Episode Transcript
[00:00:05] Welcome to Prompting Curiosity, a podcast for the AI curious. No coding background required. I'm your host, Dr. Shantae Cofield, also known as the Maestro, and I created this show to explore what these AI tools actually are. Really, though, are the files in the computer, how to use them, and what they might mean for how we think, work, create, and move through life. Whether you're skeptical, intrigued, or already experimenting, you're in the right place. All that I ask is that you stay curious. All right, let's get into it.
[00:00:38] Hello, hello, hello, my curious people. Welcome to episode 44 of Prompting Curiosity. I'm your most gratefulest host ever, the Maestro, and today we are talking about how to become an AI whisperer. You already know that title is ridiculous. Uh, and what I really want you to take away from this episode is that prompting is super important, but also not at all that serious. So let me move this mic real quick. So I actually haven't been building much with AI over the past few weeks because I have been super busy with my messaging clients and doing that side of the business, which I'm loving. Um, but the fact that I haven't been building much is one of the things that inspired this episode. Right. I'm still using AI probably every day in some way, shape or form, which means that I am prompting it. But not, you know, not once have I stressed about giving it the perfect prompt. My guess, you don't do that either. But I still wanted to devote some time and some real estate to this topic because I got asked about it on, um, a podcast I was recently a guest on. Side note, if you want me to come on your podcast and yap about, AI will love it. Send me a dm. We'll set it up. Uh, my one ask is that you have an online scheduler. It just makes everything easier. But I would love to come and. Yep, about AI on, um, um, but I was on that podcast and they asked about prompting, uh, uh, and I also listened to an interesting YouTube interview with Anthropics Philosopher. Yes, folks, they have a philosopher on staff. And she discussed this very concept of, you know, becoming an LLM whisperer she was talking about. So you can listen to that. I will link that, that, um, that interview in, in the show notes. But to me, the most magical and the most impressive thing about AI, specifically LLMs, is that you can communicate with them, um, via natural language. You don't need to know how to code. You don't have to have a developer background. You literally just type in whatever you want, whatever it is that you want it to do or make or whatever or answer and it, it understands. That's incredible. And if you can't type or you don't want to type, you can just dictate. You could also have it in voice mode where it's going to like, have a conversation with you, uh, and like, it'll speak back to you. But you can also just, you know, use, talk to text and dictate. This is insane to me. I'm not sure that folks grasp how incredible that is. And I really do believe that this is the technological advancement of our generation, our lifetime. Like, this is incredible, all right?
[00:03:09] It removes pretty much all the barriers to entry outside of hardware. Like do you have a computer, do you have a phone? Like you have access to this thing.
[00:03:17] All right?
[00:03:18] But for the billionth time, at the 11/70 billion time, AI marketing and how it has been introduced to consumers is 100% backwards, right? The correct way is to do market research and then you see what people are talking about, what they have problems with, and then you build a product that solves a known problem. AI marketing, completely backwards. They said, here's a solution, go find a problem. To that end, despite the fact that this is so incredible, this technology, many people have no use for it, right? It is absolutely not for everyone. I was having a conversation with my, my bestie Jill, and she was like, I just can't understand the utility of it. I'm like trying to like, figure it out. And I'm like, it's not for you, right? It is not for her especially. And um, I'm saying this in the kindest way possible, if you really trust yourself. Let me flip that around this.
