#75 The Future of Prompt Engineering
20:13 Share

Hello friends and welcome back to the Future of UX podcast. I'm Patricia Reines and I'm super excited about today's episode because we are diving into something that's buzzing in the tech world, prompt engineering for designers. We are going to break it down, talk about why it's important and also give you some tips to get started. So

Prompt engineering is a super important topic. Last week, actually the week before I went to the US, I gave a workshop for a client on AI for designers and generative AI for designers and there prompt engineering was also part of the training. I think it's super important for designers to learn to know how to use large language models to really get the most out of it. And before we are diving into the topic, I have a little announcement.

If you are interested in diving deeper into prompt engineering, I have a masterclass coming up next week. It's free for the members of the Design Vision News community. So if you decide to join today, this week, you will be able to join the live masterclass for free. Next week, next Wednesday it is.

It will be all about prompt engineering for designers. We will do a super interesting deep dive, lots of learnings, lots of tangible things that you can use, lots of hands-on exercises. So I highly recommend to join. But this podcast episode is a little intro, so an amazing starting point. First, a bunch of information if you haven't heard about prompt engineering, if you want to get some basic insight. I think it's a very good starting point.

So in this podcast episode I'm going to cover first of all what is prompt engineering, why is prompt engineering important especially for us as designers or UX designers. We are going to talk about key skills and techniques for prompt engineering and we'll talk about some challenges and last section will be about a little outlook and some learning pathways. I would say

Let's dive right in and get started with what is problem engineering. So alright, what exactly is problem engineering? Think of it as figuring out the best way to talk to AI systems like chat GPT or even image generators like Midjourney.

It's all about crafting the right questions, the right prompt or command to get the best response. Because the way how you communicate with these tools is via text. And a prompt is basically a text prompt, a text command that you enter into the AI. The AI, this little black box, produces an output.

Before AI got popular, nobody really talked about prompt engineering, but now it's everywhere. For example, a few years ago, you'd have to use very specific commands to get anything useful out of an AI. And now you can ask JGPT to write a story, help with your emails, or even design a logo if you know how to ask properly. So a quick example, if you ask JGPT, tell me about the history of Rome, you will get a very general, very generic overview.

But if you say, summarize the key events in the history of Rome from the Republic era to the fall of the Empire in 300 words, you will get a much more focused and useful answer. And something that I'm...

noticing quite a lot and I think this is fascinating for me is when I talk with clients with people about the use of AI and JGBT and traffic perplexity these different language models a lot of people still say that the outcome is not useful because it's not good enough it's very generic and being like this is sorry for say it like this but this is really bullshit

The outcome or the result is highly dependent on the input. So everything that you put in, if you create amazing prompts, you will get amazing outputs. It's easy as it is. If you come up with very like generic, very simple prompts, of course you will get what everyone gets. So these tools are amazing. You will be able to use it. But of course you need to be a little bit more specific. You need to craft your prompts and you need to learn how.

So let's move to section number two. Why does prompt engineering matters? Why should we even care about prompt engineering, especially as designers? We are not engineers. You know, we are working with Figma. We are working with Miro. Why do we actually need that? Well,

First of all, there is a huge demand for it right now. Companies are also paying big bucks for people who can write effective prompts. Forbes even mentioned that some prompt engineers are earning around 300k a year. So those are like very niche, interesting positions.

And more and more companies are hiring for these roles because they realize how important it is to get the most out of their AI tools. And it's not just about their money. This skill is becoming crucial as AI gets more integrated into our daily lives and work. So you can see prom engineering a bit like knowing how to Google. Maybe you've had the experience that you got a question from some of your colleagues and you felt like you could just have Googled it.

Googling is not so easy sometimes because you really need to input the right information so that you get to the right output. Bromgeleering is very similar, but a little bit more complex. But there's the same concept. Bromgeleering is something like Googling. It's an amazing skill that you need daily for every position, for every job, for your private life, for your business life. So yeah, see it as an add-on skill, a little mini asset to your skill set.

I assume in the future that there won't be proper prompt engineering positions, but this is more like knowing how to Google basically, right? Okay, now let's move to the next section, key skills and techniques for prompt engineering.

Brownie naming sounds so boring and so uninteresting, especially if you are a designer, but this is so helpful to really get the most out of it. I can talk a little bit about my own experience and I am using Chachafriti for like one and a half years now. Every day, also perplexity, clothes.

