11: Using AI for Effective Prompt Engineering with Edward Frank Morris

On today's episode, we're joined by prompt engineer extraordinaire, Edward Frank Morris. Edward and I met through LinkedIn, where his detailed posts related to prompting excellence caused me to reach out to learn more about how to work with him. He's recently achieved a rare recognition from LinkedIn when they designated him an official LinkedIn Top Voice on the subject of prompt engineering. On this episode, Edward shares his thinking on what it takes to get the highest quality output from models like Chat, GPT, Midjourney and others. And after we recorded the episode, he was inspired to create a guide on how to get hired as a Prompt engineer and offer it to the listeners for free. Check out the show notes for the link. Thanks, Edward. Now let's get started. Welcome to Using AI at Work. I'm your host, Chris Daigle. Each week we'll be learning how today's business owners, entrepreneurs and ambitious professionals are getting more done with smart use of tomorrow's tech. Let's get started. So I'm going to let him introduce himself, but our guest today is Edward Frank Morrison. And I just want to encourage everybody that if you're not reaching out to others on social, you're missing out. That's how Edward and I connected. I saw the type of content that he was putting out on LinkedIn. In particular, wildly impressed, reached out to him, and now here we are. You know, after several calls and emails, here we are, we've got a guest. Fantastic guest for everybody today. So, Edward, why don't you take a moment and just tell everybody a little bit about yourself.

Oh, that's a good question. Why do I even begin with myself? So obviously we're on an AI podcast, but I have a degree in journalism, so it's a little bit strange. And I get this quite a lot. And my background is actually in communication, so I started off as a journalist. I did a lot of social media marketing, and then I did quite a lot of copywriting as well. And I've always kind of had a foot in AI, so I've always kind of been involved in it in some way or another because it's always been something that's interested me from an ethical standpoint. So to me, I kind of like talked about AI from the perspective of, like, this is good, this is bad, here's what we can improve, here's how we can implement this. And it just so happened that, you know, around two years ago, it's more like a year and a half or even less, ChatGPT came onto the scene and everyone started using it. And then we started realizing the importance of these GPTs and these kind of transformational models, and so I kind of jumped in. And ever since then, it's just been a bit of an upward rise, really. I still do a lot of AI consulting, but now I'm one of the only prompt engineers in the UK who's actually done it. You know, who's not just said, here's how you could become a prompt engineer and here's how you can work in AI. Like, I've actually gone out there and wrestled the beast and made it mine. So, yeah, that's kind of my background. I've been featured in Forbes, been on Yahoo News, Yahoo Finance. Worked with some. I can't say them by name, but some massive companies. And I've worked with one company who's recently been in the news. Can't talk about what I've done for them either. But.

LinkedIn Top Voice, of course, that's a recent accolade.

That too, yeah. I was the very first LinkedIn top voice, specifically for prompt engineering. And yeah, it's. It's just been a crazy ride, really. It's weird because I don't think I'd be growing this fast, but here we are. You know, I'm on the podcast with you and. Yeah, it's how we met, man. Like LinkedIn. LinkedIn. Who'd have thought?

You know, it's interesting that you bring that up unintentionally. Some of our past guests there, there's been a theme of them having copywriting skills. Right. And one of the. And I didn't realize that about your background, but it makes perfect sense because one of the connections that I've made is that like copywriters, sales copywriters, they spend a lot of time thinking about specific words and how to use those words to convey a meaning, a message, to elicit an emotion. So they're. They pay more attention to the language than most of us do, who are just, you know, casual talkers. So I. Can you address how you think that that may have had some impact on. Combined with a journalism degree, which, again, with that degree, you're studying how do words influence thought? And with all of those things, it seems like because most people, they think about Prompt Engineer, or prior to today, they would think a prompt engineer is the Python coder who's getting in there and doing a lot, which. That is one definition of a prompt engineer. But those guys don't have the mastery of the language. Right. They understand the engineering side, but not the communication side. So can you kind of talk about, like, how that has given you an advantage maybe over somebody who just had the, the, the technical understanding of prompt engineering?

Yeah, for sure. Yeah, it's, it's quite interesting really, because I think a lot of the time we, we look at AI, as you've rightly said, from like a coding and Python kind of perspective, which is fair enough, because that's all we've ever really known about AI for the past decade, really, because you need to have that kind of knowledge, mathematics and in order to make an AI tool. But I don't remember who said it specifically on, on X, but someone said that communication in English is now the top coding skill. It's like the top programming language. And the reason why it's so important is because with these large language models, you have to communicate to the large language model what you want it to do, you know, and one of the biggest tasks for any copywriter in the world is to get a person to commit to an action. That's the end goal for everything. It doesn't necessarily have to be for sales. It can be for clicking a button, it can be for subscribing to a newsletter. It can be anything, really. But the ability to command and get action out of something or someone, in this case is the most important skill to have as a copywriter. And so when ChatGPT came onto the scene and we had all these people who are fantastic at like, guiding people on how to do things and making them do things, it seems like it was a match made in heaven. Because now those direct language skills can transpose themselves into ChatGPT. Because essentially what Chat GPT is trying to do is it's trying to understand what you're ordering it to do. So if you can do that in either a very precise and a concise way, or if you can do it in a way that gets the AI to interact in a particular kind of way with its own subset of data, you tend to get very good results. And so, like, just to wrap the point up, when Sam Altman came out with ChatGPT and then all these prompt engineers started rising to the surface, and Sam rightly said prompt engineering isn't like a role that is going to be out there forever. I think he was partially right. I think prompt engineering as it stands at the moment is incredibly important for a lot of companies. But I think what it is, is for the average person, prompt engineering won't be as important to them. But if you are a business that needs AI to do very specific kind of things and you need someone to whisper to the AI to do specific things. If run of the mill isn't good enough for you, you will need a prompt engineer and that's where the communication skills come in. Because you can get the best coders in the world to write something and get a large language model off the ground. But if you don't know someone who can talk to it and get it to do the things you want, you have a load of data, but you can't use the data on the way you want. At least you can't charm it. It's like having a. A snake, but you just don't have anyone to play the flute and get it to dance. Yeah, you know. Yeah, exactly. So that's how I see at least.

