The people and stuff we mentioned on the show:

The Copywriter Think Tank
Kira’s website
Rob’s website
Sam’s Twitter
The Copywriter Club Facebook Group
The Copywriter Underground
Free month of Brain.FM

Full Transcript:

Rob Marsh:  You’d have to be living on a different planet to not have your inbox clogged with emails about AI. Talk about artificial intelligence is everywhere. Some people are saying that it means the end of content, copy, and copywriting. Others are saying the opposite, that AI is the biggest opportunity for marketers in generations. And the truth probably lies somewhere in the middle.

I can’t remember who said this, but I recently saw a comment that said, “AI won’t take your job, but someone using AI will.” So learning about these tools and how to use them is not just a good idea, but quite possibly the best way to ensure that you’re still working as a copywriter in the coming decade.

Our guest for this episode of The Copywriter Club podcast is copywriter and AI expert Sam Woods, and we grilled him with our questions about AI, ChatGPT, and what it all means for the future. You are definitely going to want to stick around for this one.

Kira Hug:  Before we get to our interview, the podcast today is sponsored by, actually, it’s sponsored by our new podcast. So we have a new podcast that is launching soon featuring other experts like Sam Woods, and so today is a preview of what’s to come on the new podcast, which is called AI for Creatives. So if you like today’s episode and you want more of that, you can just check out our show notes and there’ll be a link in our show notes so you can get on the list and hear all about the new podcast when new shows come out.

Rob Marsh:  And that podcast is we’re interviewing experts in AI, experts who are developing their own AI tools. It’s really all about how we get better at using artificial intelligence in our own businesses as creatives.

Kira Hug:  And this podcast episode is also sponsored by the Copywriter Underground, which is our membership for copywriters, content writers, creatives. And we’re creating a new series of AI trainings in the membership so that you can figure out how to use these tools, how to apply them to your business. And so actually, Sam Woods has a training in the Underground that you can access where he shows a demo of how to use the tools in your own copywriting. And so, if you want access to other trainings like that, definitely check out the Underground membership.

Rob Marsh:  Yeah, listen to this episode with Sam, then go to the Underground, check out what he shared there. That’s good for now. Let’s kick off our episode with Sam Woods.

Kira Hug:  How do you see AI impacting society and the economy in the next five to ten years?

Sam Woods:  Yeah, I’ll make a prediction and everyone will then tell me that I was wrong five years from now. No, well, so it’ll have… And it’s hard to say these things because it sounds like hype, and it sounds like overstated, and it sounds like everything else that’s been hyped. Right before this you had Web3 and crypto and all this other stuff, right? NFTs. And they were all hyped and so on.

But what’s different about AI is that it’s an infrastructure play, as it is being integrated into all the things that we are already doing. Web3 is not infrastructure, Crypto is not… Even though crypto could be infrastructure in terms of payments and coins and tokens and everything else, but it wasn’t, and it probably never will be. But what you can do with AI is at an infrastructure level in society as it can be implemented into, not everything, but most things.

And that’s what we’re seeing now. Like Bing. Microsoft made the investment of the decade years ago when they invested into OpenAI, probably the investment of, not just a decade, but the century. And they’re integrating ChatGPT into Bing, they’re going to integrate it into Microsoft Word, the Office suite. All the tools that Microsoft has, they’re going to implement this little chatbot type thing and have it be your writing assistant, your presentation, like PowerPoint assistant.

In any tool, any app, any software you can imagine there, a lot of companies are integrating AI into it. Notion.so, the tool that a lot of people use, they’ve integrated AI into the tool, and it does simple things like summarizing or does whatever. So we’re seeing that happen, and it’s going to accelerate over the next few years to the point where five years from now, it’ll just be a part of our daily routine of any kind, anything from work, to play, to family, to social life. It’ll just be there in different ways and in different capacities.

Now, the interesting thing is everyone thought AI was coming for the blue-collar truck driving jobs first, but they’re coming for the white-collar information worker, creative works first or creative jobs first. It’s coming for the creative class, if you want to call it that. So designers, writers, photographers, videographers.

What we’re seeing with generative AI is how it’s becoming as good, and what we’re seeing now is early. This is an early version of what it can do. Just imagine if this is version one, imagine version 50. And it can create texts, images, video, audio, either from nothing or just based on a minute of you speaking into a microphone. And then it can take your voice, analyze it, and then replicate your voice and have you say anything. So the capacity for you to create deep fakes of yourself is, unfortunately, for anyone to create deep fakes of you, is the tech is there and it is both good and bad. It’s terrifying and exciting at the same time.

But five years from now, it’ll be part of everything we do, most things we do. Anything that involves a computer or a cell phone, somewhere in there, there’ll be an AI bot doing different tasks, anything from writing to analyzing, to reminding you, to you name it. And it’s going to create a crap ton of new jobs that didn’t exist before.

It already is. If you think about prompting, which is how you interact with something like ChatGPT and GPT-3, that is becoming a job where I believe companies are going to start hiring. And who knows what they’ll be called? Some say they will be called prompt engineers, prompters, chief prompt officers.

Rob Marsh:  Copywriter.

Sam Woods:  Yeah, exactly. Prompt writer. I like to just call it the prompt librarian I’ve seen as well. So I just like to call it prompt craft, because you’re really designing language that you then use to give instructions. So it’ll create jobs like that. It’ll create jobs for people who know how to integrate AI into any company’s workflow, whatever it might be, and processes. So it’ll create a ton of jobs, and it’ll also eliminate a ton of jobs, and it’ll reshape the economy in different ways. Not so much destroy it. It’ll just change it. Everything will change. That’s the short version.

Rob Marsh:  So you’re making a prediction. I want to actually take you back in time to 2015. We were together in Texas, I think it was the first time we met in person, and we were actually talking about AI at this conference. And a person at this conference, I don’t know if you remember this or not, we’ll call him Ed because that was his name, so we were in 2015 and he was telling us, he said… This was April 2015. And he said, “By October, AI is going to completely replace copywriting.” And you and I both laughed.

Sam Woods:  Yeah, because he was full of it. But anyway, go ahead. Yeah?

Rob Marsh:  We’re like, “No, no way. Not going to happen.” But I think at that time we said, “Eventually, yes, it may happen, but we’re years away from that.” We were right. I mean, it has been years. It probably came a little faster I think than some people were expecting. The interface that ChatGPT in particular presents I think has scared a lot of people.

But let’s just go a little bit broader as we talk about AI, because ChatGPT is one tool, but it’s not the only tool. And there are all kinds of different AIs. We should also mention, we say AI, it’s not true intelligence, right? It’s artificial, but these are algorithms, programs that are designed to do particular things. But let’s talk a little bit more about some of those other applications, too.

Sam Woods:  Well, so the funny thing is that the biggest game in town is GPT-3, which comes from the company OpenAI, and they also release ChatGPT. But ChatGPT is just a version of GPT-3.

Now, all, most up until now, and I think that’ll change this year, but up until now, almost just about every single text generating tool out there, Jasper, Copy.ai, you name it, all of them, tap into GPT-3 via API. So they look different, and they have different ways of prompting the API and interacting with the API and giving you stuff back, but it’s all based and taps into GPT-3. So up until now, the biggest game in town, and really the only game in town, has been GPT-3.

Rob Marsh:  Which is OpenAI.

Sam Woods:  Which is OpenAI. GPT-3 is a language model created by the company OpenAI. So OpenAI is a company, and before GPT-3 there was GPT-2, and before two there was one.

