POVs: AI will kill us all VS. Let's figure this AI thing out. Plus staying curious, and not letting FOMO win

Announcer: 0:00

Welcome to the Prompt Engineering Podcast, where we teach you the art of writing effective prompts for AI systems like ChatGPT, mid Journey Dolly, and More. Each week we explore prompting techniques, interviews with experts and newbies, and tips on selling your prompts. Here's your host, Greg Schwartz.

Greg Schwartz: 0:19

Welcome to the Prompt Engineering Podcast. I'm your host, Greg Schwartz. So we have a guest with us today. Go ahead and introduce yourself.

Costas Schuler: 0:29

So it's funny, my dad was an electrical engineer, so he always pushed me for those things. Ended up doing some calculus in high school. Then I'm like, dude, I'm out. I don't get the jokes anymore. I don't belong here. But then I ended up going into graphic design, which is for me, the perfect combination of like technical and creative. So I was able to consolidate the two ideas and mourn on the creative. And what's interesting about the I got exposed to ChatGPTI Alex Hermo video, he did a video and he's you gotta watch this. And then I watched it and it was like, oh my God, I gotta get on the, on this. So I got it and I started using it. And then for me, it's for me, I think your experience matters because then you know, you have an under. Certain questions you're gonna ask, right? And so I'm also part of this coaching group, and part of the thing that we do is we ask better questions.

Greg Schwartz: 1:25

Yeah,

Costas Schuler: 1:26

questions lead you to better answers. But if you don't know what question to ask, you won't get the right answer. So it's a very interesting because you don't know what you don't know. I have no idea about coding. So like, when I see some of these, prompts and things like that on the internet, on YouTube, I'm like I wouldn't even know where to start to ask. I don't even have a language to ask those questions, so I don't go there. So I ask other things that I'm interested about. I ask those kind of questions. And because I'm I'm a curious type, like I've worked on, like if my car doesn't work, I want to get in there and see why it doesn't work, and then, That's what I do.

Greg Schwartz: 2:03

One of the things that I was curious about is what, since you were saying you, you aren't an engineer and that totally makes sense. A lot of people in this aren't. What techniques have you been using?

Costas Schuler: 2:14

For instance for my business, so I do graphic design and I do websites. So real basic emails like I write an email out and I say, Hey, can you make it a little bit better? Boom that's great. So that helps me with that and organizing my thoughts. For instance, if I have a convers, discovery conversations with a new client, I write all this stuff down, all these ideas, and I basically just punch 'em in the ChatGPT and so it goes through and organizes this, and then I see it. I'm like, okay, can can you make it a little bit less formal? Can you talk about this? Can you add this? Can you take, so I'm. It's been amazing, like for proposals, like I hate proposals, but this is I can get all my random thoughts and these, organizes it all.

Greg Schwartz: 2:53

That's awesome. So it sounds like mostly zero shot prompting, probably not much in the way of role-playing is that.

Costas Schuler: 3:00

Okay. So I did some role playing in the beginning. I was like I did some sales training. I'm like, okay, pretend you're helping me with sales and let's like have, let's have a conversation. You're the client. I'm the, the salesperson and that I'm asking you question. To get to, to know you a little bit better. So I, I did that a few times, so that was interesting. And then it gets a little wobbly and I'm like, I gotta get it back. So I try to get the program back on track.

Greg Schwartz: 3:25

Yes. Yes, totally.

Costas Schuler: 3:27

There's two views of people out there. Friends of mine, they're like, oh my gosh, the apocalypse is gonna kill us all. That's the one side and the other side. They're like, okay let's like figure out how this thing works, so I always tell people like, listen, apocalypse aside, I'm gonna figure out how this thing works. It's interesting to me.

Greg Schwartz: 3:44

Yeah. What originally got you into prompting and ChatGPT?

Costas Schuler: 3:50

Alex Hermo, he's all over the internet. He puts a video out his, he says, you gotta check this out. So I was like glued to his explaining what ChatGPT was about. I'm like, that's like for me, if I hear something that like. Takes a hold of me in such a powerful way. I'm like I'm on it. I'll stop when I'm, I'll start, just, I'm logging in, I'm starting, I'll get on it, and so that's why I got it like this last weekend, this interesting. I was like I got it in my noggin I wanna do NFTs. So I use ChatGPT is okay, if I'm a beginner wanting to do NFTs, what is the process? What is the steps I need take? And so boom. You gotta get a crypto wallet, you gotta get this, you gotta get this, and then give some suggestions. And then I'm like, okay, what is this? What is this? And I ask a question. And then within an hour I was on I had a NFTs up, like I got the crypto, the Coinbase, I got this thing, I got the thing on the browser, and I got the NFT three, like NFT websites, and I did some art. I loaded it up there and that was like really amazing. Like it helped me like do this within an hour or so.

