Prompt Engineering Best Practices: Soft Skills [AI Today Podcast] AI transcript and summary - episode of podcast AI Today Podcast: Artificial Intelligence Insights, Experts, and Opinion

Go to PodExtra AI's episode page (Prompt Engineering Best Practices: Soft Skills [AI Today Podcast]) to play and view complete AI-processed content: summary, mindmap, topics, takeaways, transcript, keywords and highlights.

Go to PodExtra AI's podcast page (AI Today Podcast: Artificial Intelligence Insights, Experts, and Opinion) to view the AI-processed content of all episodes of this podcast.

View full AI transcripts and summaries of all podcast episodes on the blog: AI Today Podcast: Artificial Intelligence Insights, Experts, and Opinion

Episode: Prompt Engineering Best Practices: Soft Skills [AI Today Podcast]

Author: AI & Data Today
Duration: 00:10:59

Episode Shownotes

Generative AI is one of the most accessible forms of AI currently available. While in the past, you might have used AI without knowing it, you can use Generative AI purposefully in ways that have immediate and dramatic impact on your daily life. In this episode of AI Today hosts

Kathleen Walch and Ron Schmelzer discuss what soft skills are necessary to get what we want out of Generative AI. Continue reading Prompt Engineering Best Practices: Soft Skills [AI Today Podcast] at Cognilytica.

Full Transcript

00:00:01 Speaker_00
The AI Today podcast, produced by Cognolitica, cuts through the hype and noise to identify what is really happening now in the world of artificial intelligence.

00:00:10 Speaker_00
Learn about emerging AI trends, technologies, and use cases from Cognolitica analysts and guest experts.

00:00:22 Speaker_02
Hello and welcome to the AI Today podcast. I'm your host, Kathleen Maltz.

00:00:26 Speaker_01
And I'm your host, Ron Schmelzer. One of the interesting things we keep talking about generative AI is as if it's the only subject in AI. But of course, we've been talking about AI since 2017 here just on the AI Today podcast.

00:00:38 Speaker_01
AI goes back to the 1950s and we definitely didn't have generative AI in those first few waves and all those AI winters.

00:00:47 Speaker_01
Because for so many people, generative AI and prompt engineering and that whole space with large language models is their entry point to AI, that this is the place where a lot of folks get started with AI.

00:01:00 Speaker_01
I think it's important to understand how to really make the most out of generative AI because I think for most people, the way that AI is going to impact their lives and their jobs is going to be first and foremost through many of these tools that are accessible to everybody these days.

00:01:16 Speaker_02
Exactly. And Ron and I have been doing a lot of speaking recently. We've been speaking at a number of different sessions, including a lot of PMI events and speaking to project managers.

00:01:28 Speaker_02
And there's still a lot of fear when it comes to using, you know, large language models. And so that's why we thought it was so important to have this Prompt Engineering Best Practice series on our podcast so that we could talk about it,

00:01:42 Speaker_02
you know, really outline it in the way that we do at AI Today and how we explain things.

00:01:47 Speaker_02
And so for today's podcast, we wanted to talk about the soft skills of prompt engineering, because, you know, we've mentioned earlier that there really is no downside to testing and trying different prompts and really getting in there and using it and playing with it, because at the end of the day, it's just for you.

00:02:03 Speaker_02
You know, nobody else is going to be seeing this. You don't need to be getting help from others to run or have utilization, you know, with utilizing different things at your organization.

00:02:14 Speaker_02
So it really is important to understand like this is for you and don't be scared. Embrace it and you can move forward with this. And we always say, you know, think big, start small and iterate often.

00:02:25 Speaker_02
So continue to iterate on your prompts as well and continue to iterate and feel comfortable. You're not going to be excellent at first and that's okay. You know, everybody's learning. We always talk about how this is a journey.

00:02:36 Speaker_02
But soft skills are becoming incredibly important because, you know, generative AI really is accessible to the masses now. And almost anybody, you know, who has access to a computer, you don't even need to pay

00:02:52 Speaker_02
for an account to use many of these large language models. Yes, you can to get premium access, but to get, you know, basic access, you don't.

00:03:00 Speaker_02
So the great thing about generative AI is that it really doesn't require those hard skills that you think of, like programming, or math, or analytics skills, maybe database or data engineering skills that other parts of AI and technology do.

00:03:16 Speaker_02
It just requires us to be able to talk, right? And, you know, be able to type. So there's a lot of benefits that can come from this. You don't need to have classes in statistics and probability or calculus or linear algebra.