Rob Marsh:  And there’s rumors of four being available later this year.

Sam Woods:  Four exists. It’s just being beta tested by select companies. So it exists. And at some point this year it’ll be made available to other people.

But it’s a language model that’s, and I’m going to try to keep it simple, is that language model that’s been trained with the help of what’s called machine learning. And what happened is that it consumed almost all the internet and then it’s become a very… It’s a text predictor. So it’s a very fast, very smart text predictor. It can predict whatever text is supposed to come next. So it’s not intelligent, it’s not artificial intelligence the way we think of it or have talked about it or the way it’s portrayed in movies. It’s not sentient. It doesn’t understand. If you ask it a question, it’ll answer, but it does not understand your question. It just predicts what the text is that should come and that goes with your question. So it’s not smart, it’s not intelligent. It’s really machine learning. Everyone calls it AI because it’s just a buzzword, but it’s really machine learning. But all these other tools are based off of it.

That’ll change. There are plenty of other language models out there. They’re just not as widely adopted or used yet, but they will be. Google is obviously making a bigger… The funny thing about Google is they’ve had this technology available to them for years, they just never released it in the way that ChatGPT came out. So these tools are mostly based off of OpenAI’s language model, which is GPT-3. I think that’ll change, but right now they are all very similar. Models for images and audio are different, a different training set, different models altogether.

Rob Marsh:  Yes. So for clarity, when we say models for images, we’re talking about things like image recognition and image creation. And it’s different because we’re not using words, a predictive model on what word should follow, but it’s similar in the way that it is predicting what colors should go here, what the brush stroke should look like based off of the input that we tell the machine it should be using. Correct?

Sam Woods:  It takes instruction like you’re talking to it and you tell it what you want, and then it’ll interpret what you want and give it to you. So all of this falls under generative AI. So AI for creative work where it literally generates stuff is called generative AI, which is a different kind of AI than other types of AI. I’m not going to get into all the tech details, it’s too deep and too far, but just so people can think about it, the text model that’s called GPT-3, language model really, falls under the larger category of generative AI, which is exactly what it sounds like. It’s AI for generating X, Y, Z, whatever, anything.

Kira Hug:  Okay. All right. So how are you using these tools in your business and your creative work and your own systems and processes on a professional basis?

Sam Woods:  Yeah, I’ve used, so if I talk about poems, I can’t, that’s not professional. Is that professional too? I’m paid to write pulp fiction with AI. But anyway, we’ll get to that point. So-

Kira Hug:  We can talk about that in a little bit.

Sam Woods:  We can talk about that in a little bit. I was fortunate just because of people I’ve come to know in my work over the years. I was fortunate enough to have access to GPT-2 back in 2019, and then in 2020 GPT-3 before it was made publicly available. And ever since, I’ve used primarily GPT-3 because it was the most advanced and just the best one available in the process of anything from ideation, brainstorming, drafting, writing, rewriting, and editing copy.

I’ve also used it for things like optimization, optimizing a landing page or an ad, or whatever you want to optimize. And it became a huge part of what I’ve been doing for clients over the past few years, ever since.

Now, before that, the only exposure I had to artificial intelligence so to speak, or machine learning, was just via data analytics and data crunching because AI for those things is far more developed than AI for text generation. AI for data analytics and numbers crunching and business intelligence and so on and so forth is very far advanced and has been around for a very long time and it’s nothing new. But that was my first exposure to it. And then eventually, it turned into using tools for primarily copy, but then as these image tools became better, as in producing better output, I started incorporating that as well into essentially any client work that I’ve done over the past few years.

Rob Marsh:  Yeah, let’s talk about how. Let’s talk about how you’re doing it. So let’s say a typical client comes to you. I think you mostly write sales copy, a lot of long form. Maybe there’s some ancillary things that are attached to that. You sit down with a client, or you’re ready to work on that project. At what point does an AI tool start to play a role in the work that you’re doing?

Sam Woods:  From day one until the very end of the project, until everything’s delivered. So my work as a copywriter has changed a lot over the past few years, and if it’s helpful to think of what I do as sales copy, then we can stick with that, but it’s not really what I do anymore. It’s part of what I do, but it touches on a lot more than just writing a long-form sales page. It covers-

Rob Marsh:  Let’s use that as an example and then let’s expand on how you’re using it in your business because I think it’s at least helpful for me to start with a specific example, walk through it.

Sam Woods:  Yeah. For long-form sales pages, AI is incorporated at every stage and from day one. So day one of a client project starts with some form of research. You look at the product or the service they’re selling; you look at the market, you look at competitors, all the usual things we know to do as copywriters in the research phase of things. You look at features of benefits, promises, how something works, outcomes you can promise, who the avatar is, what their fears, problems, frustrations are, what they want, what desires they have in mind, what dreams they want to come true. Whatever you do as a copywriter in the research phase you can do with the help of these AI tools.

I mostly, 95% of the time, use GPT-3 and at this point ChatGPT together, and they’ve been my primary go-to as tools. There are other tools that are great to use, Jasper, copy.ai and so on, and they’re fine. I tend to use those the most because I have more control over what I do with it and what I get back.

But from day one, when I do research, I use it to discover things about my avatar. I’ll describe my avatar and then ask specific questions. I examine the product or the service with the help of ChatGPT or GPT-3 where I analyze the text about it and have these tools tell me things about the text that I give it. So it’s research, it’s ideation, it’s brainstorming big ideas, it’s brainstorming angles and hooks. It’s doing research.

And not that I rely only on these tools to do the research. I’ll have it do some, I’ll bring research I do to it and have it analyze the text for me. I’ll pull together reviews from Amazon or some other e-commerce site if that’s applicable, or Reddit forums or anywhere that I can find what people are saying, the voice of customer research. I’ll take interviews or services that I do, anything that’s the voice of customer, I’ll take, give it to ChatGPT and GPT-3, and have it analyzed it in different ways.

And then it’ll tell me things like out of these 2000 words of reviews, here are the five primary desires expressed. Here are the human biases expressed, here are the negative sentiments, the positive sentiments. Here are the needs in Maslow’s hierarchy of needs expressed. Here are the objections people have. Here are the positive things people say. So it’s become a tool for analyzing voice of customer research and any other research that you do when you write a sales page for a client. From day one, it’s there. And then eventually I use it for drafting and writing as well, but-

Kira Hug:  Can you talk more about the analysis part? Because that’s not quite clear to me. I’m dropping some of my research into the tool, but what am I asking to get back the information I need? What am I looking for?

Sam Woods:  You ask any question you want to have an answer to. And I know that’s the simplest answer, and I’ll explain what I mean, but I want people to understand this. You can ask any question, give it any text, and ask any question you want to know about the text, which means that what you can ask is unlimited. There is no limitation to what you can ask.

So as an example, the way you interact with ChatGPT and GPT-3 is through what we call prompts, and prompts are just statements. It’s either a question you ask, or a function you give it, or a direction you give it, you tell it to do something. Write me 10 headlines. That’s a simple prompt. It’s a really bad prompt, but it’s a prompt example.

But for the analysis, literally, and let me know if I’m either too simple and too advanced. Truly, literally what you do, let’s say you have 2000 words of Amazon reviews for let’s say your client’s product. Make it simple. You literally put that into something like a Google Doc, clean it up, make sure the formatting is fine, remove any useless information like someone’s name. We don’t care what the name is, we only want to know what they said. Clean up the text, format it nice and clean, remove names, remove stars, whatever. Clean it. All you want is the language.