Greg Schwartz: 4:56

That's awesome.

Costas Schuler: 4:57

And that's amazing. I could do stuff I could learn about things and I, I asked you questions like what is like the other last night I saw I'm a weirder I saw we're watching movies and these guys clink their glasses. I'm like, okay, where did that? So I asked, I said, where did Clinking come from and what is that all about? So he told me the history, and then it was all about the clink and the G, they would spill into each other's glasses. So they would know that it wasn't poisoned.

Greg Schwartz: 5:23

Is that actually true? I don't know.

Costas Schuler: 5:25

That's one idea that it came up with and it had some other ideas. But that was like, that's an interesting, so I do things like that. Something gets in my head, I'm like, what is that?

Greg Schwartz: 5:32

It sounds like you're doing less of I am crafting a prompt for a specific thing and I've been working on it for three weeks or whatever, and more of just the, I have a question or I have a problem, or I have a project like the NFTs and I wanna learn about it. Is that right?

Costas Schuler: 5:48

Yeah. And also, the weight training, it's yeah, I wanna create myself a six week program so I don't have to write it all down and do the math and percentages and Excel files, and I just wanna do this. And that's what I'm working on. If I can create myself a six week program strength training, which is heavy on Monday's, light on Wednesdays, and medium on Fridays with these different exercises with warmups and all those things then it's like I can punch it in. I could create programs for people based on these prompts.

Greg Schwartz: 6:17

So I'm putting it on screen now. I'm curious to hear, how did you iterate on this? I'm noticing some pretty specific, you're putting in the starting weights, you're, giving it specific instructions about like when to do different things.

Costas Schuler: 6:32

So basically this program is a linear progression. So with squats, you start with squats. On Monday, 200 pounds say okay, on Wednesday you do like light, 80% less, but, and on Friday you do medium. And so actually I'm 51 years old. I'm intermediate. If you're a young at 20 something, you're basically gonna be adding five pounds every time you do squats three times a week, you go to a 200, 205, 200. Monday, 215. 200 2200 like you'll add and you'll get huge, like I started doing this stuff at 45 pounds, now I'm up to 210 pounds squatting.

Greg Schwartz: 7:08

Nice.

Costas Schuler: 7:09

Okay. So you get a certain level, like intermediate like you plateau a little bit and you this age, so you have to mix it up a little. So I'm trying to do this sort of, heavy light, medium on Monday, Wednesday, Fridays. And that's what I was trying to. There is an actual program that does it. I just don't wanna have to do the math. I just wanna be able to say, okay, start here in six weeks, I'm gonna end up here. So the numbers keep adding up over the six week period.

Greg Schwartz: 7:35

That totally makes sense. Okay. Since we've got people both watching and listening, The watchers are able to see this, but the listeners can't, obviously. So let me just read this prompt off real quick. As your strength training coach, I will design a three by five program based on the Starting Strength Program by Mark Reito that suits your requirements. You are a 51 year old intermediate lifter looking for a heavy, light, medium program, and you want to squat 200 pounds on Mondays, Wednesdays, and Friday. Deadlifts will be on Wednesdays only on odd weeks. You want to bench press on Mondays and Fridays and overhead press on Wednesdays, on even weeks. Oops, on even weeks. You want to do overhead press on Mondays and Fridays and bench on Wednesdays. Your starting weights are 200 pounds for squat, 115 pounds for overhead, 160 pounds for bench press and 296 pounds for deadlift. Each lift will have a warmup set starting at 45 pounds with three reps, three, three sets of five reps, one set of three reps, and one set of two reps. After the warmup sets, you'll do three sets of five working sets for each lift. This program is specifically designed for an older intermediate lifter like you. So I'm curious, just have if, go ahead and read out the some of the output.