So the average you a user or listener of the podcast, their demands for it aren't going to be. They're going to have a wider tolerance of. Of hallucination or, or a rewrite the stuff that you're working on though you're suggesting that the. The output needs to be extremely precise because so many downstream activities or upstream activities are reliant on the quality of the output that you're eliciting from the models. Is that right? We'll be back after a quick break.

Do you want to start your own unsexy business offering a local service or a trade to your local community? Come check out the Owner Ops podcast. We bring on people like Sam, who has built a $22 billion roofing company in five years. I've also brought on people like Brock, who generated $1.3 million of revenue in his first seven months installing septic systems. If you are looking to build your own local service or trade business, come check out the Owner Ops podcast. That's ownerops.com O W N R ops.com.

Yeah, I think it depends entirely on. On the user case because I think for most average people, like even non average, even executives and people along those who have very specific intentions for ChatGPT, I think a lot of them, although they would benefit from taking a prompt engineering course or something along those lines, I think a lot of them are quite content to saying rewrite this email in a professional tone and then it does the thing. But when you're let's I haven't worked with them so I can actually mention a name. Let's say you're blackrock for instance. You may want a GPT and you may want a large language model that can take 50 pages worth of research data and it spits it out in a very particular kind of format where paragraph three is longer than paragraph one or whatever. And that's where Prompt Engineer really needs to come in. So on one hand, if you're like a normal person, you'll find benefit in large language models. But if you are a company that can see all of this data being formatted, or even if it's just technical writing being formatted in a particular kind of way, and you're like, that's going to save us of time. You're going to see prompt engineering a completely different light to the average person.

So when you were talking about how Sam Walton said that people didn't really need to learn how to prompt, a couple of takeaways from that concept was one, most people can get that synthetic output that they're looking for just from brute forcing, you know, beating the keyboard. Right. They don't need to be precise. However, the type of stuff that you.

Do.

You know, is, is there some feedback or are there tools that you're using to improve prompts? And, and let me, let me get at the context. So if somebody is new to this and they think, oh, all I have to do is just treat it like Google, right. Like, let me ask something, right? That's you. And I know that that's, that that's a weak prompt. It can be a lot more.

Yeah.

Effective than that. Are you using any tools or have you heard of any tools that an individual who's not good at it can use to rewrite their prompt in a way that would get them a better output?

You know, it's interesting because I do see the tools like this. You know, I do see prompt perfect. I do see extensions you can add on to Google. And what it does, it essentially beefs up your prompt. But I think ultimately I don't use any, any kind of tools whatsoever. That's the big thing. And I do anything, you know, I do anything from simple, kind of make me an email with 500 words and make it sound in a particular kind of way all the way to seriously complex prompts that require multiple chain entries. But I've never really seen the benefit of using a prompt perfect kind of thing from my perspective, because of prompt engineering. I mean, from the perspective of like the average person, I think the best thing for them to do when they learn prompt engineering or when they try to get into the world of large language models is to also work on their communication skills. Because ultimately, if you want your AI to work in a particular kind of way, you have to communicate that like It's a human being because it is ultimately meant to be mimicking a human being. Right. That's an artificial intelligence. So if you're, if you're treating it like Google, you're using 5% of its capacity, if that, and you're getting 5% results. So it's just like what every other AI influencer says on LinkedIn, which is garbage in, garbage out. So you have to be specific with it if you want specific results. Diamond in, diamond out. Right. So if you want a 1,000 word paragraph in, written in three line stanzas and it's written in a Harvard Business style, you could do that. But you have to tell the AI that because it still can't read your mind. You have to be specific. And I think a lot of the time when people get a little bit disenchanted with AI or they're like, we're not quite there yet, we're not there, it's terrible. ChatGPT is awful at doing stuff. I think there are a few reasons why they think that, and I think it's because they pick it up and they give it something impossible to do or they misunderstand the system completely, which is where I think education comes in quite a quite heavily. You know, I see all the time on, on Reddit and Twitter or X and LinkedIn and people say, you know, I told ChatGPT to do something like three plus three or two plus two and they couldn't do it. I'm like, well, yeah, because it's a large language model, it's not mathematical, but people have their expectations, you know, they have their misconceptions and it's so easily fixed. You know, it just takes a bit of time to learn, but once you learn it, it's the most powerful tool you can have.

Yeah, that's one of the things that we routinely explained to executives is the limitations of the models. Right. There's these assumptions that it's, it's perfect and if you know how to use it, it can be pretty darn perfect. So for individuals who are now starting to get this concept of hey, I don't need to be technical in order to, to prompt. Well, the advice of learning how to communicate was very good. Are there some specific people that they should follow or are there some specific. How would one, how would a non technical business person who wasn't aware that they weren't communicating well until they, they listen to this, where do they go from here to kind of get the, the hang of talking to the models?

I mean, it's a, it's a great question, right? Because I think it would be very easy for me to say Open AI, Anthropic and other companies that depend on prompt engineering for you to communicate with your system. They have courses and they have material out there that you can use and learn from, you know, But I think one of the big things about AI of this style is that it is a communication thing ultimately. And if, if you're a poor communicator, you're not going to communicate very well with the AI. And so from my perspective, I just see it and I'm like, well, if you don't want to communicate with the AI like an AI and you, or human, then you have to communicate to it as if it's like a, a system, right? So you would then have to bullet point what you want to do. And so instead of you saying, hey Chatgpt, I want you to write me like a 5,000 word essay, whatever, in this style, in this length or in this format, you would then instead open up a notepad and then you think to yourself, well, what do I really want ChatGPT GPT to do? And then you would just start bullet pointing out what you want it to do. You know, dash, write a paragraph about X products in a luxury copywriting style. Another dash below it, write out 7,000 words or 700 words that describe how to use the product. Another dash below that, then at the very bottom put a legal disclaimer that if you use it incorrectly, we don't hold any liability. And so just by bullet pointing what you want this thing to do, you're essentially communicating, you know, as directly as possible. I think there's a bit of a divide because most of my prompts are written in that way because I just want to get the points across to Chat GPT. That way if I, if there's something wrong, if something goes bad, I can look at the line and go, that's probably what's confusing it. You know, I'm using this word instead of that word. I'm telling it, here's an example. But actually I'm not using the word example. I'm saying, for instance, they're not adding quotation marks or something. So it's just a case of, you know, making sure you can actually communicate. And I don't think any prompt engineering course out there can actually do that for you. I think it's, it's very difficult to pick up any course, even with the best intentions and say, you know what, this is going to teach me everything. There is about prompt engineering, when inherently prompt engineering is a communication skill, which is why there are so many writers out there who can spot Chat GPT written content from a mile away, because they've seen it a thousand times and they're like, ah, you're communicating using this exact pattern that Chat GPT users. And we can tell, you know, I can tell if you show me a paragraph that's Chat GPT generator, I'll be able to tell you. But yeah, that's, that's kind of where I stand with the whole matter. You know, I think if you want to get good at talking to Chat GPT, learn how to talk to people. That's, that's the thing, you know.

