Ep 1 - The Common Sense of Prompt Engineering and The Future of AI

All right, welcome back, Ira, to who knows what episode this is. I don't know. A lot of these get corrupted and lost.

And with all this A.I., you know, we think that we'd have this down in by now. But regardless, we're back, man. How are you?

I'm doing well. And we'll look back at those times and we'll say, you remember all that now B roll that used to be a roll. This is B roll.

Yeah. This is B roll. So I don't know if this will get released or not.

So I have a question off the bat. You came across a pretty interesting study on prompt engineering. It's all the rage here, right?

So fill me in on that, man. What did you find? Yeah, so I mean, it was it was a variety of different articles and other topics that I was reading about it.

But prompt engineering, I mean, I don't think it's really a new thing. I mean, it's it's kind of like the evolution of Googling. But prompt engineering is is really becoming a major topic.

I would even say above and beyond this whole A.I. craze of chat GPT and prompt engineering. What it is at the core is, you know, that if you've played around with the tool and if you haven't definitely recommend it, you know, chat GPT or any of these these tools is the better prompt, better thing.

You can give it to to query on, to respond to and these kinds of things, the better response you get. And there is a bit of an art to understanding how you can input this stuff. And it's it's less about the tool and it's more about the subject matter.

But if you're an expert or a pretty proficient person in the in the topic, you can really make some really good prompts and get some solid information back, particularly if you're giving it some context. So prompt engineering is, I think, one of the really big things that will continue to really maybe even outpace some of the buzz happening with just the general A.I. topic.

Yeah, I agree. But at what point is this common sense? Yeah, I mean, it's like it's it's getting ridiculous.

I see the the LinkedIn carousels like how to. And I'm like, this is common sense. It's a conversation like I'm just blown away that like, what were these people using to find their information before?

How are they Googling things before? Like, I don't know, maybe just because I'm in that field. But at some point, it's got to be common sense that if you want to get information, it's almost like having a conversation with somebody like, yeah, maybe maybe there's an art to it that I'm just not seeing.

But but Trevor, and this is even, I think, what separates it on LinkedIn or Facebook or wherever you're seeing these carousels. But, you know, the idea of prompt engineering isn't just asking the right question. I think it's also giving it the context, because if you're just asking it the question, OK, I mean, that is good.

That's powerful. But if you provide it, say, the transcript of this video, right? And maybe all the other videos that we've done.

And then you ask it various questions about that in context with other resources that are out there to get some sort of solid output. I think that's really where the value comes in. So the prompt engineering is as much about giving the tool, giving the platform, whatever it may be, the the research, the background, what you want to pull from, plus the other resources you want to pull from, plus asking the questions in the format that you want and to get that output.

It's like programming without programming. You don't necessarily need to have the you don't have to need to know how to do like Python or C or any of these other programming languages. But I think you need to think a little bit like a programmer if you want that kind of output.

But that makes sense. Do you think it's more about the prompt or do you think it's about the individuals just don't really know what they want? Right.

Because like, here's a great example. If I let's just assume you and I are going to lunch, right? I look at this and I'm like, Ira, where do you want to go to lunch?

That's kind of like a basic prompt. I'm going to get a pretty, you know, generic output. But then if I know that, you know, you love Mexican food and you love it and I want to kind of see, I'm going to I'm going to kind of probe you with some context around that.

Right. So I think that's the bigger the bigger issue is that because if I want to really get some research, right, let's just say I want to find some research on, you know, whatever it may be. I'm probably not just going to go put a basic prompt in and expect it to read my mind.

Maybe this is common sense. Maybe I'm way off base. I'd love to know what people think about this.

But I mean, I think what Chats GPT is shown in all these other tools is really it's scary. It's just the lack of like, I think it's lack of common sense. It might sound harsh, but like, I don't know.

But these people are paying big salaries. What was the number that you threw out? Like two hundred and fifty grand or something?

Yeah. The the salary of a solid prompt engineer. Some of the numbers I've seen floating around is two hundred fifty to three hundred fifty thousand dollars.

And, you know, I think some of that is because it's new, you know, and sure. These kind of things. But, you know, Trevor, I get it.

