#36 The Good, The Bad, and The Ugly of Prompt Engineering in Generative AI
16:51 Share

What is prompt engineering and will it be important for us as UX designers in the near and far future? This is the topic of today's podcast episode of the future of UX.

Hello and welcome back to the future of UX, the podcast where we dive into technology, user experience and basically everything around the future to make sure that you are up to date with everything that's going on in the tech and in the design industry. So today we are talking about a topic that's been buzzing in the tech world, prompt engineering and generative AI.

But first of all, let me introduce myself. My name is Patricia Reiners. I am a UX and Innovation Designer and I'm running my own business in Berlin focusing on future topics, consulting clients in user experience design, innovation, everything that makes them ready for the future.

And in this topic, we will talk about the whole topic of prompt engineering. Before all these AI tools really got so popular in the last couple of months, I haven't really heard much about that topic or about that expression of prompt engineering at all. But now I am seeing more and more companies hiring

prompt engineers or prompt designers so those are job descriptions or job names right and a couple of days ago even found an article from forbes with the headline ai prompt engineers earn around 300k salaries here how here's how to learn the skill for free

So what I think is super interesting there is that there is a huge demand at the moment for writing prompts for people who know how to manage good prompt writing, how to do it. But what's for me even more important is to have a look into the near future and far future and see where the whole job description or skill of prompt engineer is going to head.

So in this podcast, we will talk about what prompt engineering is. First of all, we will talk about some good things from the UX perspective that I think are really valuable to the user experience. And we will also talk about some problems and some things that are not so ideal with prompt engineering.

And in the end, we of course will focus on the big question of how important the skill will become in the future and if it's really worth for us as designers to learn the skill or to even niche down and become a prompt engineer, UX designer or prompt UX designer or prompt experience designer, something like that. So first of all, what is prompt engineering? Let's start with the first question.

Prompt engineering is like teaching a computer how to give better answers. So when you ask a question or give a command to a computer program, the way you phrase it is called a prompt. So engineers work on making these prompts better so that the computer understands what you want and gives you a helpful answer.

You've probably all experienced that when you open up chatGPT and you want chatGPT to do something, you need to write a prompt, something, some kind of like description of the desired output that you want to get out of this, yeah, of this interaction.

And writing this prompt is something that's super important because you need to add certain background information, you really need to describe how the outcome should look like. And the way how you phrase these prompts really makes a huge difference in the outcome that or in the result.

So engineers really test different questions and settings to see which ones work best. And these really help make computer programs like chatbots much more useful and easier to interact with. So let's focus a little bit on the good thing of prompt or prompts or prompt engineering or prompt design using prompts for service or software solutions.

We defined what prompts are, so we are all on the same page. And now what are the really good thing about having so many text input fields and different kind of software or different kind of tools, not only TechiePG, but all the image creation tool like Midjourney, Firefly, Dallee, everything you describe with a prompt, all the AI chatbots,

All the AI plugins in Figma, for example, where you just write down what kind of wireframe you want and then it creates everything for you. You need to write this down. It's also like a prompt as well as all the integration in our day-to-day tools, designing websites, etc. So what's really great about it? I think generative AI and especially prompt-based conversational AI like ChatGPT

has been definitely game changer is probably like a little bit too extreme but has definitely changed a lot of how people interact with a software or a service

or it still is for a lot of interfaces are pretty difficult. When you want to get a certain result, you definitely, you need to know all the single steps that you need to take to get to a certain result. For example, you are opening up a design tool like Figma, for example. You want to design an interface for a library app. You need to know a lot of things to get to that point. You need to know how

how to design the single elements, you need to design how to, you need to know how to color them, how to edit typography, all these kind of things. So there are maybe like 100 separate steps from knowing what you want until the desired outcome. And entering those text prompts where you enter, hi, I would like to have an interface for a library app,

You only describe the result and then the machine designs or the software designs or creates the outcome and you don't need to do all the single steps in between. So I think this is super interesting for users who are not so experienced using these kind of tools. It's much easier to try different things out. You also save a lot of time.

So this is definitely good and also very innovative, right? So this definitely changes a lot of the ways that we interact with software, for example, right? Because you don't need to go through a certain interface, look through different kind of content, but you just enter what you want and then you get to the result in the best case. What are some challenges? Of course, not everything is super great and amazing.

What is definitely a big problem and a lot of experts are also talking about that is that a lot of the AI systems are black boxes, you know, really making it hard to predict the outcome.

So the first thing there is that you don't really know what are the single steps that the machine takes to get to certain results. The data in the background that is used can be biased. Thinking about mid-journey, for example, I talked about that several times in the podcast.

