases. And as I mentioned earlier, I’ll be putting out a course related to ChatGPT for LinkedIn and careers and job hunting, combining it with Notion, which is a great information and document management tool to copy your resume into different pages and all of the variations for job descriptions.

I really want to create an operating system for people to find their dream job.

Kip Boyle:
Oh, that’s fantastic. Well, I’m so glad we had you on this episode. And listen, when you get that launched, you got to come back and tell us more about it. And when I listen to you talk, what are the takeaways that I have from this conversation is, and I was a little kind of confused about this, but Jason and I have spent a lot of time thinking about, okay, we’re hiring managers, how do we hire, what do we look for? And so in our course we share all this information about what makes a good resume. How do I know when I’m interviewing somebody that is sharing with me information that makes me find them irresistible? How do you do show that on a resume? How do you do that in an interview? And I think what I’m hearing is that you’re saying our advice, if somebody was to take our advice but then use ChatGPT to just automate the things that we’re saying that they ought to do. If that’s a winning formula, would you say that’s accurate?

Sean Melis:
Yeah, I see ChatGPT as these mini applications that you can create. Each conversational thread is its own application. So you are creating a workflow that uses a conversational input to execute on that application. So whatever advice you’re receiving and whatever work you need to get done, you are able to create your own personal application to do that and you can reuse it over and over again.

Kip Boyle:
So Jason, I think we need to create some canned prompts for our folks to sort of paste this in ChatGPT and then ChatGPT will act just like Jason would and you can get coaching and feedback from ChatGPT almost the way you’d get it from Jason. That’s really powerful.

Jason Dion:
So I think what we’ll do for that Kip, is for anybody who would like to get those, we will put those in the mentor notes along with this episode. We’ll do it on the next one after this episode. So this episode is on June 23rd, so for the first week of July mentor notes, we’ll put those in there. So that gives us a little time to get those together. And that means if you want to get those, you’ve got to sign up for the mentor notes, which you can do over at yourcyberpath.com. They’re completely free right on the homepage. Just go ahead and sign up there with your email and that way you’ll make sure you get those. And Kip and I will work on giving you some really good prompts that you can use like a one-page PDF attachment to that mentor note that week that you can use as a cheat sheet and a study guide.

I was just going to echo what Sean said is I use ChatGPT a lot when I need something complex that I know how to do, but I don’t want to take the time to do it. So for example, earlier today, Kip, we were working on some practice exam questions and one of them was this little chart that we created with four different types of way to do a risk, right?

Kip Boyle:
Mm-hmm.

Jason Dion:
And it said, “Hey, if you do this, it’s going to cost you a million dollars upfront and half a million dollars a year for the next three years. This one is going to cost you $250,000 now, but a million dollars a year,” and had all these different formulas. And to answer the question of the total cost of ownership, you had to basically do four complex calculations and then compare them all. Well, it would’ve been a lot easier just to take that and go “ChatGPT, if I have option A, option B, option C, and option D of these styles, which one is the cheapest? Which one should I select?”

And it will go, “Oh, you should choose C because here’s the calculation and here’s why you shouldn’t choose A, B or D because here was those calculations,” and you can see all that very quickly. I was doing a business deal where we were going back and forth on negotiations and we were trying to figure out how much was going to be upfront as a down payment, how much was going to be paid over time, what the interest rates were. And there was three or four different categories that I could have chosen from. And I literally pulled out my phone with ChatGPT and started typing in and saying, “Tell me which one of these four is the best.”

And it came up with, “This is the right answer, choose that one.”

So I could do it all my own, but I didn’t want to. And it was like that instead of having to spend five or 10 minutes doing all the calculations.

Kip Boyle:
That’s great.

Jason Dion:
So it’s great for stuff like that and keep that in mind as you’re using it. The other thing I would say is that ChatGPT is sponsored. It’s owned by OpenAI, but a large portion of their funding came from Microsoft. And so Microsoft is now integrating it into a lot of their tools. For example, if you go into Microsoft Office, including PowerPoint, Word and Excel, they have ChatGPT integrated in so that when you’re doing different complicated formulas in Excel, instead of having to do all those complicated formulas, you now can just say, “Excel, given the data in sheet one, tell me what is the average across all my sales for the last 12 months,” or, “Which is my best selling course?”

And it will do that analysis for you without you having to go through and do it. So it really is a great assistant and a really good helper. So definitely something you want to check out. And it continues to evolve. They’ve gone from cheap ChatGPT three to 3.5, and now to four. Sean has been keeping his course updated with it. I just looked at it today and he does have ChatGPT four in there along with those 750 plus prompts and it’s been fully updated. So definitely check that out. If you want to get that course-

Kip Boyle:
Yeah, if you need to get it. If you want to stay computer lit, computer literate, you need this.

Jason Dion:
Yeah, it’s 10 to $15 on Udemy, super inexpensive, really good quality.

You can get the direct link to it at linkedin.com/in/seanmelis. That is S-E-A-N-M-E-L-I-S. And we’ll also put that in the show notes as well. So you can click on that directly and get that ChatGPT 1 0 1 course. I just want to thank you again, Sean, for being here with us, and that marks the end of another episode of Your Cyber Path. We’ll invite Sean back once the new LinkedIn course is out for job hiring, because again, I think that’s going to-

Kip Boyle:
The Udemy course.

Jason Dion:
… I think it’s going to really help all of you in the audience. And yeah, so in the meantime, if you want to connect with Sean, go over to LinkedIn, linkedin.com/in/seanmelis. He’ll be glad to talk with you there and any questions you have, just let him know and he’s there to support you guys, as well as us. So we’ll talk to you next time on the next episode of Your Cyber Path. Thanks all.

Kip Boyle:
See you, everybody.