Costas Schuler: 8:51

It did do good. It got a little confused, honestly. For instance, I wanted to do Bench on Monday and Fridays and overhead on Wednesdays. And then the next week you switch it. So you do overhead on Mondays and Fridays and bench on Wednesdays. But it wasn't quite understanding that, so I, I ran into a wall, not a wall, but I didn't quite complete my task or get it to do it the right way. I was gonna do week one and week two and give it an example and then say, okay, now do this for the next six weeks. Figure out a different way to to tackle the problem. Because it was, I think I confused it.

Greg Schwartz: 9:25

That makes sense. And just for, the audience, that's an example of giving a shot prompt. So we've talked a little bit about shot prompting on previous episodes, shot prompting is either, I'm not including any examples of the output, which is called Zero Shot. I'm including exactly one. His example of here is what Monday of week one would look like and it's, this lift, this number of weights, et cetera. And then few shot prompting is, let me obviously give a bunch of those.

Costas Schuler: 9:55

I think sometimes it needs it.

Greg Schwartz: 9:57

Yeah, it definitely does. And did it help to add that in this case?

Costas Schuler: 10:02

Wait I haven't, I also, I was actually trying to do it while I was waiting for you to get on call, so I haven't finished. I think it's gonna, I think it's gonna work. I just haven't completed the task, which I'm gonna do this week.

Greg Schwartz: 10:14

Cool. So just for the listeners here's an example of the output. Here's your revised program with squats performed on all three days and deadlifts only on Wednesdays. Obviously it's still getting confused, but week one, Monday, squat three by five at 200 pounds. Warmup sets 45 pounds, five reps 95 pounds, five reps, 135 pounds, five reps. Wednesday, here's your, so I'm not gonna read this whole

Costas Schuler: 10:39

Yeah. No

Greg Schwartz: 10:39

get the idea.

Costas Schuler: 10:40

It's like it's geeking out for the weight training crowd. If you look at this, your eyes are gonna roll back in your head and you're gonna, just, you might as well tell me, explain code to me. I'll have the same reaction.

Greg Schwartz: 10:51

So actually one trick that I have seen that I actually haven't personally used, it just hasn't been relevant yet, but is asking ChatGPT to format the output as a table. And that might be useful for you in this case, just so that you're not, here's a paragraph and have to break up each of the things by day and,

Costas Schuler: 11:09

it was doing that originally. I asked it to do the columns

Greg Schwartz: 11:12

Nice.

Costas Schuler: 11:13

and then at some point it's no, I'm not gonna do columns anymore. I think I'm gonna do your list. You know it, it's it

Greg Schwartz: 11:18

as always, yeah. Once things roll out of context, it'll start forgetting and you have to go in a table please.

Costas Schuler: 11:26

that's right. That's right. But the tables are very useful and it's, I was like, even the first reiteration, when it spit out the numbers, it was like, it wasn't very impressive to see that rollout and because, I know what the numbers should look like, so I could see that it was adding the five pounds and the 10 pounds to the deadlift. So it was doing its job. It was doing it, and it, it wasn't perfect, but the fact that it was cranking out these numbers, I was like, dude, this is cool. Getting pumped. I'm getting pumped.

Greg Schwartz: 11:56

Yeah. Getting pumped for weightlifting. Yeah, absolutely.

Costas Schuler: 11:59

And I've only been doing weightlifting for three or four years. So I was like never a gym person. Gym bro. I'm not a gym bro. But it's as you get older, you need to get fricking, your muscle mass and your bone doesn and it goes down. So you need I don't wanna fall down, break my hip.

Greg Schwartz: 12:12

Good call. Yeah, good call. In the vast experience and I'm joking when I say that cuz earlier before we got on the recording, we were chatting about how almost everyone I've talked to is oh yeah, I have two months of experience doing this." "I have six months of experience." But in, in your vast experience, what would you say are the key factors in making a prompt successful?

Costas Schuler: 12:36

As a veteran of three months I would say, that's funny right there. I would say you have to remain curious. You have to be like, this is just, it's a tool. And in order for you to succeed, it's like you have to stay curious. You have to keep asking questions you have to ask better questions. You have to it, it helps if you know what you, you want the answer to be. So it's basically just massaging it talking to it and asking better, different questions to, to get what. It, it takes time, but it's if you're not fascinated by this process, you're not gonna like doing this. But if you're curious and you're like, oh, interested, if you're geeking out on this kind of stuff you'll be there for hours, doing this. And it's, I think, stay curious my friend.