00:03:32 Speaker_02
You don't need to have a computer science degree. You don't need to be a data scientist. So this means that now a lot of people can do it. But with this, because you don't need those necessary hard skills, you need to have soft skills.

00:03:46 Speaker_02
And so what do we mean by soft skills and why are they so important? Well, soft skills are things like communication and planning, creativity, maybe interaction with others, you know, your communication skills.

00:03:58 Speaker_02
So these soft skills really are becoming so incredibly important when it comes to effective prompting. And these are things that, you know, some people have and that you can develop.

00:04:09 Speaker_02
But as we talk to a lot of project managers and people in this role, we continue to see, you know, why soft skills are so important.

00:04:19 Speaker_01
Before we dive into those soft skills and just go through them, I want to explain something really important here, and that is that when we are communicating with an LLM, we're communicating. Actually, it's exactly what we're using.

00:04:32 Speaker_01
We're writing prompts, which is the art of communication. It's not the art of programming, as Kathleen was mentioning. Traditionally, what we had needed to do to get value out of an AI system is we had to build models.

00:04:44 Speaker_01
We had to use the models specifically in the way that we built them. You might think that, well, we have computer vision and we have recognition systems. They seem to be pretty accessible. But the truth is that the average person really could not work

00:04:58 Speaker_01
with the average image recognition model. I can't just give it to a secretary or to an average person or a project manager even and say, hey, here's an image recognition system. Go use the model. They'd be like, well, how do I do it?

00:05:11 Speaker_01
Well, you need to load your Python library and then you need to take the image. You need to make sure it's exactly 256 by 256 and normalize it. Then you basically apply it to the model. And then you get back this sort of big matrix

00:05:26 Speaker_01
That is how most people have interacted with machine learning models until very recently. It's only really been in the past two or three years that the primary way that we interacted with models is simply by communicating with them.

00:05:42 Speaker_01
And this requires effective skills in communication. You might think, well, everybody knows how to communicate. Heck, I'm doing it right now here on this podcast.

00:05:50 Speaker_01
But actually, the truth of the matter is that the way you communicate has a huge impact because it's the only way that these systems work. The way you communicate has a huge impact on what you get out of the large language models. And in the previous

00:06:04 Speaker_01
episodes in this series, because if you're listening to this prompt engineering episode for the first time, this is actually the sixth of six in this series on basically doing prompt engineering.

00:06:15 Speaker_01
We spent the first talking about the best way to format prompts and prompt chaining and using tools and hack and track and all this sort of stuff.

00:06:22 Speaker_01
But really, a lot of that comes down to just figuring out how to communicate effectively so that you can get responses.

00:06:28 Speaker_01
And some people we've talked to said, like, well, that sounds a lot like English class, or it sounds a lot like a liberal arts education.

00:06:36 Speaker_01
We're like, yeah, now you can put that into powerful use because using those skills effectively actually allows you to get results from these systems. If you don't know how to do it, it doesn't matter. You could be a great programmer.

00:06:49 Speaker_01
And if you can't communicate effectively, you can't make prompt engineering work. On the flip side, you can be a poor programmer, and you can be excellent at creating prompts and communicating, and you'll get fantastic results.

00:07:01 Speaker_01
So as we go into these soft skills, we want you to think about it from the perspective of getting value from these large language models, how each one of these soft skills basically helps you get the value that you want from these large language models.

00:07:16 Speaker_02
Exactly. And it is all about, you know, continued iteration as well. So when we think about soft skills and what soft skills are needed to become effective and excel at prompt engineering, we think about things like creativity.

00:07:31 Speaker_02
So being able to think about the art of the possible when it comes to generative AI is incredibly important.

00:07:40 Speaker_02
And it helps you seek advice from others and be influenced by others as well, maybe who have written prompts that have been successful and how they've used generative AI to help them with various tasks.

00:07:54 Speaker_02
We had a previous podcast where with Sabi and we'll link to that in the show notes, but he was talking about how he was using generative AI to do things like help dress himself, you know, so that he looked color coordinated where he said, you know, if his

00:08:08 Speaker_02
daughter isn't around and he's not able to maybe pick the best option for when he's up on stage presenting, he uploaded an image of a jacket that he wanted to wear to Generative AI Solution and said, OK, help dress me. What color shirt should I wear?