So you, you bring up an interesting point. There's plenty of courses out there to learn to learn prompt engineering. However, you can learn how to prompt and some frameworks, but until you're interacting with it and realizing the variation and output based on what you're doing, like, understanding how to prompt is not learning prompt engineering. Learning prompt engineering is going to come from beating up the keyboard, going like, oh, that's not what I meant, that's not what I meant, that's not what I meant. And so this is great advice. So anybody out there who may feel like they haven't gotten the hang of prompting, keep prompting, but couple it with a framework like what Edward was talking about here, where you're very clear, you don't have to be an erudite prompt prompter, you just need to be a good communicator. So that's a great distinction. Now obviously the models are a daily tool for you outside of Chat GPT and Claude and any of those kind of tools. What are, what are some tools that you're getting a lot of value out of that are AI powered, AI driven in your, the operation of your business, not necessarily delivery of your, your value, but in the operation of your business?

Yeah, this is, this is another interesting point. I mean, at the, at the very core it is Chat GPT and generally large language models that my life is surrounded with at the moment. And the reason for that being is because there's so much written content out there. And so the way my business kind of works and the way I work as an individual, I will look at a business and I'll say there are certain areas you can optimize. And it seems that most of my clients, they fall into one or two categories and it's either they want to implement AI into the business and upskill their stuff and how to use ChatGPT efficiently, which is fair enough, because we all want to write stuff, we all want drafts and we all have long letters. We just want to avoid writing altogether. And so that's completely fine. Or alternatively, they'll go down the route of we want our own specific API with Anthropic or chatgpt or whatever it is, and we want it to have one core function and that's a lot more intricate because you have to be good at fine tuning. And at that point, that's where actual prompt engineering starts mattering. Because we talk about prompt engineering a lot and I think there's like multiple levels to prompt engineering, but for the sake of the listeners, just, just to kind of give you an overview. When you enter into the prompt engineering world at that level of the API and stuff, you have to start tinkering with how the AI interacts with your words. Does it want to answer in a 100%, this is as hard, factual as I can get kind of way, or do you want it to be a bit more imaginative and human feeling? And these are some trade offs you have to start making as a prompt engineer. And then there's other things you had to account for, like model drift and all the rest. So from my day to day, it is always chatgpt, it's always large language models, and it's always, you know, these kind of areas. But I think one thing which has saved me a lot of time, even though this is a bit of a cheat, because it still boils down to a large language model, one thing that saved me a lot of time is meeting, recording AI, you know, fireflies, otter AI. This kind of stuff, it's, it's invaluable. You know, there will be stuff which we'll talk about, about on the podcast and I'll think back and I'll think, did I really say that? Like, can I play that back? And sure, I could listen to the whole podcast, but it'd be far easier if I could just hit control F and then search through the meeting files. Right? It'd be way easier. So these kinds of things are really useful and there's other things coming out which are very, very similar to it, which I'm even considering getting, even though they're a little bit invasive and creepy. Like for instance, AI that records your live yachtings so that when you're in a face to face situation, you can just put the phone down and take all the notes so you can send those notes off to the other party. Right. Really, really useful tech and it's stuff like that that I think is probably the most beneficial to me, especially because meetings seem to be the name of the game these days.

Yeah, I think in modern business, the stat's crazy. I think that. That the average individual in a. You're entrepreneurial, I'm entrepreneurial, it's a little different. But for the person that does the nine to five and shows up to the office, I think they're spending like a third of their week in meetings. And if you can more intelligently conduct these meetings or document these meetings, like, exactly the tool that you're talking about. And here's what's crazy, dude. We built that tool for internal use four months ago. So we've been using that exact thing. So not as creepy because we were all, like, participating in the development of it, but I can see the wearables that those might be. I don't know. Yeah, like, that kind of stuff is a little.

It's insane. Who's the company? I can't remember the company. Who wants to do it. It's. It's like a necklace that goes around your neck and it, like, records everything. And I remember the company and I reached out to them, was like, hey, is this thing, like, always recording? And they're like, no, Ed, don't be ridiculous. It's not always recording. It's just when you press the button, it's like, oh, okay. So it is always recording. There's gonna be one guy who's like, record everything.

Just like Siri and Alexa aren't always listening. Yeah, you got to give up all privacy. So how did you go from journalism and copyrighting into. Obviously you'd been. You'd been an AI enthusiast prior to the big release in November of 2023, but what does that transition look like? Because we've got people out here, maybe we've got copywriters listening to this for sure, who would like to. Even though they're well paid as sales copywriters, the application of that skill into an AI enabled environment raises their value, their, like, commercial value considerably in the marketplace. So for individuals who are interested in moving more towards identifying as a prompt engineer, what was that path like for you and what advice would you give them now?