[00:04:05] LLMs are great for people who, who have a tough time trusting their own decisions, right? And they want a second opinion on something. It's phenomenal. It'll give you a second opinion. But if you're just like, I know what I want, I know what I want to say. I have ideas, I can iterate really quickly, ideate really quickly. Uh, I don't need to talk things out. Like, I've been in this business for a long time. I know my, my clientele, I know what I want to say. Just not for you. At least not that use case. Again, you folks know that I think that the, the best use case for, for AI and LLMs M, namely for the cloud code, is to build personal web apps. But what I'm trying to get across here is that it's not for everybody. As, as amazing as this technology is it's not for everybody. And I'm, um, I'm, you know, largely speaking to information workers and people who have online businesses because people may think like, oh, it's perfect for you, but just because you work with computers and you work online doesn't mean that you need AI or that it's going to be helpful for you. Right? Honestly, all the grifters that are out there that are selling these courses and trainings on how to use AI to like, make a zillion dollars in five minutes by having a bunch of agents do a bunch of stuff in your business, in your online business, it's whack. And it's just like, you know, the latest flavor of, of a, of MLM. So that's interesting. That's an LLM, but we're talking about MLMs with that case. But back to my main point, which is that, uh, for folks who are interested in AI or who do have a use case for AI, becoming this proverbial AI whisperer, AKA getting really, quote, unquote, good at using it, having it, you know, produce things that you want the first time around is, in my opinion, largely, if not completely dependent on two things. Number one, you, your ability to articulate and communicate your thoughts. And then number two, your willingness to go back and forth with this model. And the better you are at articulating and communicating your thoughts, the less back and forth you'll have with model. Right? It is not about having the perfect prompt. You know, it used to be a lot more, and I would offer that it was never that serious. I think that, you know, the tech bros just wanted to feel important and they just like discovered communication and they were like, wow, look, if you are better at communicating, then it does what you want. And you're like, I didn't know.
[00:06:10] Uh, but, um, it's. It's not about having the perfect prompt, right? It is just simply about knowing what you want and then being able to express it. Um, I do think it is important to not humanize these models though. I do believe in taking an extremely ethical approach to using them. And I am so interested in how this moves forward and the ethics around it and sentience. And I want there to be more philosophers and, and people that are in the field of ethics and you know, all of them working together to really, you know, come up with rules and laws and more discussions around this. But I think it's important to not humanize these models because, you know, I just. There's a line there. And I think that the more my attempt in not humanizing them is because I want to try and avoid, ah, AI psychosis, right, where people are just like, leaning into this thing, leaning on this thing so much. I think there's so much utility. I think that there are so many use cases. But I'm also like, you know, it's the both and.
[00:07:17] Right. So first part of that sentence is I think it's important to not, you know, superhumanize these models. But I also do think it's important to think about how you communicate with humans and then proceed similarly when working with an LLM. Right.
[00:07:31] It just makes you have a better, A better output. I'm trying not to give it 1 million instructions at once.
[00:07:38] Be as specific as possible. Ask it to confirm that it understands what you're asking it to do.
[00:07:45] Um, that's how you're going to get a better output.
[00:07:49] I will say that the context window, like how much you can put in and how much you can type. And the memory for these models has improved exponentially since the first LLMs were rolled out. We know this. So yes, you can type way more, you can enter way more than before. You can ask it to do way more than before. But as per always, the better the input, the better the output. Right? So a clearer, more organized, more streamlined input is way more likely to give you the output that you want and require less back and forth. We know this.
[00:08:17] So I'm in intention. I'm looking at the time. I'm intentionally going to keep this episode short because one, I don't have much to say about it, and two, I want to, like, make it meta and speak to the simplicity of becoming an AI whisperer, as nonsensical as that title is. Right. Um, but I do want to bring up one other point, which is the idea that these LLMs have a personality. Right? And this is something that Amanda Askell, I think, is what pronounces her last name. That's the philosopher and anthropic. She spoke about this in that interview that I linked in the show notes. But if we lean into this idea of becoming an AI whisperer, aka becoming really good at communicating with LLMs, part of that is understanding and acknowledging that each model does have a personality. Right. And yes, that, uh, personality changes when they release new versions of the model. We saw the incredible backlash that happened when, when OpenAI retired. What was it, GPT 4. 4. 4.
[00:09:11] Um, whatever model it was when people went crazy because it doesn't have personality and the new one didn't have the same personality. And I Bring this up as a data point to consider when communicating with your LLM of choice. Right. The more you use it, the more you learn it's, you know, personality, AKA how it's going to respond.
[00:09:31] And that can and probably should shape your prompt so that you get the answers that you want. My, my example, my, you know, my version of Claude Sonnet 4.7, that's my daily driver, it asks a lot of clarifying questions. Um, and yes, I could also change that in the instructions that I've given it, um, that are like, you know, the general instructions for it, but I like having that as the default. And so if I just wanted to answer something and no clarifying questions that I think it's going to ask, clarifying questions, I'll get ahead of it by saying, like, hey, don't ask me any questions. Give me an answer to this thing.