Like if you're talking about where you want to go to lunch or these kind of things. But first of all, if I ask Siri, she's going to ask me seven times what I said, because, you know, she's just not a good personal assistant. But that's a whole separate conversation we can have.

I mean, I love her. I have her on my phone. You know, I'm an Apple fanboy, but man, she needs some work.

But, you know, when it comes to Chats GPT, you ask about where you want to go to lunch. Yes, it's you know, but you need context about, OK, maybe go figure out where he's been before, whoever you're going with. But it's that background.

And that's where I think the prompt engineers that you're seeing, these big salaries, it's not just them doing the query, it's them providing the research. And if you really look at what's done, like I've heard reports that especially with coding and there's so much available out there and coding that you can pull from resources is I've heard of people downsizing significantly on their coders and reducing it down to one, two, three coders, and they're able to produce the same amount of work with fewer people using Chats GPT or other tools like it to help generate the code. Now, you have to be a subject matter expert.

So like going back to your example of food, like you have to know the person. Hey, this person likes Mexican food. They don't like Italian or, you know, whatever it is.

You go down through the list and maybe it gives you a better recommendation than if you're just going in cold. And the same thing with coding, like you need to know what's outputting is the right context, is it the right information? It's probably not going to give it to you in a perfect silver platter.

You're going to have to go through and edit it. But going through and having a generate a significant amount of code could save you a ton of time. That's my guess.

I just think it's like not even the prompt as much as knowing what you want. Right. So if I want to say I want to write a website and I'm not going to say, hey, write me a website, you know, write me a copy for my website.

I'm probably going to say, OK, what do I want? I want a website that's, you know, focused on high conversions. Right.

I want a website that's using, you know, modern copywriting formulas. So I might say to it like, hey, act as if you're a copywriting specialist or something, then I might go even deeper and say, act as if you're, you know, Dan Kennedy, the copywriting legend. And then I would ask you, like, do you understand his his frameworks for, you know, headers and stuff?

And then just have that conversation. To me, it seems like common sense. Maybe it's not to the masses, but I think it stems from like users having to understand what they want.

Now I can see 100%, you know, in your context of like coding. Yes, you have to have some background knowledge and like understand how to ask follow up questions and like, you know, rearrange things to get what you want. But like, I think this notion of just like, I can't fathom that this is going to become a mass scale thing in terms of like prompt engineering, because you could just ask chat GPT to help write the prompt, which we're starting to see now.

Currently. Yeah. And I think that it depends on how you're using it.

So if you're talking about it from a general user perspective, first of all, I do think there is a skill in Googling as well as in chat, if that's a word, probably will be if it's not already. Um, but you know, cause you got, you got to know how to do that dialogue. Cause I know people and I'm sure you do too, that you're like, how could you not find this?

Like seriously. Um, but, but that's why there's a site. Have you ever seen the site?

Uh, let me Google that for you. No, I haven't. It's awesome.

It's called let me Google that for you.com. And so what you do is you go there, you type in the search result and then it'll give you a link and then you send the link to the person that asked you. And when they click on it, it like pulls up Google on the homepage.

And then it's like typing in whatever it is that they just ask you and then it pulls up the results for them. So that's awesome. Yeah, that's great.

But you know, on this prompt engineering thing, so I think it's more of a data science thing. Cause, cause I agree with that. I agree.

Cause yeah. Cause if you look at it from like the data, yeah. If you look at it, okay.

So just give me a website, whatever it was, but think about it as a company. Right? So if you're a company and you have a couple of employees and you're going to onboard somebody, right?

So an onboarded person, they're gonna have a bunch of questions about the company. You know, what are my HR policies? Where do I find the 401k stuff?

Where do I, who do I go talk to about this particular topic? Who's the, who's the SME on that subject matter expert, or who, who do I, you know, put my time off request to all these general questions, there is no reason that couldn't be laid or provided in some sort of context, I'm sure it's written down somewhere and even the best organized companies, it's hard to find this information, but if you just put it all into some sort of database and then you were able to query that database and more of a general kind of way where the prompt engineer could potentially make sure that the prompts that a general person could put in will give them the right results. That's gold.

I mean, think of the time savings there. That to that point though, that to me gets a little bit away from prompt engineering and more into, uh, and that use case, cause you're in a container here working with a very specific set of data, right? Company, you know, their, all their data.