I think I'll just get straight to the pitfall. The biggest mistake I ever made in the prompt engineering and just generally the AI journey that I went on is I only started pushing myself in AI and that kind of field properly, I think in around June, July. Ish. So quite late into the hype cycle. Of ChatGPT at the time. Like before then, I was just lurking on Reddit and commenting on posts and seeing what people are making and doing. And that was fine, I was just upping my knowledge. But the thing is I, I was already using ChatGPT far more than other people were at that point. Like by. I got access in November of 22. And so by that point I had used it way more than the rest of the world had. And so I think the big thing, this is the biggest pitfall. If I could turn back time, this is why it would change. I would just go straight in and say, right, we're going to talk about AI on LinkedIn from the very start. I might be a newbie, but every piece of new information I come across I'm just going to share and people might benefit from it because it's such new technology. Probably my biggest regret, because if I did that, I'd probably be up there with like the six figure followers. Although it's like a vanity metric, it does make a big difference when it comes to trying to get clients, jobs and contracts. Yeah, it's an authority thing more than anything. And so that would probably be my biggest thing. So if you are considering going from the copywriting route to the AI route, you know, go out there and do your thing, but from the actual transition side of things, that's a quite interesting question because again, my foot's always kind of been in the AI sphere. It's always kind of been there. But going from journalism to copywriting was a purely monetary and a moral kind of thing for me. Because in journalism I kind of looked at the world and I was like, riyadh, like, here's the world, right? And how do you make money with journalism? You report on the truth or what you consider to be the truth. We can debate about that if you want, but I mean, that's how it's done, right? Like you, you find a sad story, you write about the sad story, you promote the sad story, hopefully gets a few clicks, you get a bit of ad revenue and you've profited off of someone's misery. Now I thought to myself, that's not how I want to live, right? I can't live like that because at some point I'm going to be pushing out like six of these stories every single day in a 9 to 5 office and I'm going to get so desensitized that at some point in my life I'm going to look at a sad story or something tragic that has happened and I'll think, oh, that won't get me that many clicks. I'm not going to make that much money off of it. And, and I didn't want that, I didn't want that on my side. And so I said, where else can I go? You know, what else can take me on or where can I take these copy while these writing skills. And it led me to social media management. And social media management at the time wasn't a super, super popular thing, but it wasn't very writing focused. It was always kind of like, let's make a viral like video clip and see how good it does. And eventually the market just became too saturated and thought there's no point competing because there's some person halfway across the world who can charge like, you know, 300 pound an hour or 300 pound a month rather, and that's their monthly salary. And in Meanwhile, in the UK you need about 1.6 thousand pounds to live per month and that's on a minimum wage. And I'm like, I can't compete with 300 pound a month, like at all. Like I can't. And so I said, okay, well, I've got marketing, I've got writing. Just transition into just pure writing. Let's just do some features, some blogs, some product description, you know, let's go down that avenue. And that works out really well for me. Yeah, so I just kept on copywriting and just kept on doing my thing. And I think the initial transition from journalism to copywriting was actually the hardest part because for me, I'm now taking the skills that I have and had where I would tell the news quite factually and then flip it onto its head and say, instead of me just telling the news, I have to describe the news and I have to make the news. Yeah, it changes everything. So when ChatGPT came out and I saw how people were using it and I thought, this isn't what we thought AI was going to be. Because we all thought that AI was going to be coding and all this stuff. Like, how is this possible? So Twitter was, back then, was flooding with videos and pictures and memes and people were making all sorts of crazy stuff, you know, stuff that these days we would consider mundane. You know, shopping lists, they were making full documents for leme as the whole thing. And we were like, wow, this is insane. It's crazy. Let's see what else we can do with it. So I got Chat GPT. I was very, very lucky because there weren't that many users at the time. So I Got the beta quite early, and then in November, I started using it and I was like, this is insane. Like, I can get it to role play fictional characters. There weren't any safeguards on it back then, so I could get it to emulate a doctor, I could get it to emulate a judge, a lawyer, nurse, all of these high professions. And I was like, even if the output isn't 100% accurate, it's still a proof of concept that at some point we can get something which is so close to accurate that it'll be better than a human being. Then, lo and behold, ChatGPT4 comes out and it says, well, OpenAI says it's passed the bar exam, it's passed the medical exam, it's passed all these exams. And I'm like, this is it, go in, Ed. Like to put all your chips in. This is it. This is the end game. And so I've still got friends who are doing copywriting, and I still do copywriting every now and then, but, like, on a more casual kind of basis. And even they're like, why did you transition? You know, why did you do it? If you're that skilled at copywriting, you should never have moved, you should have had faith. I'm like, you're kidding, right? Like, I can take everything I've ever written, put it into a notepad document, feed this into AI technology that is now a year old, and it will spit out content that sounds like I wrote it word for word, you know, And I think a lot of people don't understand that. Like, sure, it's, it's. It's almost like you're selling yourself out. But if this technology exists, it doesn't really matter what I think, because this thing exists. You know, we can't put, we can't close Pandora's box. It's hoping we know what a large language model can do now. You know, I mean, even if the government of the UK were to come along and say, no more ChatGPT, no more Gemini, no more Anthropic, there's no more large language models. There are millions of people in the uk, just like myself, who have downloaded offline large language models like Mixtral and Llama, and we can get it to run on our PCs, the people who possess those things are superhuman because if they're generating work in seconds, I mean, how can a copywriter, a hard copywriter, compete with that? They just can't. And so I was like, you either pivot now, Ed, or you're going to get left behind. And so I pivoted. I've never looked back. Since. Prompt engineering is something which you have to keep on playing with, you learn the basics. You can start with a few frameworks if you want. I think the most popular framework at the moment is the smart framework. But once you've graduated from that framework, you then just like copywriting, you start doing your own thing. You know, there's loads and loads of copywriting frameworks. There's a pain aggravate solution. And most copywriters, yeah, yeah, they start, they start with, they start with that. But then as they get older into their copywriting journey, they start to realize actually the frameworks like training wheels, we need to let go of them so that we can grow. And then once you let go, you, then you start growing. Just like prompt engineering, you can start with some frameworks. Once the frameworks have done their purpose and you know how it works, spread your wings, you know, go crazy. That's, that's, that's it.

You mentioned a pretty commonly encountered scenario for people who are using the models to create content. It doesn't sound like me. I can't get it to sound like me. That's a common refrain that I hear. And you, the example, you have the benefit of a body of work that you can train the models on. Right. Like you've been writing. So you can say, this is how I write. Here's 5 years, 10 years worth of work. Do it like that. The average person doesn't.

Yeah.