[00:10:05] Right? So again, just knowing who you're knowing, and if we're talking about a human, we're knowing who you're speaking to, right? Knowing their personality, their preferences, what they might say, how they might respond, and you can look to get ahead of it or just like, be aware of it.
[00:10:19] I realize that. Or hopefully none of this is revolutionary, what I'm saying to you. And to me, that's kind of the point. And I do think that it's worth highlighting, right? Using these LLMs is.
[00:10:31] It's becoming and it is just so second nature and it's such a regular part of one's workflow that. That in and of itself to me is worth noting. I, I made this podcast to create, to give myself space to just riff on this stuff. I find it fascinating just in all the aspects of it, not just the technology, but the, you know, sociological and anthropological implications of this.
[00:10:52] Right? And as we use the LLMs, like I just said, and they become second nature, right? I'm not here to predict the future or anything like that. I think it's. That's very futile. But if you take a step back and just look at the bigger picture, perhaps this is how market disruptions occur, right? When something is just so easy to use. But on the other hand, when it's easy to use and it's embedded in everything, does it then just become a standard and expected feature? Something like spell check, that perhaps doesn't disrupt any industries at all.
[00:11:24] Who's to say? I just like throwing this stuff out there, like thinking about this stuff. It's fun to me, right?
[00:11:28] But as related to this episode, all that to say, when it comes to using LLMs, nothing has changed in that the better the input, the better the output. The main limitations that exist when it comes to, you know, quote unquote, becoming an LLM whisperer and getting it to do what you want is how quickly, how clearly that's. It's ironic that. Right. Stumbled over the word clearly. How clearly you can articulate what you want, how willing you are to go back and forth with the model. And a third thing that I mentioned earlier, but that would be how much money you're willing to spend.
[00:12:02] Uh, because especially if you're using Claude. Ah, Claude does have token limitations. And if what you're doing is very token heavy, you will max that out, and you either have to wait for it to reset or you buy more.
[00:12:15] Um, this depends on what you're doing. If you're just kind of like talking back and forth, you'll be fine. But if you're having it build things for you.
[00:12:20] Um, I have not had it. When it was. It was. There was like a little time period where Claude was going through some, um, and people were reporting that. And I did hit my limit during that time, and I was like, wow, I've never hit it before. I'm not doing anything crazy. Um, since then, it has seemingly gone back to normal. But I've mentioned this before. Do I think that this is a way that Anthropic will ultimately look to recoup costs? Yeah, we saw. It's straight out of Uber's playbook, right? You. You subsidize things, you get people addicted to it, and then you jack up the price and. And people are addicted because it's so easy to use it. And so you're just like, well, I guess I'll pay more. But that would be the third thing. Uh, the third factor that would limit, you know, the third limitation when it comes to becoming that LLM whisperer, is that do you have the money, which I hate saying, I do not want this to become some exclusive technology, but, uh, uh, you know, do you have the money to spend if whatever you're doing is very, very token heavy? But the flip side of that is that the better you can communicate, the more clearly you can articulate things, the less back and forth you have, the less tokens that it uses. So we see there's like a little bit of a circularity there. All right, let's wrap it up. Let's go to the final section, and then we'll be done. Done. Which is how I used AI this week. Each episode I share a quick example of How I used AI AI that week. If you're new, welcome.