And for me, it has been a huge game changer. Having this helper by my side, writing content for me, preparing assets for me, helping me summarize, helping me, especially with the UX design process. But to do that, you really need to come up with your own prompts. You need to be very strategic about it and you need to be very flexible and also experiment a lot.

So how can you get good at it? The first step is to be very clear and specific with your prompts. You need to provide some amount of detail because those details are important for the AI to understand what you want. And to do that, you first need to think what you actually want to get out of the system. I think it's super fascinating, especially if you use image generation tools, because this is a little bit more visual, so a little bit easier to understand.

So today I am so currently I'm working on an AI project with a client and we are preparing different scenarios for testing. We haven't integrated AI yet, so we are prototyping the whole experience. We can't really basically design the whole experience with AI, although it's an AI application.

So we are basically like, yes, how do you say it? Like tricking or withering of us the whole experience.

So we are preparing like different scenarios for the testing and prepare it. So I am preparing like different visuals to explain the scenario that the user is a little bit better. And to do that, I need to think about like what does the user really need to see on this visual that I'm going to create to understand how the scenario is. For example, like a user

standing with their phone on you know at a bus station waiting for the bus and I don't know

wanting to talk to people but being too scared to talk to them. Something like this where you're like, okay, how can I actually visualize that? This is not the scenario that I'm talking about, but I can talk about the project. So this is something else. But we think like, how can I actually get this message across, right? So this would be describing how like a man standing at the bus station, maybe looking slightly to the right to the people, looking a little bit scared of the other people are standing in a group talking.

You know what I really want to say with it? You have this idea of something that you would like to have, but you really need to do this thinking of how do you get this message across? What do I need to prompt the AI to get an output? How does that need to look?

And this isn't that easy, right? Because especially with a designer, we are used to like, okay, let's try things out. You know, let's put the button here and let's experiment a little bit. Let's place an image there and see how it looks and then test it out. And working by doing it with an AI, it's a bit different because you need to prompt it. You need to think ahead, basically. And this is a super important skill, something that will become super relevant with AI, really think ahead. Think

think what is the outcome how can i describe it to an ai and for this you really need to think about the whole scenario now we've talked about the basics of prompt engineering i would say let's dive into some practical tips that you can start using right away so here are two key tips for writing effective prompts my tip is and i think this is something that everyone can

takeaway from this podcast episode is start with a very basic problem and refine it. Test different rephrasing to see what works best. It's a bit like having a conversation where you keep tweaking what you say to get the best response. So you start with something, you see how the outcome is, and then you iterate. This is super important. It's about iteration. It's very unlikely that you get the first

prompt perfect right so you always iterate and then if you come up with a good prompt you save it somewhere that could be a notion doc or anywhere else where you have this prompt and maybe you want to reuse it at some point and the more you do it the better you get at like crafting your own prompts on the go and here's also another mini trick that I love

I call this the power prompt method. We are also going to talk more about the power prompt method in the masterclass that I'm teaching next Wednesday. Again, if you want to join, sign up for the membership. You're going to join for free then. The membership is something that I just started. It's basically an opportunity to grow where you meet like-minded people, where you get tutorials, hands-on work, where you exchange opportunities.

all about AI. So become a design visionary and then you will be able to join the free masterclass next Wednesday. So the power prompt method is basically thinking about crafting prompts at the building Lego, building with Lego bricks, right? You have those different power prompt elements and then you conclude them or build them together.

I always use the example of building a castle with Lego, right? Like if you want to build a castle with Lego, you can only use like the small red Lego brick. It's not going to work. You need different Lego bricks. You need different colors. That's also how prompting works. You don't need too many Lego bricks, right? Because then it gets too much. So with an AI or with the large language models, it's important that you have the right amount because you can also overwhelm the system.

and then doesn't really know what to do. So use the information that are relevant. Don't throw everything in it. It won't be sure what is relevant and then creates like crazy outcomes. So be very specific and detailed when you're writing a prompt. The most specific and detailed you are, the better the AI can really understand and deliver what you want. Think about including the context, one part of the power prompt element, maybe the desired length, the style that you want to use, any other relevant information.

Example, instead of asking "write me a story about", you could write a 500-word science fiction story on a set about a scientist who discovered a new form of life. Be specific. Because this detail prompt gives the AI a few instructions about setting Mars, the character, a scientist, and the plot, discovering a new form of life. And this specifically helps the AI to generate a more relevant and cohesive story.