Do you have any. Although this is, this is. Everybody wants the models to sound more like them. How do I. Are there any prompting techniques or any, any frameworks that you found that are particularly good at helping an individual create a closer clone of their voice for the output?

Good question. I think this is going to be a little bit strange because I'll get to the large language model aspect in a second. But with midjourney, when people try to generate their own pictures, I think a lot of people mess up. And the reason why people mess up with Mid Journey, and for the listeners who aren't quite aware, Midjourney is for image generation rather than text generation. I think what a lot of people tend to do is they go into Mid Journey and they'll say, I want a picture of a man wearing a black tuxedo, nice hair, sunglasses, and it's their idyllic version of themselves. But in order for you to get something that even closely resembles you, you have one of two options. And your two options are one, you give Midjourney a picture of yourself and you'll reference the picture, but it won't be perfect. Or two, you take a long hard look in the mirror and you take note of your face. You know, what is it round? Is it triangular? Diamond shape? Are your eyes almond shaped? Are they, you know, wide? Are they small, Are they big? You know, do you have a tomato nose? Like all sorts of things, you just be as meticulous as possible. And then you feed into mid journey and then mid journey will spit out something which is closer to you. It will still be 90% of the way there, but when you finally change it over and you do a face swap with the newly generated image, you then get something which looks pretty scarily accurate to use, like a 98% thing. And you can go farther, you can get close to 99. Now the reason I mention it is because copywriting and getting large language models to behave in the exact same way is no different. So you have to be very aware of how you write as a person. Like all of us have habits. So some people, they'll start all their messages off with an introduction line or a joke. Some people start off with just a straight up nice to hear your voice after so long and this kind of stuff, or a compliment, whatever it might be. But we all have our habits. And I think when you're trying to get a large language model to act like yourself again, you have two options. You can do it the easy way, which is taking all the content you've ever written or tweets or LinkedIn posts or Facebook posts, I guess, and you, you plug it into a notepad file, you just copy and paste everything you've ever written and then you feed it into chat GPT and you tell it to use as a reference point and you tell the model, write the post like I would. And here's the reference point and it will. I've done it many, many times, not just for myself, but for copywriters. Like there are many luxury brands which I thought, I really like their copywriting style, let's see if I can emulate it. And I just take the copy from their site and feed into a notepad file. And yeah, it does the job, but there's always that 5%. And then you have to make those edits and then you have to make sure that it actually fits who you are because there'll be certain things where you'll go, that's very close to what I would say, but I'm just going to erase that line or edit that line. But it's a 5% change that you make that will make it really sound like you. And I'm sure there are going to be people listening and they're going to think, well that's not good because the beef you still have to edit it. It's not 100% there. But keep in mind that this AI, this AI can generate, you know, 2,500 words in 30 seconds and we're asking you to edit a paper for like two minutes. I think, I think it saved you enough time, you know. And so yeah, yeah, that's, that's how I view it. You know, you can get things very close.

So this kind of is tying into the, the concept that you shared earlier where in your journalism career you reported on the news. But then in copywriting you're almost like you said, you're describing the news with the mid journey example. Here's a photograph of me versus let me describe me. So this, this concept of, of running the, the, the what you want to communicate through the. How do I describe that? That instead of just saying it or just typing it. Right. It's a very interesting concept and if somebody can make that. Because one of the things that I'm always looking to do with, with guests and clients and that sort of thing is call it get them to start thinking in AI. And what I mean by that is that yeah, if I see you do something that's cool, but I don't know why you did it. It's when I can start understanding why these things are happening, then I begin to start thinking in AI and I become agnostic to tools or to frameworks. It's like you were talking about earlier, the copywriter who starts by using a framework. Once they achieve virtuosity or you know, a level of talent they have, they don't need the frameworks any longer because they've almost internalized how to, how to do this thing. So that thinking in AI, this, this idea of describing it as compared to just communicating with it, I think as a heuristic is a shortcut for people to start thinking in AI. It's awesome takeaway. Thank you. Hey, so where are you getting your news from? I know you, you're an enthusiast, you've early adopter beyond even most of us. Where do you pay attention to what's happening as far as tools, developments, progress, things like that?

Oh, that's a great question. So I usually get most of my news from LinkedIn, but that's like just generally browsing through. I'm not exactly super active on X I, I, I wanted to be, but I just found that growing on X and being your own personal X was just too, too strange. Like I don't know what it is, it's just something, it just doesn't have the same spark as it did before. I mean, I'm subscribed to a couple of newsletters. I'm subscribed to the Rundown, which is pretty good. TLDR tech, that's another good one. And so there are these kinds of newsletters. I've got my own newsletter which I'm currently working on at the moment.

How can people get access to that?

Oh, so it's actually on my LinkedIn page. So if you were to find me on LinkedIn, Edward Frank Morris, you would just scroll down and there's like a quite big banner you can subscribe to, like my, my newsletter. But the difference is, is that my newsletter is a bit more thought leader orientated. So I don't, I don't really push tools which I haven't tried yet. So the problem is in the sphere there's a lot of noise and understandably, because it's a hype kind of cycle at the moment. And so you'll get people who say, oh, I've tried this fantastic new AI tool and it does all these things and it might do, it's just that, that about 85% of these AI tools these days are a wrapper of some kind. And when you're in the prompt engineering sphere, you kind of know how to spot it. It doesn't make the tool bad per se, but when you're paying people influences, specifically money to say, hey, here's like 300 pound to promote my product. You have to wonder if that product is actually good because it is good and because the influencers tried it, or are they saying it because they've been paid off? And it's a quite tricky thing because I'm friends of a lot of these AI influencers. It's a very tight knit circle and they're quite vocal about it. They get some tremendously bad offers. They'll be offered £250 to £400 for a shout out and they've got 10,000 followers on LinkedIn. But they take the offer up even though the actual system itself isn't really anything special, it's just a copywriting tool. But they do it because 400 pound in their currency is a monthly salary. And I'm like, you have to eat. I get it, it, it's a good point. Yeah, it's, it's an integrity thing. So weirdly it boils down to the journalism background again. So when I talk about AI tools 9 out of 10 times because I've tried it or because I have belief in it. So I've been talking quite a lot about Sora lately and the reason I believe in that so much is because it's leagues ahead of any of the competition. And I, I genuinely do believe it's going to have tremendous effects on society. And I think that's going to be the next big thing. Similar to chat GPT, it's going to be what everyone's going to talk about because of the societal effects is going to have on us. But that seems to be where I get it from. Mostly I, I try to avoid the major newspapers. I try to avoid, you know, major magazines and stuff for news. I don't like them per se. It's, it's, it's. Yeah, it's, it's, it's because they tend to blow things out of proportion. So it's like a chat GPTs here they take all of your jobs and here's why. And then they link in like Klarna and it's like, well, maybe, but like you're, you're overhyping it, you know, it's. You've made me read like for 10 minutes and you could have summarized into a sentence. It's really annoying. But it's just the nature of the beast at the moment. You know, there's a lot of fear in the air. There's a lot of potential in the air, but I think take everything with a pitch of salt.