And then you write a prompt, and depending on what you want to know, so if I want to know what the primary desires expressed are in this body of text, then it’s something a bit more detail in this, but essentially what I’m asking in the prompt, I write a prompt out, which is just a statement and a direction, and I say, “Analyze the text below and tell me what the five primary desires expressed are.”

There’s more to that prompt, but I’ll keep it simple for now. Literally, truly, the prompt I have is maybe five sentences, but it’s literally asking what are the desires expressed in this text? I take that and I highlight that prompt together with a text that I wanted to analyze. I paste it into ChatGPT and I hit enter, and then it tells me what the desires are in the text.

And replace primary desire with anything you want to know. Objections, positive sentiment, negative sentiments, what Maslow hierarchy care needs are expressed, what human biases show up in this text? And truly, it’s very low-tech for being a very high-tech tool. Truly what you do is you literally copy and paste stuff into the little box and you hit enter.

And so I’ll read out a prompt so you hear how the level of detail that’s in it. And so this one is for primary desires. What I’m asking it to consume any text I give it and tell me in this text, what primary desires show up, what primary desires are expressed?

Now, and here’s where the prompt starts. Here’s what I’m saying to ChatGPT. You are an expert on human emotions, behavior, and language. You can easily and expertly detect human behavior, thoughts, and logic based on language. Make a bulleted list of six primary desires people experience based on the texts provided below. Mark this list with a heading, six primary desires. Also, make a separate list of what desires are not present, but should be if I want to appeal to a specific avatar.

And where I then change is I describe who the avatar is at the very end. So I will say, “Make a list of things that could appeal to copywriters in their late thirties who started a family and are wondering how to feed their kids.” Then I give it a list. “Here is a list of desires that I want you to look for.” And then it’s a list of about 15 or so desires that I just paste in. And then I say, “The text you should analyze is here,” and then I paste the text, hit enter, and it’ll analyze the text and tell me what the six primary desires are, and it’ll give me a list of the desires that could be or should be in a text like that that could appeal to whatever the avatar is.

Replace the primary desire and change a few words around, and you can insert human biases, psychographics, logic versions of motion. Positive sentiments are for things like what are the goals expressed? What outcomes are expressed? What wants, hopes, and dreams? What features and benefits do people want? What relationships are expressed?

Go to negative sentiments and what pings are expressed, what problems, what fears, what worries, what frustrations, what uncertainties are expressed? Maslow’s hierarchy. What needs in Maslow’s hierarchy needs are expressed, and which aren’t, but should be? What objections do people have in this text? List them all. And analyze the voice, tone, style, and diction of this text and give me a profile of what the voice, tone, style, and diction is like.

Rob Marsh:  Now, to be clear, Sam, as you’re sharing a lot of these prompts, you’re usually not combining them all, right? You’re doing them in a single conversation, but you are asking it separately, each one of these things, right? Or do you throw all of that into one big prompt?

Sam Woods:  No. Well, Maslow’s hierarchy of needs is one prompt. The desires are a separate one. Now, I have a process that I go through, so you open up… And I’ll use ChatGPT because people have, I think, most experience with that as opposed to GPT-3. Funny side note, GPT-3 is better than ChatGPT, but it never took off because it wasn’t as easy to interact with it as it is to interact with ChatGPT.

Rob Marsh:  So there’s a copywriting lesson for you. Make your thing feel like you’re talking to a human if you want humans to talk to it.

Sam Woods:  Yeah, exactly. So side note. Anyway, the point is that I’ll open up a new session, and for any client project stays within the same session. And you can save your session and so on. And so when I start a project, like a sales page project, I will start with prompts and several of them that’ll help me do research around the avatar research, around the product or service that I’m writing for. I will then form a profile of who the avatar is. I will uncover what the unique mechanisms are. I will find big ideas in the research that I do with the help of ChatGPT.

And then when I’m done with that, I’ll have a document that has all the output that I want to keep, that profiles the product, the unique mechanism, the big ideas, the features, the benefits, and the avatar that it’s for. And I’ll know deep things about the avatar, like what relationships matter the most for these people? What true deep fears do they have, and how do they show up? What does a day in their life look like for this avatar? All the deep things that we usually do research for are there.

Then I move on, stay within the same session, and then I move on and then take, with that, all that generated output, I will then start drafting copy. Headlines, sub-headlines, section copy, product copy. All the different types of copy that exist in a sales page, I will just start drafting it with the help of ChatGPT.

Kira Hug:  What are some concerns you have? Before we get to the drafting phase, you’re talking me through this, this is making sense. I can get the research and analysis from the tool. What are you thinking about as you’re getting back this analysis? How are you thinking about it objectively so you can discern what’s worth paying attention to, what’s not worth paying attention to?

Sam Woods:  Yeah, so the first thing you should know is that because it is a language model, it is not a fact model and it’s not a scientific model. It can replicate and tell you facts and scientific things and so on that are true and accurate, but it can also tell you things that are wrong. So this is why I keep saying, and any chance I get, that this is a tool for collaboration and not your single source of truth.

So if I have research that I need where I need to be absolutely sure of certain facts, I don’t ask ChatGPT those things. I will find the facts.

Now, there are other AI tools where you can more easily find facts. There are websites like elicit.org, I think consensus.com, talktobooks.com or something to that effect. And so there are platforms in SCI space where you have access to all the world’s research and scientific papers. And then with the help of AI, you can consume them, summarize them, and ask questions about them, and have AI help you understand them and do the research for you.

So I will use those tools to get the facts that I need, and then I will bring the facts to ChatGPT for the sake of summarizing, rewriting, and analyzing. But it’s a collaboration. You never rely on, whether it’s Jasper, GPT-3, copy.ai or any other tool, they’re not meant, they’re not coded, they’re not trained to be fact-checkers. They’re meant for language. I say it again, they’re language models. So you bring the facts that you need to be sure of and the references to them that you need to be sure of, you bring that to the table, so to speak.

Now, there are search engines popping up that will give you references that use AI to produce results. There’s u.com, neva.com, and soon bing.com will have ChatGPT integrated, and it’ll give you not just answers on ChatGPT, but ChatGPT will be able to actually give you references in the search. So it’s coming, but so far you use it for language, not for scientific fact. So you just have to know it’ll tell you things, but you can’t always trust what it tells you to be actually accurate. However, when it comes to language and writing and analyzing, it’s flawless.

Rob Marsh:  Yeah. And as you’re mentioning some of the tools, we should probably note, Google obviously competing with Bing has their own AI, Sparrow, that is rumored to be, I don’t know very many people have played with it outside of Google, but it’s rumored to be very much like OpenAI or GPT-3, but with an up-to-date internet connection, which would be a step up from the data set that’s in GPT-3 now.

Sam Woods:  Because GPT-3 and ChatGPT are cut off at 2021. So it doesn’t know events or things that happen after the end of 2021. Now, Google is releasing what they call Bard, which is an odd name for their little tool, but everyone is now… Everyone. These companies are now competing to release these little AI chat tools all at once. And so if this is version one, again, I’m telling you, when version 50 comes or version two comes, it’ll be far beyond what we can do right now. It’s not getting worse. It’s only getting better.

