← Everyday AI Podcast – An AI and ChatGPT Podcast
Everyday AI Podcast – An AI and ChatGPT Podcast
Creative Frameworks for Problem-Solving with Generative AI
00:27:49
Loading summary
Transcript49 lines
- [00:01]
Podcast Host
This is the Everyday AI show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business and everyday life.
- [00:15]
Jordan Wilson
Be honest with yourself when you go into your favorite large language model of choice, whether that's ChatGPT, Gemini, Copilot, Claude.
- [00:24]
Promotional Voice / Announcer
Whatever it may be.
- [00:25]
Jordan Wilson
You're probably just looking for an output, right? Maybe a quick answer to your question or something maybe you can copy, paste and modify and use at school or at work. That's probably not the best way to use it. If you've been listening to this show at all over the past three years, you've probably heard me rant over and over that you need to be using large language models as thought partners, as helping to augment your own intelligence, not just kick off and grab answers and try to move on as quickly as possible. I think those that are finding the best results, both individually, on teams and as companies, are those that are using large language models to solve problems in a creative way and to actually make their own human outputs better. So that's the topic that we're going to be tackling today on Everyday AI.
- [01:19]
Promotional Voice / Announcer
Welcome.
- [01:20]
Jordan Wilson
If you're new here, what's going on? My name is Jordan Wilson and welcome to Everyday AI. We do this every single day. It's your unedited, unscripted, daily live stream podcast newsletter helping everyday business leaders like you and me make sense of all the AI craziness that's happening nonstop. Hopefully help us learn it a little better so we can leverage it to grow our companies and our careers. If that's what you're trying to do, awesome. It starts here, but if you really want to take it to the next level, you're going to have to go to your everyday AI.com Sign up for the free daily newsletter. We're going to be recapping all of the important highlights from today's show as well as bringing you all of the AI news like we do each and every day. So if you want that, make sure to go check out the daily newsletter. All right, I'm excited for today's conversation. Have an very experienced, fantastic guest with in great background and we're going to be talking, like I said today, about using AI in a little bit of a different way. But enough of me chit chatting about it. I'm excited to bring on my guests. So live stream audience, please help me welcome Leslie Grande, lead Executive in Residence in the Executive Education program at the University of Washington. Leslie, thank you so much for joining the Everyday AI Show.
- [02:30]
Leslie Grande
I'm such, I'm such a fan of this show. I'm excited to be on it.
- [02:34]
Jordan Wilson
Fantastic.
- [02:35]
Promotional Voice / Announcer
As am I, like, excited to be.
- [02:37]
Jordan Wilson
On it every single day. Right. But, you know, real quick, before we get into it, tell everyone a little bit about your background.
- [02:44]
Leslie Grande
Sure. I had been an unconventional journey. I worked for 13 years in the film industry, and I'm a member of the Directors Guild of America. And after I left the film industry, I got into technology product management after getting my MBA at the University of Washington. And I worked for 25 years at companies like Apple, Amazon, and T Mobile. And at the end of 2022, I started thinking about writing a book because my experience in innovation really showed me how important it is for my teammates to have the creative confidence to really drive. Drive innovation and not just implement innovation. And I think before AI, people struggled with it, and they still struggle with it, but now they have a tool. And so one of the things I wanted to do is help people build that creative agency and creative confidence and use AI as a thought partner to do that.
- [03:35]
Jordan Wilson
I love that. And maybe could you walk us through, because we're going to get into some detailed strategies and really helpful frameworks, but maybe could you walk us through even some of your own personal, you know, findings? I guess through originally working with large language models. I've talked about mine for many hours over the last last few years. But, you know, what was it like for you and what were some of those initial aha moments in the earlier days of using large language models?