To me, that just seems more along the lines of like an algorithm, machine learning, et cetera, right. More so than a prompt, because if, if, uh, and we have a guest that we're going to bring on soon, Jamison Rots, who he can really enlighten on this, because I mean, he knows way more than we do, but it's interesting. I think it's interesting.

I don't know where these salaries are going to come from. I mean, it's like, at what point today, I think it would be a good time to. will be really in the backend building it for like maybe internal data sets.

But like, I can't see companies at, at scale going out here now saying we need prompt that a prompt that is incredibly important. I do agree, but like, I just can't see companies hiring people for this. When the whole premise of what they're using, meaning like open AI and chat GPT is constantly learning and evolving and can essentially write the prompts for you.

And it's just, I don't know the whole thing. It's, it's interesting. If there are no right or wrong answers, I'm just kind of fascinated by this statistic.

Yeah. I think it's a great statistic and you know, it is kind of interesting and that's kind of the whole point of this podcast, right? We can kind of get two different points of view on it.

Cause, cause I can completely see the salary ranges of this. Now, will everybody hire these kinds of people? No.

Yeah. Um, but maybe there's, what type of companies, what type of companies you think would hire this type of person? The logical thing is any kind of big company, any kind of company that has, um, maybe more than a hundred employees could use this internally from an onboarding perspective.

And then if you look at anybody that does any kind of, um, there is doing anything with data and they, people need to extract that data in some way, shape or form, which pretty much be anybody, whether it be business results or production results or financial results or whatever it is, I think all these things are good and you're, you're right. I mean, it's as much as data collection and making sure the data is available. As it is, you know, doing the prompt engineering, but I think the prompt engineering position is so important because having the data available as part of a model.

Yeah, that's just piece of it, but making sure that if you query it in a particular kind of way, like a normal person would, and not thinking necessarily as hierarchical point of view, and you actually get the right result. Okay. Well, that's, that's where the real value of that person comes in.

And if you can eliminate a ton of phone calls, you can increase the productivity of a company exponentially. Yeah. Yeah.

That's, that's where our buddy Jamison Rots at Nearly Human AI is going to come on and he's, he's working on that for companies, so it's going to be interesting to hear his perspective and more details. Cause you know, I know what he's working on, but not to the extent. So I think he's going to answer a lot of these questions.

He's he's knows so much, man. He starts talking and then it's like, what's going on? Yeah.

Yeah. We'll try to keep them at a general level, but yeah, it's, I think it's fascinating, you know, there are apps out there that do it, so I know a lot of people probably use the Chrome extension. I'm going to I flip over and I just use the chat GPT, you know, I'm so it's called, um, we got it policies.

So my company computer, I have a hard time installing some of these Chrome extensions. So, okay. It's called, I don't, I don't know how to pronounce this, but it's AI P R M.

And it's essentially chat GPT prompts. And there currently are 2,500 of these in the database. And so they work really well.

And it's a Chrome extension. And then when you open chat GPT, what it does is you can just search anything you want and it's pre-written prompts already. And some of them are pretty incredible.

They're, they're really detailed. Now that's a great starting point because you can start there and then ask those follow-up questions. Uh, but yeah, I don't know.

That's pretty incredible to me. These salaries, I'm not surprised where I actually think the bigger, more alarming issue is going to be is, you know, and this is kind of in like the business, you know, corporate world is like how companies create content. And I'm going to give you a couple of examples here where I think is happening is that tools like mid journey are, you know, they're creating some of the best graphics and it's not going to take them that long to, to further develop that.

So I think companies that are having, you know, I don't think graphic designers are ever going to be replaced, but I certainly think there's going to be a more of a use case now for those types of roles that are like almost like in that thin layer on like, you know, the fence of saying, this is what the goal is. You know, here's on the other side, we've got, you know, all this technology. I'm the one that understands how to get them that image pretty quickly.

And we can do it at scale though. I think that's going to be one case that we're going to start seeing. I think the more interesting thing though, that, that this technology is really opening up and I'm seeing this all the time now is that like, let's be real 99% of companies content is self-serving and boring.

Right. So I think this opens up the gate and gives them the opportunity to really not only produce content at scale, but also this is kind of that lifeline. A lot of companies have needed in terms of, okay, well, maybe you didn't have the resources before, or maybe, you know, you didn't have the creativity to think about this.