[00:13:39] This week, I use Claude to build my entire PowerPoint deck for a Vibe coding presentation that I gave to my MAFIA membership. I wrote the entire outline for the presentation, remember better the input, the output, uh, super detailed. I broke it up how I'd want, like, the slides essentially to be. And because this is how I was going to present it, it was, like, in order. And I just asked it to make me slides. I already knew my brand, colors and my preferred aesthetic, my preferred fonts, things like that, because I use it so much. And so I actually asked it first. I was like, hey, if I paste in my outline, uh, can you make me slides? And it was like, yes. And I was like, is it best to just paste it into the chat or should I, like, upload a markdown file? Like, what. What's the best way for you to read it? It was like, no, just paste it here. Uh, and so I pasted it, and it took maybe five minutes, downloaded the slides, opened it up, and it was done. I went in and I changed what I, you know, a few things. But, um, I have that much to change because it was 100% my words, right? It just made it pretty on the page. And I've never been one. I don't really like making. I don't mind making them, but I never made them very flashy. And I always just start my presentation off by saying that, like, I am the entertainment. The slides are the slides. Like, the words are on there. Um, but so to have it put the words on there for me, super helpful, right? If I can have the robot do it for me, I will have the robot do it for me. So that was. Was fun and, uh, super helpful. And speaking of Vibe coding presentations, I finally locked in the date and the time for that intro to Vibe Coding workshop that I've been talking about hosting. So I ran it with my MAFIA first, just to see, kind of test it out, and it went so well. It was so much fun.
[00:15:19] Uh, and that has allowed me to be like, okay, I can lock in it for the date and the time. So, uh, it's going to be Thursday, May 28th. So if you're listening to this on the day that it drops, uh, uh, it will be. This will be May 21st. So one week from today will be the workshop. It's gonna be at 12pm Pacific. It's gonna be two hours long. That is the middle of the day. If you're on the Pacific, it's around the west coast. And I did that intentionally so that your brains are fresh. It is tech heavy. I ran through this, and I will hold your hand through it. But there. There you. It does require tech tenacity. And, like, if it's not a type of thing, you're just gonna, like, sleep through it, right? It's $150. 10 spots only. You can use the link in the show notes to register. I will drop it. If spots are open, then they should still be open. I'm opening it up first to my. To the interest list. Um, but if the page is up and then their spots are open, uh, so two hours.
[00:16:14] Come and learn how to make your own personal web apps with Claude code. Learn how to 5 code. My promise to you is by the end of the workshop, you will have made your first web app app. I, uh, will hold your hand through it, and then you will understand this is the foundations, the fundamentals. And then you'll see the sky is the limit. And you can use everything that I will have taught you to go and make whatever it is that you would like to make.
[00:16:38] All right. That, my friends, is all for today. Hopefully you found this episode helpful if you did consider leaving a rating or a review. We hit 30 ratings, friends, and we got our first, like, not so good rating. I don't know. I think it's, like, two stars, probably from a white man. I said. When. I said there was no review associated with it. It's a rating. But let's focus on the positive because we did get a new review as well from EC Runner 34. Thank you, EC Runner 34. And that person said, I love this show. Shante has been so helpful and inspiring as I explore how to incorporate more AI into my business and personal life. The walkthroughs and examples of how she uses it are really helpful. She keeps it light and makes it all feel doable. I also love her personal takes and colorful language. Yes, this is my podcast. I'm gonna do what I want. And then the person said, keep it up. Five stars. Thank you, E.C. ron. That does these literally, folks. These explode my heart. Podcasting is such a unidirectional medium. I'm just sitting here talking to the screen, looking at my mic, talking to the screen. Rupert's here, my cat. He's here with me. He didn't even. He don't even sit next to me. He's sitting in his bed. And so to know that this is landing for you, you're finding it helpful. You're enjoying really does makes me really happy. So thank you for the reviews, thank you for the ratings. I know that there are some ratings on Spotify as well. Thank you for that. I really, really, really do appreciate it. If you have any comments, uh, you have any questions, you have any requests, just shoot me a DM at the movement maestro. I'm happy to hear them. I've had some good chats with some of you, and, um, I'm always looking for topics for the podcast. So by all means, slide on into the old DMS. Or if you want, you can send a text. 310-737-2345. That is my sideline. The text will be green. I don't have an Android. It's my sideline. Uh, but hit me up. I would love, love, love to hear from you. And yeah, last things, don't forget, I have a companion newsletter, the Curious Companion, that drops every Thursday. That is basically. And by basically I mean exactly, the podcast episode in text format. So if you prefer to read or you just want a written record, you can join the newsletter fam. You can head to prompting curiosity.com um, newsletter. Or you can check out the link in the show notes.
[00:19:00] As always, endlessly, endlessly, one more time. Endlessly appreciative for every single one of you. Until we chat again next Thursday, stay curious.
[00:19:15] It.