You don't really know what kind of data they use to train the software or not for all software. Sometimes they are pretty open about it and very transparent like Adobe Firefly, Adobe in general, but not all companies are like that. That's definitely a little bit tricky that this really feels like a black box. And the second thing is the whole topic around prompts. You need to describe what you want.

There's a huge difference when you say, please design me an interface for a library app or design me an interface for an app where I can store all my books that are written. I can highlight certain pages to reread them afterwards.

the tool automatically writes me summaries so where I'm going much more into detail what are the certain features that I would like to have what is that actually about then just like being really open and letting the software in the end choose how this app looks like so defining the prompt

is a big challenge and we are also seeing that with chatgpt there was a huge hype in the beginning everyone used chatgpt but now a lot of people got a little bit frustrated because not always because the results are not always that great they are very basic

They are not actually what they wanted and that really frustrates people. And the reason why these outcomes are not that great is because the prompts are not great. And writing good prompts takes a lot of time and also has some kind of like a learning curve. So you need to learn and understand how you actually craft these prompts, especially for tools like ChatGPT.

A few other problems that I still want to mention, I don't want to go too much into depth here, but I still want to mention that, is the whole topic of accessibility and inclusion. There is definitely an elephant in the room here, which is accessibility, right? So what happens to users who can't write well-formed sentences or to those who face any kind of language barriers?

Imagine an elderly person who's not tech-savvy trying to interact with like a prompt-based system. They might find it difficult to get the help they need, leading to frustrating experiences. There are not only tech-savvy people out there. There are a lot of elderly people, people with accessibility issues. And there I think it's super important to also think about them.

Of course, all the tools that we are seeing there are more or less like first generation tools. So of course, they will be improved and accessibility is definitely a thing where most of these tools need to be improved. Right. Okay. Now let's go to the last point, the future. So what I really like is to compare things a little bit.

And the whole AI revolution at the moment really reminds me a little bit of Google, because Google really revolutionized search. And I think that we need a similar leap in the usability of generative AI. So in the early 2000s, we really needed specialized query skills to find academic papers. And Google Scholar or Google Search really changed that.

And with Google search, you can go through the internet, you can find everything. It's super easy. Everyone can do it. It's very simple. And we need something very similar, like a similar transformation in AI.

So better usability isn't just like a nice to have, it's a competitive advantage. And there it also gets really interesting for us as UX designers because I am seeing so much potential in many of those AI tools

who are not using basic usability principles. Why not? Because they don't hire UX designers. They have like a really small percentage of their employees are UX designers. They are not...

seeing the urgency at the moment so interesting field also for us as designers to focus on these things because there will be a huge demand in the upcoming month and years to really rethink redesign these tools at the moment come up with better patterns and for that you need to do a lot of testing and research and understand how to interact with ai

So going a little bit back to the topic of prompt engineering, prompt designing, is that something that we as UX designers need to learn? That is a super important skill for us. I think for the near future, it's an absolutely amazing skill to have. Not only to be more productive in your work and in your environment, also for your personal structure,

There are a lot of studies that show that you are much more productive when you use these tools, especially like ChatGPT. So I think it's super important to understand how to frame good prompts. I would absolutely dive into that if I would be you.

But I also think that the far future will look very different. So I think in the far future, we won't have difficult prompts where we need to define everything and be super specific about...

what we actually want to get out of it because that's too difficult for people and we are already realizing that. So interfaces will change with, I assume, certain things that are pre-selected or that you already see in the interface like the entire GPT that could be like the tone of voice, for example, or how many words you want, you know, some kind of like a pre-selection and then also a little description, but not everything.

so in depth that you would call it prompt engineering or prompt design. So for those of you who are considering a career in prompt engineering or prompt designing, who read the Forbes article and want to earn 300k, remember that the landscape is ever evolving. So there is a huge demand at the moment, but this will probably change in the future, at least I guess.

What will definitely bring you high paying roles in the future is to focus on AI and really know how to use AI, how to design with AI for AI products. There's definitely a huge demand already starting. I'm seeing that on LinkedIn, on all the job platforms that a lot of companies are specifically looking for AI UX designers. Super interesting, right? So...

friends i hope um you got some insights there and if you want to learn a little bit more of like how to design prompts what are good prompts i created a few um freebies that are interesting about chat gpt i will link them here so you can download them about certain um i would say like prompts that you can use for chat gpt to be more productive but

One thing that also might be really interesting for you is the AI for a designer course that will launch late summer/autumn where we focus on these topics specifically and where we just really talk about these things and where you can get AI as a skill to your skill set and really keep up to date with all the things that are happening in that direction.

Okay, friends, I hope this was helpful to you. I wish you a wonderful week. Hopefully hear you next week. And of course, my friends, hear you in the future.