Greg Schwartz: 13:19

I can definitely agree with that. I've had many a late night of "Wait, it's what time?" Yeah. We're not alone on that one. I've heard that from a lot of people. What would you recommend to people who are starting with prompt engineering? As so many of the listeners are where they're just like, ChatGPT sounds cool, but I don't really know how to use it. Like, where would they start?

Costas Schuler: 13:41

I would start with what is it that you do and you're interested in trying to basically like figure out a way to use ChatGPT to help you do what you do. Whether it's your hobby, your business, your, your interest, anything. It could be anything just. Figure out a way or come up with a way to help you do a, an aspect of it. Even if it's like, Hey, this thing helps me write better emails. Boom. I don't have to I hate typing. I do this because this is how I feel when I'm typing. I don't have to, like I'm

Greg Schwartz: 14:13

some very T-rex like

Costas Schuler: 14:15

yeah. That's how it feels. Like I'm the clucking. I'm, English is second language. I don't know how to spell things over and over. Though just certain words, I just don't get it. For years now. So start where you, what you know, and then build from there. And then as you go keep watching videos, keep watching podcasts. Be interested and find a niche or something interested and start, it's like a a thread on a sweater. Like you find something, you're like, oh, what's that? And you start pulling that thing and all of a sudden you discovered a whole nother, old take your. That you can't even imagine,

Greg Schwartz: 14:44

Nice. So your prompting, are you doing it in a prompt engine? Are you doing it in the playground? Are you doing it in just the ChatGPT sort of web app?

Costas Schuler: 14:55

ChatGPT if there's something else I should be in I'm, send me link. I'll sign up. You said there's like a playground? Is that something, that's something I'm.

Greg Schwartz: 15:03

So playground is a more technical version of the ChatGPT interface. It's got a bunch of interesting controls that you can use around what are called temperature and top p, which basically are ways of saying how random do you want the answers to be? If you're doing math, you want basically no randomness cuz you want the actual answer. But if. I don't know, brainstorming what to write for marketing copy, for example, you can turn the randomness up because you don't want tell me about problems that people have who are, guitar players in it to be like selling tickets yeah, no, duh, tell me something else. So you turn up the randomness and it'll be like, people, I don't know what it's gonna say, but, people stealing their sheet music or something like that. It's higher randomness leads to more creativity. But also more noise. So yeah, it's a double-edged sword.

Costas Schuler: 15:57

I think giving ChatGPT context is important, I think that's, that's helped a lot If you give it more information on background about where you going. I actually saw this video yesterday. It was very interesting. It says, then you ask it if you have any questions before you start. And then it will ask you if he has clarifying questions and then you can run the program. So that's very interesting. So I'm watching ChatGPT videos while I'm waiting, resting between sets. So I'm like, this, I got chalk everywhere, so yeah. Context is important.

Greg Schwartz: 16:25

The whole, Here's what you, I want you to do, but first ask me questions to improve the output. That nuance is really useful. I've seen that used on a few prompts.

Costas Schuler: 16:36

It's fascinating, man. I mean, the fact that there's already jobs that people do. There's a part of me that's. Fomo, there's a FOMO aspect to it. I don't wanna miss out like it, I'm interested in what this will do to my business, am I gonna be out of a job? Rather than be afraid of it, I'm embracing it because I'm like, I want to know what's happening and where it's going, and I want to be, on top. I just, I guess I wanna be on top of kind of tracking it so I can make moves if I need. If I need to pivot in my business, if I need to get out of the business or whatever it is I'm supposed to be doing, at least I have a my eye on something, then I can make a more informed decision.

Greg Schwartz: 17:15

That totally makes sense.

Costas Schuler: 17:17

It feels like I sh I should be doing that. Like I, I don't know how people are, like, I'm afraid of people who are like, oh, it's evil, and it's, it's gonna take over. I'm like, yeah, those are. Concerns that are valid, but don't just ignore it.

Greg Schwartz: 17:34

have you sold any prompts or bought any prompts?

Costas Schuler: 17:37

No I just got here.