Rob Marsh:  So let’s talk a little bit about some of the limits, because as you’ve talked about the research capacity and capabilities, this seems to me like the best part of OpenAI and ChatGPT right now. The writing part, to me anyway, there’s still a lot lacking there, but let’s talk about some of the… We can talk about that in a minute, but-

Sam Woods:  Sorry I’m dropping you, I’m just… Go ahead.

Rob Marsh:  Yeah, no, no. There are clearly some limits. We just mentioned the data set is cut off. You were talking about accuracy, you can’t just pull facts out because it’s not an encyclopedia and necessarily it’s a predictive model. It has trouble with some math. Obviously, it can predict easily something like three times three because that’s easily predicted, but if the math gets complex or calculus, it can’t actually predict the outcomes a lot of the time. I think there are some limitations also with programming languages. It can predict what code should say in order to produce an outcome, but it doesn’t necessarily have the thinking capacity to actually figure out is this truly a bug, or will it actually do the thing that it’s going to do? So maybe it’s 70, 80% there, maybe it’s more than that or a little bit less. But there are definitely those kinds of hangups.

And there are also some biases built in because of the data sets that we’re working with, whether it’s racial bias, and it’s not just that, but there are biases that we have to be a little bit careful of. So like you were saying, I think it’s really important to bring our copywriter brains to this, and it’s not just, hey, we’re turning over an assignment to an AI machine, but we’re partnering basically with this tool and we have to suss out some of the stuff that it’s not there yet.

Sam Woods:  Yeah. But what you also have to know is that these companies, especially OpenAI, are, if not daily, they’re certainly weekly filtering how we can interact or not interact with it. There are prompts that worked a couple months ago that don’t work anymore, as in it’ll decline to give you output based on the prompts. So it’s changing.

Rob Marsh:  I mean, I think some specific examples of what you’re talking about is there’s people who have asked it to do things like build bombs or do illegal-

Sam Woods:  Yeah, yeah. Okay.

Rob Marsh:  And it won’t answer those questions anymore?

Sam Woods:  Well, it will, if you give it a certain…

Rob Marsh:  You have to do the walk, the end around though. It’s like you have to write a script in which a character prepares-

Sam Woods:  If you fictionalize it, if you ask it to write a story, then to a degree, it’ll write whatever you want the story to be. And people are talking about prompt injection, prompt hacking, which is how you… So it’s a language model, which means that you can trick it into saying anything you want if you just know the right way to trick it. It’s like a human being. I can trick you into saying something if I just know how to manipulate my words in a certain way so that you then start saying things. So it’s possible to jailbreak it, whatever term you want to use. It is possible for someone to make it say things that it’s specifically programmed not to say.

And there are sub, what’s a Reddit thing called? Sub-Reddit forums? Whatever the heck. Sub-Reddit, where people are competing about how to jailbreak it. And so they’ll make it say things, and they share tips on how to make it say things.

The point is that that’ll always happen because there are hackers who still try to break into computers all day long everywhere in the world. So there’s always going to be someone who’s trying to break into this thing.

For you as a copywriter though, you have a specific use case if you use this for client work, or for your own stuff. And it sounds weird, but talk to it as if it is a human. It’s not a human, it doesn’t understand what you’re saying, but it’s trained on human language. And so whatever showed up on the internet when they consumed it all and processed it all and so on, it’s there. If I ask it, what would a 50-year-old male fear who’s… A 50-year-old male who struggles with joint pain, what does he fear every day when he wakes up? It’ll give me an answer, and it’s most likely a very, very true and accurate answer because someone, a 50-year-old male who struggles with this somewhere on the internet said this thing, and then it picked it up.

So you can jailbreak it and you can spend your time trying to make it say things that it is programmed not to say, and that aren’t nice things to say, but use it as a copywriter or use it for any kind of writing, but if you think about your use case, you use it for a specific purpose, which is to write things.

Rob Marsh:  Okay, Kira, there is a ton of really good stuff. In some ways I kind of don’t even want to comment on a lot of this because it’s so good, and Sam says it so well, and sharing so many really good ideas, but it’s probably worth just underlining some of the stuff that he’s pointing out, stuff that we need to keep in mind. So do you want to kick it off with maybe one or two things that stood out to you?

Kira Hug:  Yeah, I mean, the importance of this transition in what we do as copywriters and creatives is really important. And even just giving yourself a new title and helping your clients in a new way doesn’t mean you have to blow up your business, doesn’t mean you have to shut it down, doesn’t mean you have to stop doing what you’re doing. But if you can start to experiment with prompts and use this tool, and use all the tools just so that you understand how they work, and you can start to even advise your clients and share insights with them, and learn how to use these tools to do your job even better, that gives you a market advantage because there are many writers listening and other creatives who are not going to do that for many reasons. It’s overwhelming. There are tons of reasons not to do it.

So if you are doing it, maybe you even give yourself a title of prompt engineer. Maybe you call yourself a prompt copywriter, a prompt marketer. Throw that title on LinkedIn because that’s going to start to attract a new audience for you, because there are people and clients who have money to spend and who are looking for these experts now. And there’s not a school that is farming out these prompt marketers. We’re all starting from the same place here. So I think there’s a lot of imposter complexes that can trigger in our minds, well, who am I to say I’m a prompt marketer? But who is anyone else? We’re all figuring this out at the same time. So if you’re interested in this, own it and learn it, and don’t be afraid to market yourself that way.

Rob Marsh:  Even if you don’t take on a new title, even if you decide to stay a strategist or a copywriter or a content writer, whatever you are, adding this to your skills bucket, having that ability to engineer prompts, is going to put you ahead of all of those copywriters, content writers out there who are afraid to play with this tool, or who are afraid that it’s going to take their job and so they’re doubling down and saying, “It’s not good content, or it doesn’t create good copy.” And we’re seeing those comments all over the place.

But the fact of the matter is, if you’re using the tool properly, the way Sam describes here in this interview, but also in the training that he presented in the Underground where he literally demonstrates how to write prompts, how to add functions, how to get the right voice out of the tool, it changes the game. And so it’s really important to have that in your skillset moving forward.

Kira Hug:  I don’t have anything else to add here, other than I think this part of the conversation that we just listened to is just a great education on what we call things, prompts, sessions, how it all fits together, what a language model is, what it actually does. So this was like it was AI or ChatGPT 101, and definitely helped me get an understanding of what this is so I can continue learning and go a little bit deeper. So it was helpful. What else stood out to you, Rob?

Rob Marsh:  Yeah, two things. I think at the very beginning, Sam mentioned that AI is an infrastructure play. So unlike a lot of the stuff that’s been overhyped over the last decade or so, it’s being wrapped into so many different things that we’re seeing in our lives. If you use Netflix, if you watched Netflix in the last five years, you’ve already been using AI at some level. Or we’re starting to see facial recognition at TSA at the airport, right? It’s being integrated into all of these places in our lives already. And so embracing these tools and using them, the tools that have, especially for marketing for our clients, only makes sense because it’s just going to continue. This is not going to go away.

It’s not like NFTs where the market has dropped out of them, and people are laughing at that stuff now. The only way that’s happening is if the government comes and says, “Hey, this is an existential threat. We need to shut this stuff down.” Other than that, it’s going to continue. So I think we need to use it.

And then, I guess the other thing that I would say is, and I just want to underline some of the stuff that Sam said he was using ChatGPT in particular for, he mentioned looking at products, services, competitors, features and benefits, promises. And this is a long list. How something works, outcomes that you can promise, who the avatar is, their fears, problems, frustrations, what they want, what desires they have, the dreams that they have, research, ideation, brainstorming, figuring out human biases, psychographics, logic versus emotion, goals, outcomes, wants, hopes, dreams, features, benefits. There’s literally not a limit to the stuff that you can get out of this tool. And that’s why it’s so important to start learning and get this stuff into your skillset.