This, this blows everything out of the water now, because it's so easy to have somebody go out there and do tons of research and find, you know, different headlines, different talking points, different sub niches of things and craft, you know, a content calendar and start putting out content where the real challenge is going to come. And this is where most companies I think will fail unless, you know, they're really embracing this is just in the delivery of the content, right? Because my LinkedIn newsfeed is just flooded right now with just garbage.

It should be called LinkedIn GPT. That's what they should name the site too. Cause it's either post about chat GPT or, and you know, the irony here is we're talking about on the podcast, but it's just completely AI generated, by the way.

This will probably be in a LinkedIn newsfeed. So if I had any credibility, it's not the window, but the idea here though, is that it's, it's never been easier to stand out online because you have the ideas, but I think where we're hitting the point now is for originality. And I think we're kind of getting in this space where, you know, it's always been human to human, but I think now we're seeing it more than ever.

And this just kind of accelerated it. You know, when COVID first hit, we started to see this evolution of companies stepping out saying, okay, we've got to try different things because the newsfeeds are getting more crowded because everyone's locked at home on these digital platforms. So a lot of companies that had success before with like company pages and like traditional posts, those are now getting buried because of the algorithms.

And now content really needs to be educational. It needs to be entertaining or it needs to be really aspiring. Right.

And where. Kind of what I'm seeing is that we're on this fine line now, like edutainment, it's working people that are educating, but in like a unique and creative style, but even that gets burned out pretty quick. And so where I think companies are going to stand out is leveraging this AI, you know, to get ideas and help it accelerate the process, you know, and that, that can be whole downstream of this content creation too, that could be getting the transcripts that could be, you know, telling you how to repurpose it, help and write some, you know, summaries and snippets and all that stuff, but where the value is going to come.

And, and again, this is my intuition speaking here, and I think we'll see this in the next six months is that those in the, in the corporate space, or, you know, if you own a business and are listening to this, that want to win, I think it's going to come down to doubling down and going all in on your originality and creating content that is, you know, original and creative, that's not necessarily, you know, trend jacking or hopping on trends or, you know, things of that nature, but it's more, I'm thinking more like, I don't want to say cinematic, but more like if you go to YouTube and watch some of these YouTubers, they have their own feel and their style of how they're presenting this content. I think that's what we're going to start seeing standing out a little bit more in the news feeds.

And the caveat of why that works is because it actually takes time and it takes creativity.

And what we're going to start seeing then is when the masses are on these platforms like LinkedIn, for example, and every post in the news feed is some regurgitated chat GPT prompt, that's, you know, trying to educate their market. What's going to stand out is those that actually have a personality and a heartbeat, whether it's an individual or whether it's a company that's actually creating that and sharing the story and things of that nature. So that's what I'm doubling down on.

You know, I hear you say that, and I actually completely agree with you. I completely agree, but at the same time, I don't think anything's changed. I just think the bar is moved.

So while I look at it and I think about content, right. And you can think about LinkedIn in particular, corporate content is always struggled. Why?

Because it's typically not that exciting. Okay. Maybe if you're a super, why do you think it's not exciting?

It's not, it's not exciting because a lot of, a lot of the companies post a lot of companies posts are 100% self-serving. So unless you're super interested in a particular product, like, look, if I'm following Apple and Apple's just giving me stuff about their product, I'm good. Because I really want to know about that, but the majority of companies, it's not really what you're looking for.

Right. Yeah. Because it's because it's the delivery.

Yes. So I do think these tools will make that be a little bit more edutainment or, or trendy or whatever else, because it's very easy to create that it's another known. So you can go from, you know, your subject matter really, really, really well.

How do you apply that to the, to other areas, these kinds of things? Well, that's application and that's a little different. You know, I think these AI tools will, will just like you said, you can go type in a prompt, Hey, how does this particular product work in industry?

Or what is a big problem in this particular, and then you just copy and paste that, and that gives you, it makes you appear as a subject matter expert. And I think that that line has now moved to what you said, where there is an element where you can still use these AI tools as a subject matter expert to even further you even more. And that kind of goes back to what I was saying originally, where I use the AI tools, I use them as a, as a marketing professional, and I also use them in generating stuff for my posts, but I'm not going there and saying, Hey, write me an article on XYZ.