So you mentioned a couple of things and for the listeners, Sora is a text to video release by OpenAI. Correct. It's like best in class at this point. And then you also mentioned Klarna and were you referencing the study, the research that Klarna just showed that their customer service chat bots had like done the work of 700 customer service agents. Is that the thing you were.

Yeah, that's, that's the one. Yeah, yeah, yeah. So SORA is going to be the big. The reason SOAR is going to be the big one isn't because people are going to use it. It's probably because it's the proof of concept that comes with it. And I think the fear of AI starts with stuff like that. The uncanny valley where people can't really tell what is real and what's not anymore. And so that's why I think it's going to pop off. But as for KLARNA Yeah, they're kind of like the talk of the town at the moment. They may not have gotten the hype that may have, they may have wanted. But to be honest, I heard about it when Klarna. Well, yeah, when Klarna came along and said, look, we've replaced 700 people with an AI chatbot, 85% of our customers are still happy and they can still go to human to talk. And we're expecting to make an extra $40 million on top of it. And when, when you read stuff like that and as a normal person, that's terrifying as a company, even if you're an SME, you're reading and you're thinking, holy shit, I can like, I can have all of like these like bots and I can have this stuff. I can have entire company of like AI tools and, and, and that's, that's opportunity right there for companies. Whereas the average consumer and people who may be scared of AI hear that and they're like, I'm going to be replaced. And I understand that it's an interesting.

Topic because you and I, we're immersed in it and we have a keen eye for, oh, wait a minute, that's not human generated, that's synthetic. Right. We're testing with, we're, you know, we train people on how to use AI and small business. And so we test a ton of tools and we're testing some voice to voice AI right Right now that is conducting phone calls on our behalf with people who have customer service issues or we've got a few use cases. And even being somebody who is ever like aware of use cases or opportunities to use AI, it's still, it's so good at this point. And it's hard for me. If I didn't know it would take me, me 60 seconds at least before I was like, wait a minute, are you real? So yeah, here's the thing. Klarna with their app, removing 700, you know, or doing the work of 700 humans, I, there, there is that, that, you know what? I, I still want to talk to a human. However, what if that human can't get you the answer? And what if the AI can say, hold on, snap of a finger, here's what you want, plus here's a few additional resources. It becomes this question of do you really want the person or do you just want the result?

This is the interesting thing, isn't it? Because I think we want that human connection. We all do. But when we call up a customer service line and I'm not really the type of guy to get angry at customer service agent. I know 9 out of 10 times it's not their fault. But when they can't solve the issue I'm having, or when they give me a response that I don't particularly find satisfactory, I'm just like, I've wasted 30 minutes of my time trying to get this fixed and I went nowhere with it. But you're right. Like, what if an AI could get you the answer you actually want? Because this AI, it can read through like, you know, megabytes, different gigabytes worth of content and FAQs, and it's never wrong. It's never wrong unless it's manipulated into being wrong, but it's never wrong. A human can be wrong. I mean, I can't even, I have a breakfast this morning. An AI would be able to tell me why I have a breakfast like 12 years ago if I had the data. You know what I mean? So people talk about like AI, like I'll never be as good as a human. I, I would argue it'll be better than the human in time. I mean, just like the voice calls, right? Like if you can't tell that you're talking to an AI, like what's the difference? It's going to give you the answer you want. If you want to talk to a human and get the same response, but in human, then fine. But I mean, realistically you have, yeah, you can't, you can't tell.

And then with Sora, obviously the concern is that it's output is so good. With Klarna, the concern is that the chat bot is so good that people are using it without even realizing that they're not using a human. And oh, there's 700 jobs that go out. So it's an interesting time. You know, you brought up, you, you had kind of dabbled in X and then ended up on LinkedIn, I, I think it was a good move. The LinkedIn top voice aside, the content that you're able to put out there is resulting in engagement. Not just from nerds like me and you who love AI, but from companies. Right. Like, like this.

Yeah.

So going back to that suggestion for that individual and I said, well, what do you, what if you're a copywriter and you want to transition out of copywriting into prompt engineering? You mentioned go there and start sharing your AI journey in social. Right. If you're listening to this, maybe you don't want to be a thought leader. Maybe you do, maybe you want to be recognized and I can tell you there's nothing negative that will happen with your career. If you get identified as, oh, she's the AI expert. Oh, he's the AI expert. That is not going to hurt your career. So if you're listening to this, and this has been engaging and you're looking for, and you want to do more, start there, start documenting your journey. Because one of the things that I know and I've heard this, Edwards, a rare exception, most people that are professing expertise in generative AI today, they never saw what, they never worked with any models until GPT 3.5 came out and the whole world got access to it. Right. Most people out there who are identifying as an expert, they're an, they're only an expert expert six months smarter than you are. Right?

Yeah.

So that was fantastic advice that you were sharing because, I mean, you're a perfect. You're the result of that. You started documenting your journey. You started getting really good at. Well, the journalism background probably didn't hurt, but now, I mean, in short order and six months, you're a recognized expert in a platform that's very desirable if you're selling B2B services like you are, for them to go, well, who knows, what's this prompt engineering stuff? Oh, this guy knows. And all of a sudden, boom, business out there.