00:08:23 Speaker_02
What shoes should I wear? What pants should I wear? And it provided some pretty great suggestions for him, especially with what he had already in his wardrobe. So this is just one example of maybe you wouldn't have thought to do that.

00:08:35 Speaker_02
But how are others at your organization using it? You have to be creative. Think outside the box. Say, how are others doing it? Or maybe, how are others doing it in different industries and how can I apply that to my industry?

00:08:47 Speaker_02
We always say, you know, learn from others. So really having that curious mind and being creative are important aspects of prompt engineering because you can, you know, don't be limited by your imagination and by your creativity.

00:09:03 Speaker_02
Really see how others are doing it as well.

00:09:06 Speaker_01
Yeah. So that sort of leads us to another soft skill. So it's creativity. Being creative is actually really the hard part. This is the art, as they say, in artificial intelligence. It's like really kind of never thought about doing it that way.

00:09:16 Speaker_01
It sort of leads you to the next real skill you need is problem solving, especially with AI systems that don't always respond the same way, right? Even to the same prompt, or maybe they're doing things that you wouldn't have expected.

00:09:30 Speaker_01
you have to figure out how to solve those problems. It's actually harder than an AI system, because it's not like debugging code, where you can be like, oh, I made a mistake.

00:09:36 Speaker_01
That parameter was the... No, sometimes just a single word, a single letter, a different way of phrasing the same thing could get you completely different results.

00:09:45 Speaker_01
So you really need to think about how you take the prompt outputs that are not necessarily what you want, especially the first time around, and you need to figure out how to iterate, how to solve problems,

00:09:56 Speaker_01
Figure out to refine that prompt until you get your desired output. Having really good problem-solving skills gets you very far with any sort of prompt engineering stuff.

00:10:06 Speaker_01
That's sort of related to adaptability, which is another related soft skill, which is how can you adapt to new circumstances, new situations?

00:10:17 Speaker_01
How can you adapt a prompt based on, well, maybe you need to take this prompt and maybe tweak it in a particularly different way to solve a different but related need? Maybe not do it the exact same way.

00:10:27 Speaker_01
You can't just take a template and say, well, I tried this exact same prompt in five different systems. Some of them worked, some of them didn't. It's like, well, maybe you could have changed it. Maybe you could have tweaked the problem.

00:10:37 Speaker_01
Maybe you could have thought about adapting things in a more creative way, right? Especially if the things are changing, the project conditions change. Maybe you get unexpected answers. Maybe the models have updated.

00:10:48 Speaker_01
Maybe you're trying a different model on a different system. Really having that sort of problem-solving and adaptability capability are some really key you know, soft skills.

00:10:58 Speaker_01
There's a few others that I think really help you with project management and doing it with prompt engineering.

00:11:04 Speaker_02
Yeah, you know, and also with adaptability too. I mean, sometimes even just, you know, slightly changing the words that you use can have a major impact on the results. And so again, this is something you shouldn't fear.

00:11:18 Speaker_02
feel confident, continue to try and test these systems, and feel empowered to go through everything that we talked about, all of the different prompt patterns, and maybe different platforms that you can use, and see how the results are.

00:11:32 Speaker_02
Another skill that's incredibly important is critical thinking. You want to use critical thinking, you know, to iteratively refine prompts. And it helps hone in on that, you know, effective questions or which prompt patterns that you should be using.

00:11:47 Speaker_02
So think about it and say, okay, which prompt pattern is going to be best for this situation? Maybe if you're unsure, try a few different prompt patterns and test them out because, you know, this also, they all relate on each other too, right?

00:12:01 Speaker_02
This goes to adaptability. This goes to problem solving as well and creativity. Are you using the right prompt pattern?

00:12:08 Speaker_02
Maybe you have assumptions and you think one pattern is going to be the best, but then you go and you use a different one and you're like, hey, I like those results better.

00:12:16 Speaker_02
Or maybe you're gonna aggregate results from a few different prompts that you've created. All of that is absolutely okay.

00:12:24 Speaker_02
The ability to analyze complex situations and make those informed decisions helps in designing effective prompts and then assessing the quality of those responses.

00:12:34 Speaker_02
Because another thing that we need to do is not always assume that whatever we get is correct. We also, you know, I mean, I mean, this you think is common sense, but it's not always. And so that's why we're saying this.

00:12:45 Speaker_02
Don't just copy and paste the results that you get from the prompt and use it somewhere. We had a conversation just yesterday with someone and they said, oh, and you went to insert here. and they copied that in. You gotta at least read the responses.