I'm using it for research or, you know, what are the big topics or whatever else? Or if I'm in a brainstorming meeting, it is fantastic. Fantastic.

Yeah. To have a neutral party sitting there right next to me that I can literally ask anything to, and it'll feed me information back. And, uh, and I use mid journey in the same way, if I'm doing creative design, like, Hey, I really want a logo that does blah, blah, blah.

You know, it's really, really easy. nice even if it's not exactly what you want, it can help your brain wander in different ways from a creativity standpoint. So I am a big one on brainstorming.

I like to have those kinds of meetings and try to fulfill these objectives. But I do think, to your point, the people that can really leverage this tool to enhance their content game, to enhance their own knowledge, and really, frankly, what things to research, it's gold. But it's a slippery slope right now, man.

It is. It's scary. I think we need a little Sheba in you next to valid posts.

Sorry, another trend topic. We do. Look, if Elon buys LinkedIn, that's going to be the new logo.

That would be Doge, right? Yeah, yeah. It's true.

Yeah, Doge. So I kind of look at it as maybe two phases, because what I heard you say and how I understood it is that companies now at their arsenal have the tool for unlimited research. Now, whether they use it or not, that's going to be a whole different story.

I imagine most won't, because they're so stuck on that this is what we have to put out versus what you said is that they could use this to add more relevant context. And so that's going to help them, because the more relevant something is and the more it speaks to an audience, whether the company thinks that that's what the audience wants or not, if that's what the audience wants, you got to give it to them. Right.

Where I think, though, that that'll have a very short lifespan is that every other company is going to start doing that. So now you still have the problem of distribution. And in this context, it's how do you get that post that is very specific and relevant because you've done the research, you know, your audience cares about it.

How do you get that to stand out on the platform when you've got individual creators out here just blowing you away with their content? And even if you look at the carousels, we're seeing all these people now have to put more effort into them in terms of like the look, the feel, because you've got to stand out in the newsfeed. So that's where I really think it's going to come back to not the content, the context will be there.

It's going to help companies rein that in and get a little bit more relevancy with their audience. But you've got to present it in a manner that's entertaining or interesting or has a unique flair to it that doesn't look like all the other content that you see in your newsfeed. Like they have to be able to say when they see you, oh, I love his look, his feel, his style.

And I think that's where we're going to get to. And I just think it's never been easier for companies right now, but it's never been harder because it's a big shift that they have to make and they have to invest in the resources and the mindset. And I see a lot of companies just over-investing in the wrong skill sets for where things are.

Yeah, it is interesting. And I know we're running long on this podcast. We could probably talk about this for quite a bit longer, but I have a decent following on LinkedIn, right?

And I am a smeeh in particularly areas of like industrial and IOT, but everything you're saying is the things I should be doing. And I guess I do do some of these things, but I guess I post for the wrong reasons just in general. Because I post because I'm researching a particular topic or I'm learning a particular thing.

And I like to add my context. If people like to read it, awesome. If they don't, I don't mind it.

But I would argue that that is what makes you stand out in the newsfeed because your posts are always out there. And I know when I... Here's what I know.

I know when I see Ira's post, I expect... I like the fact that I don't know what I'm going to get, but I do know what I'm going to get. I don't know what the topic's going to be, but I know I look at Ira as someone who's very curious and he's learning and you're not building in public.

You're learning in public as you go. Where there's others who I know when I see their content, oh, he's the carousel guy. Okay.

I've seen two or three of these. I know that there's not much value there unless the topic's interesting. But when I see your content, I'm like, oh, this is interesting because I know how you present it.

And if you look at YouTubers, there's a ton of these. There's YouTubers I follow that I don't even like the content. I like their style.

I like how they present things and so forth. And so I would argue that you're actually... You have a great case of that.

I think for you, it's just a matter of... And this podcast is going to help with that, doubling down and really enhancing your voice to reach more people now. Yeah.

And again, you know me. I like to talk, like to have conversations. I like to explore and I like to learn, right?

Because while I think I'm right about everything, trust me, I'm not. I'm not a lot. You remind me of that all the time.