Kira Hug:  All right, well, let’s go back to our interview with Sam and find out how ChatGPT can write in your voice.

Sam Woods:  And one thing you said earlier, Rob, it is excellent at writing.

Rob Marsh:  Okay, let’s talk about that because there are some things, again, it being a language model, that it can’t do. It can’t tell my personal stories or the personal stories of the person who makes the supplement, right? But there are ways, not necessarily to jailbreak that process, but there are ways to get into those kinds of stories. So let’s talk about some of those.

Sam Woods:  And if you give it the raw stuff of your story with the right prompts, it can write it exactly in your voice, your tone, your style, your diction, and you’ll read it and you’ll wonder, holy (beep), excuse my language, did I write this, or how does it know? Because there are ways you can train it on a particular style if you give enough examples. And so if I take stuff that you’ve written, Rob, and I get something that’s representative of how you write and speak, so to speak, I can train it on that style and then I can ask it to write anything in that style, and it’ll produce anything in the style, and you’ll read it and you’ll wonder if you wrote it.

Rob Marsh:  Let’s say you’ve been collecting my emails for the last three, four years-

Kira Hug:  Sam has been collecting your emails.

Rob Marsh:  I’m sure you have.

Sam Woods:  I have, yeah.

Rob Marsh:  I’m sure you’ve got them in a file.

Sam Woods:  I have a separate swipe file, yeah.

Rob Marsh:  Or the Kira file or whatever. How many of those do you need to drop into ChatGPT in order for it to be able to… What’s the word count that it needs in order to really start doing that well?

Sam Woods:  The more, the better, but in reality, you can use as little or as few words as two or 300 words. So until you can train your own model, which is coming, you’re going to be able to train your own model where you can take all your content and you’ll have a Rob bot that produces content. But until then, what you can do right now today with ChatGPT, even with the limitation that exists, all you need is two or 300 words that truly represents your style, where you are using the emotional intensity of what you’re saying is high, and the specific word choices show up, and on something like two or 300 words, you can train it on that style and it’ll replicate your style flawlessly.

Rob Marsh:  Okay, so let’s talk about some of those prompts then to get it to write in a particular style, maybe give us an example or two. I think anybody who’s been following this has probably seen the example of the peanut butter and jelly sandwich stuck in a VCR, how to remove it instruction written in the voice of the King James Bible, which is if you haven’t seen it, Google it, look it up. It’s very funny. Or maybe they’ve seen some of the stuff that people have done in Samuel L. Jackson’s voice. Our friend Jill Clark Keys posted some of those on his Twitter feed. They’re also quite funny. But let’s say you want to write in my voice, or maybe it’s one of my clients that I’m working with and I want to write in his or her voice. What does that prompt look like?

Sam Woods:  I’ll give you the basis of the prompt, the principle of it, because it can read like anything as long as the key elements are in it. So even though I will tell you a prompt, you don’t have to copy my prompt word for word in order to get it right, you just have to know the direction you’re giving it.

So the direction you’re giving it is two things. One, you’re going to tell it to analyze a text, and the text is whatever, let’s say 500 words of your client’s voice that truly expresses the client’s style, voice, and diction and so on. You are going to say, “Analyze this text,” and then you’re going to say, “Based on the analysis, create a personality.” Then you say, “And then write,” and then you give a direction on what to write, using the style, diction, tone, and voice of the personality you just created. That’s it.

Now, I can give you word for word what it should be. And so something like this, a personality with a capitalized P, for the sake of just making it an entity, a personality is a paragraph describing the writing style, tone, voice, and diction of a written text. “You are a writing bot who analyzes a piece of written text and creates the personality. Analyze the following text and generate the personality.” And then you paste in the text, you hit enter, and then it’ll give you a profiling of the personality of the text, which includes notions about voice, tone, style, and diction, things that make up our personality in text. Then you will ask it to write, tell it to write anything.

One example I like to use them as part of a workshop I did is, “Please write two paragraphs about scuba diving in the voice, style, tone of the personality,” and then it’ll write about a random topic, or scuba diving, in the voice, tone, style, and diction of the personality that it just created. And then to finish the thought, you can keep having it write whatever you want it to write by referencing the personality it created, and it’ll write it in that style.

Kira Hug:  I think the part that’s confusing to me, as someone who’s new to this, is the prompts you’re giving that you’re sharing with us where you’re talking directly to the chatbot and the tool, how do you figure out what that is? Or can you not really go wrong? I know part of this is practice. You’ve had a lot more experience doing it, so you know exactly how to word things. But am I just figuring it out as I go and seeing what gets the best output, or are there certain words and phrases you really need to use?

Sam Woods:  If you were giving instructions to a junior writer, how would you instruct him or her?

Kira Hug:  I would expect them to read my mind.

Rob Marsh:  That is so true.

Sam Woods:  Assuming for a moment that they can’t read your mind.

Rob Marsh:  The prompt is you are a mind reader.

Sam Woods:  Exactly. I’m not trying to be facetious or anything, and I’ll explain more, but what I’m trying to explain is… It sounds weird because it’s not a human being. Literally and truly, talk to it like a human being and give it instructions as if you’re talking to it like a junior writer.

Kira Hug:  Right. It sounds like you’re talking to a child, the way you’re talking to it, which it sounds like that’s maybe the way you want to do it.

Sam Woods:  Yeah. You can talk to it at a college, Ph.D. level language, you can, and it’ll understand, so to speak. So you can be very complicated in what you tell it. I’m just doing it to make it simple and try to make sure that it doesn’t misunderstand what I’m having it to do. So think of this, if I was to ask you, “What did you eat last night?” That question, I can ask you that question in plenty of different ways, but the intention is still for me to find out what you ate last night. You need to approach ChatGPT and GPT-3 the same way. As long as you know what you’re looking for you can ask it in any way, shape, or form, and if you get output that’s not what you’re looking for, change the words and ask it differently.

Kira Hug:  Yeah, that makes sense.

Sam Woods:  Even though I give him prompts, I don’t want people to get stuck on the specific words. Sometimes words matter. If I wanted to develop a personality, I’m going to say personality. I’m not going to say, “You are a piece of a rock.” And so I’m going to say, I’m going to use words that make sense for what I’m looking for. If I wanted to have a personality, why would I not use the word personality? Again, I’m not going to tell it, “You’re a rock today and therefore you’re going to do X, Y, Z.” It’s language, it’s human language. How do you communicate ideas to another person, another human being? You use language, you use words, and if they misunderstand that, you change the words. It’s the most specific answer I can give, but it’s also vague at the same time, because I get it, people want to know what words should I use?

Kira Hug:  I think the way you were saying it sounded like a script, but I get it wasn’t a script. You’re just giving us examples, and we can use whatever words we need to use to be clear. So it makes sense now.

Sam Woods:  Again, think of it as a junior writer with some experience but less than you, and you’re just trying to give it direction on what to give it. And if I’m a junior writer and you say to me, “Sam, write 10 headlines.” I go, “About what?” And then you’ll say, “About X, Y, Z,” and then I’ll say, “But who is it for?” So you got to give it context. So when you ask it to do things, you got to give it context as to what you’re asking for, not just, “Write a headline about amazon.com.” It’ll write something, but it’ll be a dog. But if I give a specific direction on what kind of headlines and context around whatever it is, what it is, who it’s for, and then if I use adjectives like be descriptive, use emotional language, write like a human, be vivid, be detailed. But I would say that to a junior writer, I would say, “In these headlines, make them mysterious.” And I’ll say the same thing to ChatGPT and it’ll write what it thinks… Not what it thinks, what it predicts is our mysterious headlines.