00:13:03 Speaker_02
And then they said another one that was a major giveaway was that it cited the sources down below with where it came from. So again, this is critical thinking, right? Or just thinking in general. Make sure that you are double-checking the results.

00:13:17 Speaker_02
Don't always take everything at face value. Dig one level deeper, even in the prompts, and this is going to help you get better at prompting.

00:13:24 Speaker_01
Right. It goes without saying that you need to be an effective communicator. I mean, if you can't put your thoughts into words, into a way that a system can understand them, you're not going to get good results.

00:13:35 Speaker_01
I mean, you know, most LLMs are, we think of them as being English based, but there are, LLMs can handle other languages too. So if you have a problem speaking fluently in English, you can choose the language you're more fluent at.

00:13:48 Speaker_01
However, if you are a native English speaker and you can't communicate well in English, Translation's not your problem. It's learning how to formulate your thoughts and explain things in a way that a system can respond.

00:14:00 Speaker_01
So becoming a better communicator, I actually see a renaissance in people taking classes and learning how to do things like write and speak and communicate and really do critical thinking and be in creativity.

00:14:11 Speaker_01
This may be a really good opportunity for liberal arts to make a strong comeback in the context of making AI systems work, right? And of course, the next soft skill is really collaboration.

00:14:21 Speaker_01
This is, I think, one of the most underplayed values in prompt engineering, which is that people should share the results they're getting. They should share their ideas.

00:14:30 Speaker_01
Sometimes you have to go to places like Reddit forums or community bulletin boards or forums or maybe even LinkedIn posts or Facebook or wherever, Twitter, X, I guess now.

00:14:42 Speaker_01
And you have to say like, oh, I did this with my, you know, LLM and it worked like this. And that's just how people are getting ideas.

00:14:48 Speaker_01
That's part of the creativity aspect but it's also part of the collaboration people need to get together and share the results are getting i see people starting to have spreadsheets of prompts that work in blog post and stuff like that if you're gonna be involved in using generative ai in your job.

00:15:04 Speaker_01
and using it for really critical things, you should be part of some sort of community where you can learn from others and not count on something you learned once in some place to work for you all the time.

00:15:15 Speaker_01
So collaboration and communication are really core to making all that work.

00:15:20 Speaker_02
And I know one of the things, oh, I was going to say, I'd like to see to, you know, organizations have these forums internally, you know, and have these these groups that people can go to for lunch and learns or just meetings.

00:15:34 Speaker_02
You know, sometimes people call them centers of excellence is how they're going to be adding to all of this as well, depending on the size of the organization. I think that there also is a lot that can happen.

00:15:46 Speaker_02
And so I'm interested to see kind of how that evolves over time. So, really what we say at the end of the day is that these soft skills are incredibly important, but just embrace that continuous experimentation and learning.

00:16:00 Speaker_02
Because when it comes to prompt engineering, there really is no such thing as failure, and that's important to kind of digest and understand. There's no such thing as failure. There's only experimenting and trial and error.

00:16:15 Speaker_02
So maybe it didn't work right the first time, or it didn't work as expected the first time, and that's okay, right? We're going to continue to iterate. We're going to continue to try. We always talk about embracing that growth mindset.

00:16:26 Speaker_02
You didn't know it yet, but continue to learn, continue to embrace it. Don't fear this, because if you are resistant to this, the rest of the world is not going to be resistant to this. And so, you know, make sure that you really are moving forward.

00:16:41 Speaker_02
We say, don't be afraid to come up also with some crazy ideas and test them because maybe they actually aren't so crazy. And most of the time, your prompts are going to be private unless you choose to share them.

00:16:54 Speaker_02
So if you wanna test your crazy ideas, go ahead and have that freedom to experiment and try new things without that ridicule or failure. And so there really is nothing to be fearful because you're not failing here. You're only experimenting.

00:17:08 Speaker_01
Yeah, I think it's really interesting because we see people using these generative AI solutions for all sorts of things I never thought of. I mean, like, it's crazy.

00:17:15 Speaker_01
You can use something to help you and augment your life in ways that you just never realized you could do that. It goes back to creativity, but also goes to experimentation. Somebody had to just try it out.

00:17:27 Speaker_01
It's like, well, I wonder what would happen if I took a picture of this and uploaded and asked a question or two about it. Turns out it works. Who would have knew? Who would have known?