Yeah, yeah, that's the, that's the thing. And it's, it's, it's, it's, it's a bit like. I don't want to say frustrating because you can't change the past, but I do wish that it is my biggest regret. I wish I started earlier because if I did, there would have been so much I could have done. But, but, you know, to even have the LinkedIn Top Voice badge, it's, it's such a. I don't know, man. Like, it's, it's crazy to be. To even have it because it's such a difficult thing to get, like, just for the listeners out there, because I know, I know this is like, you might be listening to this and thinking, I want to do that. Like, I can do better than he can. And I hope you do, honestly. But, like, the way you get a LinkedIn top voice badge and it's the blue badge next to your name, not the yellow badge. The blue badge is when LinkedIn have personally seen what you've been doing, and they say, this guy's literally the best in his domain. We're just going to give him the tick for it. We're going to give him like the blue verification tick to say, this guy's legit, like he's for real. And so it's not just a single person who does this, you know, it's. I think the last I checked was a three layered system. They're very, very secretive. So they'll find you, they'll scout you, and then they'll keep on looking at your content to see what you're doing, to make sure you're actually talking about your stuff and, you know, your stuff, and they'll make sure you're not cheating the system and gaming the system. So a of the people that rai influencers at the moment who kind of just copy and paste their content from other people, they'll never get the badge because LinkedIn are smart enough to catch on. And so, yeah, and so they'll keep on watching and eventually they'll say, okay, let's take him to round two. Let's, let's get some other experts in AI and tech to look at his work and to see if he really knows his stuff. Then once that's done, he'll go to another round, a final judgment round, and then we'll say, yep, he gets the blue tick of approval. He doesn't get the blue tick of approval. Here's why. And, and no one in the world, apart from LinkedIn's internal team know who those people are, because it's such a coveted like, like in the uk, it's not really massively celebrated and I understand, I mean, it would be a little bit strange, but it's not a massively celebrated thing. There are events specifically for LinkedIn top voice people, but I haven't been invited to one yet. But to be fair, I only got the badge like two months ago, so fair enough. But I know in dubai in the UAE, there are these huge, massive events specifically for LinkedIn and these top Voice people. And I get invited to do talks because of that LinkedIn Top Voice badge. I mean, although it's unconfirmed yet. I mean, currently my inbox, I've been invited to be on the panel for two judgment things for AI and tech. So it's very strange to me because only last year I was starting to get into AI, now I'm about to be part of two panels judging AI entries because there's just. That's it. Yeah, it's an authority. And I think, you know, again, you're right, like, if you want to do this, just document your journey, you know, I think a lot of people go on onto LinkedIn. They're like, where the hell do I start? You know, like, what do I tell people? And I think a lot of people approach this from an academic perspective, but I think just being explorative, you know, just going out there and just doing your thing, I think it says a lot, you know, and you can always look back at it and you can be critical about that. There's one guy and I'll give him a bit of a shout out. He's, he's not like a big deal, he's not an authority figure, he's not an influencer or anything like that. His name is John Paget and John is a lovely guy. He's, he's a part of a council for the government in the city, which is quite far from me. But he's a really nice guy and we met through AI and he's, he's kind of just been a very big supporter of mine from day one. He's only got like 500 connections, not that many followers at all, maybe 200 followers. But he's just out here every day. Like I've released a new custom GPT, I've released a new book on Amazon and he's just at it every day, you know, he's just going for it and he, you know, he's, he's getting places because people at work are asking him to implement AI and in his city, he is the AI guy and he's getting stuff from it, you know, and I'm really proud of him. I haven't really given him much in the way of like super advice. I just kept on supporting him and helping where I can. And it really can be a life changing thing, just being vocal on LinkedIn and just, you know, being.

Yeah, this has been very informative because just as we referenced already, you don't, can't always trust what you read. And with prompt engineering in particular, I spend a lot of time on that, the Twitter sphere for AI and there are so many just crap books for, you know, here's my 1000 chat GPT prompts to start a business. And you think, oh my gosh, I don't have to learn this stuff. I can just use these. Yeah, look at them. And they're trash. And they're like one line. Oh yeah, and all that. So the distinction that you gave with LinkedIn, that's going to be very helpful for people who are, who are looking for thought leaders to start to follow. If they don't have that doesn't necessarily mean that they're, they don't know what they're talking about. But if they do have that recognition, like for instance, if, if prompt engineering is something that you want better perspective on, Edwards profile is one that needs to be followed by you. Right, so that was, that was a, an unanticipated nugget there that as you're on the platforms, look for the people who are really doing it, not all the people people, because there, trust me, AI is, it's, it's the, the Internet boom once again. Right. Like there's all kinds and you know, people coming out saying that I'm the expert and their stuff is not going to get, you're not going to experience the transformation that can come from using AI.

Yeah, yeah, this is, this is just like, just one more point on that because I'm really hoping people listen to this and take it away and they'll be like, I want to go on this journey, but I don't know where to start. I think just be, be careful out there. I think one of the things which I really struggle with on LinkedIn and just generally kind of this particular area is authenticity. And the sad truth is, is that a lot of people, they do enjoy the 1000 prompt guides. They do enjoy the hundred. Yeah, they enjoy it. And the reason they enjoy it is because it's a shortcut to success. And that's an aphrodisiac to a lot of people. Like, oh, I don't have to learn prompt engineering. That's like a whole 12 hour course. Like on the surface to be 12 hours is nothing. But to a lot of people, they're like, I don't want to do that. No, I want to just copy and paste prompts all day long. Fair enough. You know, you want a prompt library and okay, fine, like here's a thousand prompts. And none of them are really going to be that good because they're not tailored to you. But the thing is, is there's a lot of people who do that. You know, they want you to follow them, they want you to download their 1000 prompts guide. They want you to make sure your eyes are on them. Because if your eyes are and you're into this AI, they can sell you stuff. You know, they'll take that 300 pound payment from some random person halfway across the world and they'll promote the ever loving crap out of their product because they've been paid to do it. But that doesn't necessarily mean they know what the product does. And again, for the listeners at home or in the car, wherever you may be Listening to this, I suppose, just to put into perspective how easy it is to actually make these guys. And I might make a few enemies, I might lose a few friends, but I'm willing to do that. All they need to do is they need to have reference points, right? So the reference points can just be 10, 20 prompts. And you would copy them and paste them into something like anthropic or into maybe ChatGPT4. And then you'd get ChatGPT to read, and you just prompt and say, here are my 20 chatgpt prompts. I want you to write another thousand that do similar things within the field. And that's it. That's all they do. And then they package it up into a PDF, they offer it for free, and they go, here's my free 1000 prompt PDF guide. It's free, download it. You've got no risk. But I just want your email address so you can join my newsletter and then I can, you know, email you stuff I want you to buy. That's the whole fun. Yeah, but it's, it's not value. If he dragged one of the influencers aside and you were like, hey, I want you to write me a prompt that can do very complex things, things they're going to choke. They have no idea what they're talking about or doing.