Rob Marsh:  So Sam, when you get that kind of output, let’s say you’ve written a sales page or emails or whatever, how much rewriting then do you go back in and do? How much adjusting? I’m guessing the real answer to this is it’s going to depend on the input we’ve given it, but let’s say that you’ve done your very best. How often would you take the output and rewrite something, change the headlines, rework a story? What does that look like?

Sam Woods:  At this point, I don’t rewrite a ton. I used to rewrite a lot, but now I’ve gotten better at getting what I want. And also, you can have it keep rewriting the same email with different directions over and over again until it is what you want it to be. And so when it is as close to what I even have, I might spend 10 minutes telling it to rewrite the same email in different ways. And then when it gives me a version, a couple versions that I like, I’ll copy and paste it into a doc and then make some small adjustments and then it’s done. So the more you use it, the more comfortable you’d be about how to rework text.

There’s really no answer as to when you’re done with it. You’re done whenever you want to be done with it. I could have stopped earlier and rewritten more, or because it’s creativity on tap, essentially, I’ll just have it redo the same thing over and over again. It’s like a junior writer who never gets tired and never gets pissed off at you for telling it to rewrite stuff 100 times.

Kira Hug:  How do you use it in reference to your clients as far as you tell clients, “Hey, here’s one of my tools I use,” are you upfront about that? Do we even need to be upfront? I know it’s a personal decision.

Sam Woods:  It kind of is. It also depends on how you use it. I’m upfront and I tell them, but I tell them why and how I use it. So whatever I’m working on, it is always that I leverage different AI tools to help me uncover opportunities in copy or optimization that I otherwise would probably miss. Or I’ll say, “I’m going to leverage these tools to help me create variants and then make a selection as to what gives you the strongest possible chance of success and converting.” So when I talk about it and how I use it, I never say, “It’ll help me write faster,” because it’s not true. And I never say, “It’s the thing that I use and I just edit.” It is always, “I use it as an assistant, as a tool in the toolbox, but the outcome of what I deliver is still the same excellent copy that converts.” And so it is never about the AI and it’s never not about the AI. It is always how it helps me do a better job at what I’m doing.

Rob Marsh:  Let’s talk a little bit about some of the non-standard uses that you have for AI.

Kira Hug:  Let’s talk about the poetry.

Rob Marsh:  I asked ChatGPT to write a limerick about Kira, and it did a pretty good job. It didn’t quite get the rhymes right, but it was close enough. For the first attempt, it was pretty good. But let’s talk about this. So poetry, pulp fiction, tell us just how you’re playing around with it there.

Sam Woods:  I’m using it in all the ways you can write, or most ways you can write, I’ve been using it to do those things. So I’ve practiced my Shakespearean sonnet writing with it. I have pulp fiction stories selling on Amazon under a pen name right now.

Rob Marsh:  You can share the pen name after, afterwards.

Sam Woods:  I’m doing this as an experiment because it’s published on Amazon without any notice that it is done by an AI because I want to know, will people read it and buy it? And so when I’m done with the novel series, which is six books right now and three of them are published, when it’s done, then reveal and say, “Hey, this thing…” But the point is that I’m using it to write all kinds of things. Non-fiction essays, articles, sonnets, fiction, plays. It’s great. It’s perfect for ideation and brainstorming anything.

Rob Marsh:  What Sam didn’t tell us, Kira, before he got on is we’re actually talking to an AI Sam.

Kira Hug:  I figured. I figured Sam has always been AI. Sam’s not real.

Sam Woods:  I don’t know if I should be offended, or-

Kira Hug:  No, I think it’s a compliment. How are you using it outside of the writing world, if you don’t mind sharing? If you are.

Sam Woods:  Yeah. I’m trying to think of an example. I’ll use it for… You talking about stuff that’s not writing, no writing at all? So it’s great for doing SWOT analysis on companies for investing. It is great for analyzing quarterly earning reports and financial stuff, because it’ll tell you in ordinary language you understand. It’s great for research, and not just chat, but tools like CiteSpace, which is, I think it’s typeset to Io. Their tool for reading PDFs is awesome. So I’ll upload whatever academic paper I want to read. I’ll upload it there and it’ll tell me things about it. It’ll help me understand it. It’ll rewrite things. That’s just some examples. I could keep going, but it’s…

Kira Hug:  What about with relationships? Are there any strong use cases you’ve seen there?

Sam Woods:  Oh, absolutely. Yeah, whatever. I’m not going to say that I’ve done this, but what you could do is take text messages from your significant other and input them into the chat window, and then ask them to write 10 replies or 10 messages that expresses your love toward that person. And then you’ll have text messages done for you that you can then use or not use in communication with-

Rob Marsh:  How romantic.

Kira Hug:  Yes. So romantic.

Rob Marsh:  A romantic use of new technology.

Sam Woods:  Are you talking society level, or are you talking copywriting level?

Rob Marsh:  Well, let’s talk about copywriting level because I think if we go society level now we’re talking about mass extinction events and all of that.

Sam Woods:  Yeah, which is a very real possibility. So copywriting specifically, its pitfall is what it’ll do to you is that it’ll expose your thinking as either robust or flawed. And so if you give it unspecific and vague prompts, you will get really crappy output. And so if you find yourself interacting with ChatGPT one day and you just get crap back, then most likely it is because your own thinking and how you express it in words is not good enough, it’s not clear enough, it’s not strong enough in what you’re trying to say. And so it’s easy to get stuck generating text whether the text is good or bad. And it’s hard to know when to stop and when it’s time for you to do something with the text. Part of that is novelty. It’s easy to go, holy shit, I can just have it rewrite 20 headlines endlessly for hours and it’ll keep you giving me new headlines endlessly for hours. But if you have 100 headlines or 200 headlines, 1,000 headlines, how do you know which ones are good?

So you have to have, as a copywriter, you need to practice your eye for copy, if that makes sense. Your ability to tell that copy is either strong or not, or useful or not, or good enough, so to speak. So it’s easy to get stuck and not know when to stop. And it’s easy. The pitfall is, like I said, you have it write stuff for you and generate stuff for you and you start to believe it, if that makes sense.

It’ll be very accurate if you deal with language and it’ll be very accurate in terms of research around words people use and express, the fears people express and desires and so on. It’d be very accurate. But it’s like human beings. It is a very good bull (beep) artist at the same time. So we are good at just (beep) and it is also good, therefore it is also good at bull (beep) because if you keep asking it for stuff, it is not trained to say no and stop at some point. It’ll just keep giving you stuff to the point where it’ll do what’s called hallucinate, which is when this starts to make up.

So this is why I keep saying I don’t buy into it that this will replace copywriters. I also don’t buy into it that copywriters can go about their life without ever touching it. I don’t think that’s true, either. So the approach that I say you have to take is it’s a collaboration tool, and you use it as a collaborator and an assistant.

So those are some of the issues. And like I said, it’ll be clear very quickly if your thinking is muddled and muddied, because if you can’t express clearly in words that a machine can give you output back, then somewhere in your thinking or the way you express it, something is wrong or something is missing.