Greg Schwartz: 17:39

No worries. I ask because I actually created a sales analysis for people on Prompt Base. So prompt base for context is a website where you can sell and purchase prompts. And I'm selling one about Writing a good cover letter where you paste in your resume, you paste in the job description, and then it outputs, here's a cover letter for that job position using your resume details. Anyway when I was thinking about how to build that, or really when I was thinking about what do I want to build, I went and did this analysis where I took all of the data on prompt base and basically said, okay, where are the niches that are earning? A fair amount of money, but don't have a ton of competition because it turns out some of them, like marketing as a whole, it has I don't even remember how many hundred prompts

Costas Schuler: 18:30

Oh, gotcha. So marketing a, it's kind A saturated,

Greg Schwartz: 18:34

exactly, at least on prompt base, maybe in other areas it's not a saturated, but I was specifically thinking, I wanna sell a prompt on prop base, so where should I focus my energy? So anyway I did this analysis. It is available on Gumroad at https://gregschwartz.gumroad.com. People listening, I will put a link to it, but yeah, that's why I've been curious about, if you're selling or if you're buying any, how did you think about that? So it's okay that you aren't doing that but feel free, it's actually pretty awesome.

Costas Schuler: 19:02

if I could figure out these uh, training course, training prompt. There's like ton of different like programs and that actually exist on an Excel file

Greg Schwartz: 19:13

That would actually be, yeah I bet there would be people who are like, I don't wanna figure out all the

Costas Schuler: 19:17

but the Excel files do the math. So you punch in the numbers and it does it all for you. But maybe it doesn't matter. I don't know. Who knows?

Greg Schwartz: 19:25

It's definitely worth trying. I know there are fitness prompts. I don't remember if there are any specifically around weightlifting. Most of them are around, give me six exercises to do at home, or, more nutrition focused stuff. Here's what's in my fridge, tell me what I can cook, kind of things. Is there anything in the news that has you particularly excited that you've heard about recently?

Costas Schuler: 19:47

I guess the increase in quality of ChatGPT tutorials and, insights and things. So that, that, that is exciting. It's also terrifying to see where this could go. It's like there's a terrifying and a exciting aspect, both at the same time. This is terrifying, but this is super exciting. Oh my God, we're all gonna die. But this is really amazing. So SkyNet's coming up online tomorrow. But wait, I gotta sell a couple prompts before then. I dunno.

Greg Schwartz: 20:15

Yes, I know what you mean.

Costas Schuler: 20:16

Just the more we focus on bad things the more I. Anxious and anxiety and anxious and it shuts me down. So it's a trick to be able to like, keep your peace do what you are supposed to be doing. What is it I'm supposed to do today? What is that excites me? What is it that brings me joy? Do that right? And then, We're all gonna die. Okay, that's a given. Fine. We're all gonna die. We don't know when, but we're all gonna die. But so remain in peace and do the things you're meant to do in life.

Greg Schwartz: 20:49

Yeah, and the FOMO thing is real, not falling for FOMO

Costas Schuler: 20:53

yeah, just, find the joy, find the thing that brings you joy, and go do that. And then forget about the clickbait stuff. If I can make you laugh like I made you, And to me that's regardless of whether I sell a prompt or whatever, if I have a I got to meet you, Greg. And that's the, actually, that was the most important part of this whole thing. If me getting on ChatGPTto get to talk to you was what it was all about then dude I'm good. You know what I mean? That's,

Greg Schwartz: 21:19

that's how I feel about this podcast as a whole is, I want to educate people about here's how to write good prompts and here's some really specific techniques, but also just connecting with people is so much fun. So yeah. Totally with you.

Costas Schuler: 21:32

This's not. This is a good conversation. Hopefully it was valuable to you and it brings value to your listeners. I, I love podcasts, so I'm like, I constantly listen to podcasts, so being a part of a conversation to me is super awesome.

Greg Schwartz: 21:44

Do you have any links, social media, anything like that you wanna

Costas Schuler: 21:48

Oh, I have my website I could actually I could put in, it's EnvisionaryDesign.com and that's my business. And from there's Instagram and Facebook and things like that. I don't tweet. I haven't figured that one out. Maybe ChatGPT can help me tweet better.

Greg Schwartz: 22:02

I think it can. I think it can. Awesome. All right, I will put your website in the description and thank you so

Costas Schuler: 22:09

Greg.

Greg Schwartz: 22:11

Thanks for coming to the Prompt Engineering podcast, the podcast dedicated to helping you learn how to be a better prompt engineer, how to sell your prompts if that's something you're into, or just use them in your day job. See you soon.