Another tip is that you share a problem and then say, if you need more information, you

to come up with a good result, please ask me and you will see that the large language model is going to ask you follow-up questions. Something that you would like to or that you would expect from a human, right? Like if you give the task, write a story to a colleague, the colleague would probably also ask like, how long should it be? What should it be about? What should the style be? What is the goal with the story, right? So all these follow-up questions, super helpful, a great tip, something that you can use today.

And then another super important tip is really iterate and define the prompts. Writing prompts is often an iterative process. So start with a basic prompt and then refine it based on the AI response. This helps you learn what works best and improves the quality of the output over time. So for example, explain the importance of your X might be your initial prompt.

And then you start to add a little bit more information. For example, explain the importance of your ads in creating user-friendly websites and apps. And then you add also how it can impact user satisfaction and business success. So you experiment, you try a lot of things out. Don't be afraid to get your hands dirty. The refined prompt is more focused and usually really provides additional context. And this context is key.

It specifies the area of application, in this case, it's websites and apps and the aspects to cover user satisfaction and business success. And those are all things that the AI generate and help to create a more comprehensive and useful response. So section number four, the challenges and ethical consideration, also very, very important. Let's come onto some challenges. One big issue with AI is that we call it the black box problem.

You don't always know how it comes up with the answers, right? Like it's this black box. You throw something in, something happens and then something gets out of it. But you don't really understand how exactly it happened. And this is a bit tricky. Another really big problem is bias.

AI systems learn from data and it has that data that is biased. So the AI will be biased too, right? Like for example, if you ask an AI to create an image of a doctor, it might only show you middle aged men because that's what it's seen the most. Remember, an AI system has been trained on a lot of data and a lot of the data it has been trained on is data that's from the history, you know, from the last decade basically.

And a lot of things have changed. Luckily, we are so, I think we can be, we are so grateful that things are changed. But if you think about like 50 years ago, like 100 years ago, most of the doctors were men, right? So a lot of these images are still on there, on the internet that has been used with the labeling. And this is the problem, right? Then AI has been used in the past, but it should actually show the future, right?

Which makes it more difficult because you don't have images from the future. So you need to change the input, basically, right? You need to add things that haven't been there in the past. First of all, you need to be aware of these biases and then change the training data accordingly, right?

So it's super important to review what the AI produces and be ready to correct any biases. Also try to use very diverse and fair training data whenever possible. So if you're working on an AI product,

Ethical challenges will be a huge, huge topic. You as a designer will be someone who will be advocating for ethical considerations or someone who is going to raise problems that might come up, who's going to also check the data and then come up with solutions. Here's my recommendation.

Don't be the blocker in the process. Otherwise, the team will hate you. Raise these concerns, but also help come up with solutions, for example, right? Like prepare a mini workshop where you come up with solutions and guide basically the team and how to do that. How is probably something for another episode, but I think super, super, super important topic for the future.

So last section number five, what is the future outlook and what's the learning pathway? Looking ahead, prompt engineering is only going to get more important. AI will get better.

better at understanding what we mean, but we will still need to guide it with prompts. And if you're interested diving into this field, again, I have this masterclass next week, next Wednesday, where we are diving into prompt engineering. We're learning the power prompt principle. We're learning different prompting techniques. So everything is tailored for designers. So an amazing starting point for you to get into it, to learn it.

and to get your hands on it to really get better outcomes which is saving you so much time which is getting you better results especially if you're working with clients or with a designer to really yeah stay up to date super important at the moment

So I think that's pretty much it for today. To wrap things up, from engineering is becoming a super important skill for anyone working with AI. It's not just a trend or something. It's an essential part. It's a skill that you can add easily to your UX skill set. So take some time to explore those fields, practice your prompt writing skills and see how you can integrate it in your work.

Thank you all so much for tuning in. I am always so happy about the kind and nice messages that I'm getting from people. For me, the podcast is a project that I'm doing on the side beside my client project and social media and my newsletter and everything. So I'm so, so grateful that there are so many excited people about the future out there who listen to the podcast, who support the podcast by giving the

the podcast a five-star rating. I always read the comments and I always read the ratings and I'm so grateful for it. So thank you so much. I think that this community here is so supportive and so excited about the future. I love that. Just wanted to say that. Super, super, super, super grateful for you that you are listening, that you are supporting us and that you are driving the future into

a direction that we really want to go to. So thank you so much for being here, for inspiring the community, for thinking about the future, for getting ready. I really appreciate that. So stay tuned for more exciting topics in our upcoming episodes. I would say until next time and hear you in the future.