Kira Hug:  Can you talk about the impact it’s had on your business, your life? I’m imagining, well, I thought it would be faster, but you mentioned that it wasn’t, it’s not actually faster to write.

Sam Woods:  I mean it is faster, but that’s not how I want to compete in the world.

Kira Hug:  Okay. But how has it helped you in your business?

Sam Woods:  It has helped me recognize good copy faster, and it’s helped making decisions around what’s worthwhile to test and not to test. Because it’s faster in the sense that I could probably sit down and come up with 100 headlines in any given day, it would just take a while to do it, but this thing can do it in seconds or a minute.

So it is absolutely faster, but that’s not how you want to approach it and use it. You don’t use it for speed. Use it for creativity and ideation that can help you with. Use it for the rewrites and variance. I can take any client’s email or whatever, landing page text, and have it rewritten it in endless variations, and then I can select the variations that I want to keep. I might have it rewrite an email and I’ll give it directions to rewrite it 10 different ways. But out of those 10 ways, maybe only two or three of them are applicable to what my client is selling and who they’re writing for or sending emails to.

So it has helped me not so much, even though it is faster and helps me write faster, but that’s not the value of it. The value is not in speed. The value is that it’s creativity on top and on demand. Endless. And rewrites are endless. You can sit for hours at any given day and have it rewrite the same email 100 times and it’ll just do it.

Rob Marsh:  This is a really interesting point to me actually, Sam, because I hadn’t thought about this, but one of the ways that I feel like I’ve gotten to be a better copywriter over time is by looking at other people’s copy, evaluating it, critiquing it, and giving feedback. And in some ways that’s what you’re doing with the GPT robot is it’s giving you copy and you have to basically be able to look at it and say, “Oh, that part is really good,” or, “This is really flat here.” And in some ways, just using the tool as a feedback mechanism could make us better copywriters because we’re actually just practicing the art of thinking through persuasion and all of these little pieces of copy that, again, can take us days to put together when we’re doing it on our own.

Sam Woods:  And have it analyze your own copy. The analyses that I mentioned before, I do those, I do all that and a bunch of other analyses that I didn’t mention. I do all of that on my own copy when it’s done. If I’ve written a sales page or a page or an email for a client, I will ruthlessly run through those analytical tools on my copy to see where it falls short. If I do a sentiment analysis and I expect it to have a certain sentiment and then ChatGPT says, “No, it does not have that sentiment,” then whoops. Okay, so something’s missing. I use it on my own copy as much as I use it to write anything. It’s powerful for rewriting and analyzing your own stuff after you’ve written it.

Kira Hug:  Now I want to analyze my last few emails that I’ve written, see how I did.

Rob Marsh:  I’m going to put in my emails and just say to ChatGPT, “Why is nobody reading my emails?” And see what it tells me.

Kira Hug:  What can I do differently? What about for creatives listening, right? All the creatives. They’re listening. They’re like, “Now’s the time. I want to jump in. I want to figure this out.” What are a couple steps they could take? Because I feel like it feels overwhelming. There are already all these products, people are selling courses, there’s so much to learn. But what would you recommend as just an initial step or two?

Sam Woods:  Start using it. Actually, log into ChatGPT, use chat.openai.com, create an account, log into it, and start interacting with it.

Rob Marsh:  Would you recommend the paid version, or is the free version good enough for now?

Sam Woods:  I would recommend the paid simply because it gives you uptime and availability when the free version is luck of the draw. Because there are tens of millions of people using this every single day, they have a hard time keeping up with the load, obviously. So the paid version I would, simply because you will not run into the issue of having it crash on you. It’s 20 bucks, do it for a month, you know what I mean? It’s nothing.

Rob Marsh:  We know there’s going to be other iterations of it. It’s going to be more expensive to get certain things out of it, but for now it’s definitely worth the money.

Sam Woods:  The best way is truly to start interacting with it and start asking it questions, and tell it to analyze text, and tell it to rewrite text, and play around with it, because the only way for you to get good at it is to actually do the thing. I have a workshop course that helps people do stuff, but you don’t even need it. You can just start working with the thing. And if you like copywriting and language and so on, you’ll have a great time. If you’re an English major, even better, because you’ll have all this knowledge about English or whatever language you speak. If you’re a language major of any kind and you understand how to speak to people and communicate, then it’ll be easier for you to do so. And I think copywriters have an advantage because we know words, we know how to string them together. We know how to use language. Use that same skillset when you interact with ChatGPT and ask things, tell it things, tell it to redo something. Talk to it.

Kira Hug:  I love following you on Twitter to see all your updates, so definitely recommend everyone follow you on Twitter to see what’s happening, and the latest, and your perspective on it. I think you had one quote that I really liked. You said, “AI can either dehumanize us or make us more human.” So can you just speak to what that could look like maybe on a more global societal level, in a positive way? We could go down the dark side of it, but let’s end on a positive note.

Rob Marsh:  It’s all of those text messages to your loved one to be automated.

Sam Woods:  The love bot. Yeah. Anyway, so I think I’ll start with the negative because it’ll make the positives make sense.

Kira Hug:  Wait, I didn’t ask you for anything negative.

Sam Woods:  I don’t care. I’m Sambot. I’m Sambot and I do whatever I want.

Rob Marsh:  The bot gives you what it wants.

Kira Hug:  Do negative and positive. Okay.

Sam Woods:  Well, because I need to contrast it. So AI as a technology, and it’s really machine learning, but AI and machine learning, what we think of as AI, the capabilities are there to dehumanize us. And by that I mean it’s fully possible for anyone to use these tools to degrade themselves, degrade others, and for it to be used on a societal level to control us, surveilling us, and dominate us. You see that happen in other countries. The capabilities are there, and they’re at anyone’s fingertips.

The choice is, will they do it or not? And so, if the right people make the choice to not do that and to not go down the road of surveillance, dominance, and control and enslavement of who we are, then they can also humanize us, because the same creative capabilities that can wreck us can also help us grow and evolve.

And so it is possible for these tools to be used to create good in the world, and to share and channel love toward each other, to create new things. To create new masterpieces, whether it is in music, arts, any kind of art, film, photography, text. There will be a book written at some point, or several books written at some point, that when people read it, they’re transported and consumed by it and taken in by it and changed by it, and it’ll be written by a robot. And the same capability for creative evolution and creative enhancement used the wrong way can take us down a dark road that no one wants to live in. But because the capability is there for good and for love and for positive things to exist, that’s the choice that we have to make every single day.

I can use ChatGPT right now to create all kinds of awful content. I can hack it and I can manipulate it, and I can have it produce texts in this case that would make someone feel threatened and feel attacked or feel degraded. I can also use ChatGPT to create amazing texts that make someone feel good and uplifts them, and brings something good into the world. It’s all about how I use it.

And so what we’re going to see over the next few years, especially since it’s coming for all the creative work, we are going to see blockbuster movies scripted by robots, produced by robots, filmed and made by robots in all ways. And to a lot of people, that will matter and they’ll reject it. And that’s fine. To others, it’ll be like a revelation where they go, “I can’t believe that this is possible.”

But what we have to understand is these robots, so to speak, are trained on what we have produced in the world. So the only reason it can produce good things is because we have, as a species, produced good things. That’s also unfortunately why it can also produce bad things, because we as humans have made bad things happen throughout history.