So it's okay. We'll wrap this one up, but I guess we'll... Let's make up a new segment here on the fly.

Okay. So this is... Let me get your hot take on this.

Okay. Do you feel that... We'll go...

We'll really generalize this. In general, right now, do you think chat GPT is a shiny object or do you think it's actually something that people are sleeping on despite the buzz it has in the world right now? Do you think it's overkill or do you think it's...

Do you really think that people are still sleeping on it even though there's all this noise? This is a good one. So I 100% think that people are sleeping on it.

I don't think it's a shiny object. I think it is a shiny object that is actually shinier than people think. Do you think companies should be having emergency meetings about this right now?

Yes. I mean, frankly. Do you think they should be opening up budgets to find ways to either...

If they're even... They got to be allocated towards it? If you're considering any kind of digital strategy, anything, whether it be content, whether it be internal, committed to be whatever, I think you need to strongly be considering this and researching it and making it part of your plan if you didn't yesterday to make it today.

It's just that big. And... And what would you name that line item on the budget?

ChatGPT? Because ChatGPT is just an example. It's more of the idea and the concept of it.

So whether you're using... There's just so many models out there. And there's going to be many more to come.

But natural language processing in some way, shape, or form that you can have a conversation and get valuable information out. How is that not at the top of your list? Because of the productivity gains that you can get from it.

So how do you teach it? How do you utilize it? How do you integrate it?

If you're concerned about privacy and these other things, I get it. Figure out how to do it in private ways. And I'm sure there'll be lots of things coming down the pike.

But I think that companies should definitely be having strong meetings about this, emergency meetings. And I would consider it the future of the company as far as HR and data. Yeah, I agree.

So we'll end on a follow-up question. And all I need is just one word answer. I don't think I can do one word answers.

Two words, actually. It has to be two words at the middle, I guess. When we look at this, and this keeps me up at night, this is random, but I think about this a lot.

What do you think is the window of opportunity right now for companies to hop on this and really adapt? Because the window's closing. Every day, more and more things are coming out.

You've got that first mover advantage. And yeah, it might not look the same in six months, but there is an opportunity right now that's very small that the right people and the right companies can hop on and make millions, whether it's in revenue or pipeline, whatever, just by making those things before the competitors do. And then it becomes a, oh, we're going to hop in this as well, type of a strategy.

What do you think that is? One word. How long?

Oh, okay. Well, I can't give you a word. All right.

You're five words on it. Yeah, I can't give you one word. I told you I can't give you one word.

Because, well, here's the thing. Here's the thing you just asked, right? So how long?

So I think if you're going to be a consultant or something like this, I think you need to move moderately fast. And I know that's kind of goes against what I just said, but look, companies got to figure out- What's moderately fast mean? We talking 30 days, 60 days, six months?

I think six months to a year. Yeah, I do. You think that long, I think 60 days.

I think 60 days. But see, here's the thing. Who's going to integrate it in 60 days?

So you need the companies to have, that aren't going to be developing. Well, I'm going through the consultant, the individual- They need to get up to speed and things, but they're not going to get projects for six to 12 months. I mean, companies don't move fast.

So you're looking at enterprise- Exactly. More layers of tape and division. Yeah.

If you're looking at larger enterprise, the people that can really leverage this, they're not moving fast, right? So they need to be having meetings now so they can get a decision in six to 12 months so they can hire the consultant in 13 months. I know it sounds like a long time.

And in this world, it is a long time, but- That's like dog years. I get it. But there's a reason the government has websites from the 90s and the 2000s, right?

It just takes a long time to move a big ship. And it's not saying necessarily it's a bad thing. It just is a thing.

So that's my two cents. Yeah. Hopefully this will speed it up.

I think this is going to really change a lot of companies. Maybe that's a question we should query. We should go, should large enterprise consider chatGBT and say it in some sort of weird voice and see what it says.

Yeah. Yeah. Yeah.

We'll talk about that next week. And then yeah, we'll have Jameson on maybe next week or the week after. Hopefully he can answer more of this.

But yeah, otherwise, man, great catching up with you today. Good episode. A lot to think about, man.

As always, man. Appreciate it. I love this conversation.

I love the general nature of it. So yeah, let's keep on doing this. All right, man.

Awesome. Chat soon.