I hope people are paying attention to this. This is the truth.

Yeah, it's the truth. It's. It's just the truth, man. Like, there's, there's a lot of people who call themselves prompt engineers, but you strip away the CHAT GPT, you strip away the large language models from their hands and you ask them to open up a WordPad file or whatever else and write out the prompt in front of you, they're going to suffer, you know, and that's the downside of ChatGPT is that it's giving people the ability to pretend. And I think the real people who know how to prompt and who know how to use ChatGPT to its fullest potential rise to the surface. And those who don't, don't, you know, there's no problem in not knowing how to use it as max capacity. The problem is pretending that you do and pretending that you know. Because it harms businesses. Yeah, it harms businesses, man. Like, you're taking like a prompt engineer who doesn't know how to do security settings, taking a prompt engineer who doesn't know how to do basic large language model stuff, temperature management, anything really, and you'll drag them in, you'll pay Them. I think the going rate for Anthropics Prompt Engineer at the moment is like half a million dollars. And it's, it's high. Yeah. Like if you even a startup, you know, it's, it's like in the uk I think they charge senior developer rates. I know I do. And you know you're paying like three to four thousand pound a month maybe for a prompt engineer and, and they'll write you out a prompt and it doesn't do what you want it to do. And it's like, well, we're paying you all this money, what the hell are you here for? And then you realize the prompt that they're writing is written from ChatGPT. Like, damn it, we've just put a down payment on this guy, he's a contract, we can't fire him. But they're inept and it's all because of ChatGPT. It's really strange. It's so strange. In fact, my friend works at a full time job, he's just taken on board a software developer. The software developer doesn't know how to code in Python, he doesn't know how to code in C, even Flutter or whatever else. And he just sits there with ChatGPT, the free version, Google, Bard. Yeah, well, Gemini and Anthropic, and he just sits there. All three major large language models open brazenly and he's just extracting code from each of them and none of the code works because he doesn't know how to prompt. So not only does this guy not know how to do software engineering, he doesn't know how to prompt, he doesn't know how to make it work. And you've got to be careful, man, you've got to, to be very careful. There's a lot of people who pretend to know it all, but they, you know, I think they just don't.

It's also an indication that the marketplace is looking for these people, so much so that they're willing to pay jokers to come in and do, you know, substandard work because the demand is so great and the supply of talent is so low. So individuals who are listening to this start learning to think in AI, if you can. If you can think in AI eye, none of those books matter. Somebody can drop you in the middle of the desert and say prompt your way to a glass of water. Right. And you'd be able to do it.

Yeah.

If you, if all you've got is this collection of PDFs of, oh, I know somewhere in here I've Got a prompt for that. I know somewhere you're digging around in there. That's not what we're talking about here. And that's the kind of situation where, as Edward was mentioning, if you don't understand why you're doing it a certain way, mistakes are inevitable. They're guaranteed. So, Edward, this has been a fantastic interview and my intention was to really kind of dive into the paradigm of somebody who prompts for a living. Right. And I think that we did that here. But the takeaways for me in particular was I. I keep hear, like I mentioned, man, this is episode whatever. But the people that I've talked to, there's a pattern here, here that I'm noticing and I hope that the listeners are noticing. And that is. It's not about the engineering comes from understanding how to describe what you want. Not from any technical or. Or being wired like an engineer. That's not what this. This is almost, as you mentioned at the very beginning, that quote that was, you know, mastery of the English or mastery of the. The spoken language is the new coding. Right. That's where we find ourselves with prompt.

Yeah. English is new programming.

Fantastic. And I don't know how busy you are and I really appreciate the time that you took to do this, man. So with that, how can people find out more? I know that they can go and please do go to LinkedIn and start following Edwards content because it's fantastic and it's frequent. But is there any other way that you want people to maybe reach out or connect with you?

Well, LinkedIn would be the best. Yeah, I mean, you can find me on LinkedIn. Do just search for Edward Frank Morris. Just for the listeners, it might be curious. Frank actually is my middle name just in case they think it's like a crazy weird nickname. It's actually just my middle name. So you can find me on LinkedIn. It's Edward Frank Morris. There's only one of me on there. Unless I've got a stalker of some kind, you can find me on there. I don't really use Twitter all too much, but you can find me on there, too. My Twitter handle is whose efm. So just straight up, whose E fm. You can find me there. Alternatively, if you really want to. I wouldn't suggest it because it'll take me ages to get back to you. You can always check out my website. It's www.encyigmatica.co.uk. that's enigmatica.co.uk just like the Enigma machine. So, yeah, please like LinkedIn. It's the easiest way to get at me. And yeah, if you ever need help and you want to know how to build up a portfolio for prompt engineering and stuff that people would find attractive within the prompt engineering portfolio, just drop me a message. More than happy to help. There's a shortage of actual capable people out there, so please. Yeah, beautiful stuff.

Thank you everybody and we'll see you on LinkedIn.

Edward yeah, see you there.

Thanks for tuning in to Using AI at Work. Don't forget to subscribe for more conversations about how to use AI at work and a special thank you to our sponsor, Chief AI Officer for empowering businesses with AI education and training. Visit their website for a free AI Readiness Assessment, Assessment and AI Strategy Guide to help you get started using AI at work. That's www.chiefaiofficer.com. so thanks to our producer Evan Desaulnier for making this episode possible. Follow us on Twitter at the handle using AIATWORK and visit www.usingaiatwork.com for free resources to help you harness AI in your role.