So it’s really a reflection of who we are and where we are, and therefore we have a choice. Every day when you use ChatGPT or any other tool, are you going to use it to spread good things, or are you going to use it to spread bad things? Those capabilities are there, and it’s all in the choice you make.

It can produce amazing art, film, video, text, even dance choreography, anything creative it can produce, but it does so at the direction of us and what instructions we give it and what direction we give it. So maybe not more so than any time in history, but for the first time in a long time in history, the creative capabilities of geniuses is now available to anyone. Anyone, at least, with an internet connection and a browser and a laptop or computer. There are people in Nigeria, people that I know that I met online in Nigeria who are using these tools to create amazing copywriting, amazing apps, where 10 years ago they couldn’t, for whatever reason.

So we’re right on that line, we’re walking down that tightrope every day right now where the choice you make will affect whichever way we, we, as a society go. If enough people use it for good, then things will turn out good. If enough people use it for bad, then things will go bad, because the same capability and ability is there, available at a rate that wasn’t possible even five years ago.

Rob Marsh:  Probably a good place to end this interview with that warning and maybe a charge for all of us to do better, to make sure that we’re looking out for each other, and to use the tools properly. So Sam, thanks for sharing your genius, your experience.

Sam Woods:  My robot speaks, yeah.

Rob Marsh:  All that you know, and I have a feeling we may have you back to talk even deeper about some of this stuff as time goes on.

Sam Woods:  Anytime.

Kira Hug:  Yeah, this was fascinating, and such a great way to kick off this new podcast. You over-delivered, so thank you.

Sam Woods:  My pleasure.

Rob Marsh:  So that’s the end of our interview with Sam Woods. Before we close, I think there’s, again, a lot of really good stuff here, and maybe just a couple of things that are worth underlining, making sure that we’re understanding what Sam’s talking about.

Number one that stood out to me as we were talking there at the end, Kira, was I think there’s this thing where people are talking about how ChatGPT is going to make us all faster, faster at our jobs because it’s doing all of the writing work. And it’s clear that’s not really how Sam is using it. Yeah, it is really fast. It can generate several hundred words in a few seconds. But to really do it properly, to really dial in prompts, to get the voice, the stuff that you want out of it, it takes some time. It takes iterating, and almost a conversation back and forth with the tool. And the result of that isn’t necessarily that it makes you faster, but rather that you’re getting better ideas. You’re pushing, not just ChatGPT, but you’re pushing your own imagination farther, figuring out how to get deeper into all that long list of things that I mentioned as we’re commenting on the first part of the episode.

Kira Hug:  Yeah. And Sam gave us word-for-word what he says to his clients and how to position these tools. And so this is what he says, and we can say this, too. “I leverage different AI tools to help me uncover opportunities in copy or optimization that I otherwise could miss.” Or he says something like, “I’m going to leverage these tools to help me create variance and then make a selection as to what gives the strongest possible chance of conversion.” How could you say no as a client or a prospect? When you hear that, you’re like, “Of course I want you to do that. Use those tools,” rather than saying, “It’s going to make me faster, or it’s just an editing tool,” or diminishing what these tools can do. But it still puts Sam in the driver’s seat as the strategist, the copy chief, the thought leader. You’re working with the collaborator. And so that’s a really great way to position this as you’re talking to prospects and your clients about how you’re using these tools.

Rob Marsh:  And as we were talking, I had that realization. It’s like, back and forth with myself and this writing tool actually makes me a better writer because I’m able to look at that copy, critique it, what it’s giving me, what I’m putting back into it. And so using the tool, it’s not just another Word document or something like Hemingway where you’re just doing a quick check or an edit, but it’s actually a tool that can make you a better writer if you’re actually going to have that interaction with it. And again, I can’t emphasize enough that I think more and more copywriters need to be jumping in and at least learning how to do it for their own writing.

Kira Hug:  Yeah. And I like the way that Sam talked about these tools as a junior writer or just the collaborator, because I was definitely overthinking as he was talking through the prompts, and I was really learning about it with him. I was like, “Ah, how do you know what prompts to use?” And so when Sam broke it down and just said, “Just speak to ChatGPT like it is a writer who is sitting across from you and you’re just trying to give them directions, and give them some directions so they can nail the project,” that helps me understand how to use prompts, and so that was just a really helpful reframe.

Rob Marsh:  Knowing the Underground training that Sam did for the members who are there, he put in a prompt that didn’t have enough information and was talking about selling his stuff. And it was interesting, ChatGPT actually pushes back and just says, “Wait a second, I don’t know enough about that stuff. Please give me more details on X, Y, and Z.” And that’s exactly how he described working with the junior writer. It’s like if you’re working with a junior writer, say, “Hey, write me some emails,” of course they’re going to push back and say, “Well, who are they going to? What does it need to sell? Or what is the message?” And that’s exactly what happens with using this tool. It’s like having a writing team.

Kira Hug:  Yeah. And then Sam shared towards the end of the conversation other ways that he uses these tools. Of course, it was funny when he was sharing the way he enhances a relationship by using this tool to send text messages that really nail the messaging and the voice to a loved one possibly, or using it as a SWOT analysis, using it more for business strategy purposes. I even liked how he mentioned he uploads academic papers and summarizes them using this tool, which is so helpful. And so I think just thinking outside of the box about how we could use this is really helpful in our business, and even beyond our business.

Rob Marsh:  And in order to do those kinds of things, uploading an entire academic paper or even a long list of, say, client writings, emails, input or whatever, in order to do that kind of analysis, really better to have the paid version of ChatGPT. It’s only $20 a month currently. It’s definitely worth it in order to get that kind of output out of it.

Kira Hug:  And then we touched on the good and the bad at the very end of the episode. I don’t think there’s a lot to add because Sam was so elegant in the way he spoke about the upsides, the downsides, his excitement, his anxieties over it, and he just said, “The same creative capabilities that can wreck us can also help us grow and evolve.” And I thought that was a really beautiful way to end the conversation. We will talk more about the societal level and those concerns in other episodes. Definitely episode two we touch on that as well. But that was a great note to end on.

Rob Marsh:  Yeah, I agree. We want to thank Sam Woods for joining us on the podcast to talk about AI, how it’s changing the scope of not just the copywriting industry, but the creative industry as a whole, and maybe even the entire world. Sam recently presented, as we mentioned earlier, a training in the Copywriter Underground where he walks through the process and the exact prompts that he uses in ChatGPT to write an email sequence. He showed us how to define exactly what you need to get the output that you want so that it’s actually good enough to use without a lot of editing. You’ll find that in the Copywriter Underground membership that’s at thecopywriterunderground.com. And if you want to connect with Sam and see what he’s teaching about AI and copywriting, go to copywriting.ai or you can get his AI newsletter at bionicwriter.com. And finally, if you want to follow him on Twitter, like both Kira and I do, you can find him at Samuel Woods, and I believe there’s an underscore at the end of his name there.

Kira Hug:  Well, I don’t actually follow him on Twitter because I’m not on Twitter, but you could follow him on Twitter. I scope him out every once in a while.

Rob Marsh:  Get back on Twitter, Kira.

Kira Hug:  I am not getting back on Twitter. That’s another conversation. That is the end of this episode of The Copywriter Club podcast. The intro music was composed by copywriter and songwriter Addison Rice. The outro was composed by copywriter and songwriter David Muntner. If you enjoyed today’s episode as much as we did, and I really enjoyed it, please just give us a couple of minutes of your time and leave a review of the show on Apple Podcast. We really appreciate it. And we’ll see you next week.