00:17:36 Speaker_01
And the answer is, well, probably not even the creators of the LLM knew that. There's so many situations where people are like, you use the models.

00:17:42 Speaker_01
We want to see how people are using them because people are using them perhaps in ways that we expect, but we really want to see the ways that people are using them in ways we don't expect.

00:17:51 Speaker_01
And that's what this whole continuous experimentation and learning is all about.

00:17:54 Speaker_01
And part of that, when you're doing this over and over again, you're not just experimenting to see what will drive results, but you're also learning and developing expertise in prompt engineering itself.

00:18:05 Speaker_01
At some point, you might even be known to your organization of like, oh, that person is really good at writing prompts or figuring it out or figuring out the right way to coax the system to get what I want out of it.

00:18:16 Speaker_01
Or maybe feeling that, hey, maybe everybody's using chat GPT, but maybe Claude is better for something. Maybe Gemini is better for something else. Maybe Copilot is better for something. Maybe some other model is better for something.

00:18:27 Speaker_01
And as you develop that expertise, and as we're suggesting, you should probably try to develop that expertise in multiple different LLMs and multiple different versions.

00:18:36 Speaker_01
Because then you could sort of pick the best tool for the job, and you could be like, well, I know that if I ask this to this system, it doesn't do very well, which people might tell you.

00:18:43 Speaker_01
But you could say, but wait, I could take that same idea and ask it on this system, and it might do very well. So that's what part of it. These are all soft skills. These are not coding. As we said, this is not computer science. This isn't statistics.

00:18:55 Speaker_01
This isn't data science. This is really just practicing and pushing your boundaries for what you want to use these systems for. You know, a lot of times you can buy a monthly access for very low price and basically you have unlimited use.

00:19:09 Speaker_01
Well, there's limits, but for your purposes, for experimentation, you have very little limits. So might as well just use it when you have time, just Try different things.

00:19:18 Speaker_01
You can utilize those soft skills to continuously refine and hone your prompt engineering skills so you can really use these LLMs as that augmented intelligence tool, which is really the sweet spot for LLMs.

00:19:31 Speaker_02
Exactly. So as we mentioned earlier, this is podcast six of our Prompt Engineering Best Practices series. So we'll link to all of the other ones in the show notes. I encourage you to go back, check it out.

00:19:43 Speaker_02
Also subscribe to our newsletter if you haven't done so already. And I'll link to that in the show notes. You can find us on LinkedIn and you can find our newsletter on LinkedIn as well. where we walk through this also in written format.

00:19:55 Speaker_02
I know some people like to listen, some people like to read, some people like to do both, so we have it in a few different formats. But really, at the end of the day, just continue to experiment, continue to try, learn some of these best practices.

00:20:07 Speaker_02
This still is evolving, which is really exciting because you're not, you know, too late to the game on this. It's really here and now. So we say the future is now. So definitely go and try and experiment and take advantage of all of this.

00:20:24 Speaker_02
And also reach out to us. We love to hear from our listeners. Let us know what critical thinking skills you've employed while you've done this.

00:20:31 Speaker_02
And also, let us know how you've enjoyed this series, if you've learned from it, what you've learned from it, what prompt patterns you're doing.

00:20:39 Speaker_02
Also, let us know how you've been experimenting and maybe using large language models and writing prompts in ways that we haven't discussed on this podcast or maybe that you think are unique and our listeners would enjoy.

00:20:53 Speaker_02
You know, you can reach out to us on LinkedIn. You can email us at info at cognolitica.com. You can also go to our website as well. And I will link to all of the episodes that I discussed in the show notes. Like this episode and want to hear more?

00:21:06 Speaker_02
With hundreds of episodes and over 3 million downloads, check out more AI Today podcasts at aitoday.live.

00:21:13 Speaker_02
Make sure to subscribe to AI Today if you haven't already on Apple Podcasts, Spotify, Stitcher, Google, Amazon, or your favorite podcast platform. Want to dive deeper and get resources to drive your AI efforts further?

00:21:26 Speaker_02
We've put together a carefully curated collection of resources and tools, handcrafted for you, our listeners, to expand your knowledge, dive deeper into the world of AI, and provide you with the essential resources you need.

00:21:39 Speaker_02
Check it out at aitoday.live slash list. This sound recording and its contents are copyright by Cognolitica. All rights reserved. Music by Matsu Gravas. As always, thanks for listening to AI Today, and we'll catch you at the next podcast.