- [04:06]
Leslie Grande
That's a great question. Because I really felt compelled to integrate it into my process of writing the book Creative Velocity. And one of the ways I was compelled to use it was at the end of every chapter, I, after discussing a creative thinking framework, I provide exercises for people to practice with and without AI. And I thought, any reader who does this is probably going to likely put these exercises into AI to see what AI comes up with. So I had an insight problem where I had 15 facts, and the question was really, who lived in the red house? And so you had to use all these various facts and insights to back into who lived in each colored house. So I gave the problem to AI and it came back with an answer, and it was wrong. And I knew it was wrong because I wrote the exercise. So I wondered, how did it get it wrong? And it actually skipped three of the 15 facts. It took the first answer it arrived at, considered it correct, and the other ones, other factors were easily ignored. And so in the interest of speed, it came up with the first answer and the fastest answer wasn't the right answer. And I think that was the biggest aha moment for me was recognizing that speed triumphs over smart and comprehensive. And sometimes if it lasts a little longer to get a better answer, AI won't necessarily take that extra time. And so that was one of the really big aha moments, was not to trust the first output and to really question where did that output come from? And that process really helped me inform how I teach this topic in my maven course and at the uw because I think we are all primed for speed and so we'll go grab that first answer and run with it.
- [05:57]
Jordan Wilson
Yeah, it's, it's such a good point because I think when people are trying to show an ROI on Gen 8 AI, the default is to go to efficiency and, and, and productivity and to, you know, just do a task faster and not necessarily better. And I think what that means a lot of times is people giving up their agency. Right. But really maybe outsourcing to AI, one of their most valuable skill sets. So can you walk us through, you know, when it comes to still using and still leveraging your agency, how can you do that? What are the best practices to do that? As these large language models become more and more sophisticated, more and more robust, with all the scaffolding and agenta capabilities, how can humans still lean on their own agency?
- [06:50]
Leslie Grande
Well, that's an important question because I think we're so trained to do prompt engineering to go question, answer, question, answer. And I think that sort of creates a situation where we're willing to outsource our thinking to AI. And so what, what I'm trying to do is train people to use more open ended questions that are actually organized as creative thinking scaffolding as a way to navigate a problem space without jumping to the first conclusion. Because sometimes the, the first thing isn't the best thing or sometimes the most innovative thing is a place that lives further away from the space that you're in and you need to take the time to explore that. And I think time is the resource that gets eliminated when we go into the prompt engineering mindset. Yeah.
- [07:45]
Jordan Wilson
And maybe let's talk a little bit on prompt engineering. And I know it's an ever changing definition and now the trend is to call it context engineering. And I'm sure next year we'll be calling it something else. But know, maybe for our audience that is maybe not as technical. Why, you know, why is the engine like the prompt engineering process important? Why Is it important to. To iterate and continually improve upon what a model might spit out at first?
- [08:14]
Leslie Grande
Well, I think part of it is starting with a prompt that's open ended, right where the answer isn't, isn't pointed by the context you've been given. So look in this problem space for this answer really is a narrow way of thinking. And what I'm trying to do is broaden the space of the initial conversation before narrowing it. So zooming out and asking it in more generic terms, while it still will happen quickly, it does allow you to invite something that you hadn't thought of before. And so one of the first ways to do that is a technique called the generic parts technique, where you break a problem down to its most generic functional components. Not the features, but what every piece of that solution provides functionally. Because when you do that now, the space is defined in a more abstract way and you're able then to explore a part of that problem uniquely. So a great example of this is in my maven course, I had a problem where I as a consumer can't stand when I go to a hotel or an Airbnb and I can't log into my streaming accounts easily and I have to use some lousy TV remote to log in. And I just think that should be simpler because there's a million ways I could imagine it to be simpler. So when you ask AI to break that system down into its most generic parts, you start to realize authentication is the heart of that problem. And where else do I authenticate in that workflow that I could actually leverage so I don't have to have the problem at the pain point? I experience it by talking about it in those generic terms. It helps me associate another possible authentication solution into this problem space that wasn't there before. And so starting in a more generic way with the functionality breaks that fixedness that bias towards a specific spot for an answer to exist and allows you to look at the problem space more expansively. I think that's kind of getting at the point there, which is start by being more abstract and then get more fin.
- [10:17]
Promotional Voice / Announcer
Are you still running in circles trying to figure out how to actually grow.
- [10:20]
Jordan Wilson
Your business with AI?
- [10:22]
Promotional Voice / Announcer
Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Genai. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead and Some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for ChatGPT training for thousands or just need help building your front end AI strategy, you can partner with us too. Just like some of the biggest companies.
- [11:00]
Jordan Wilson
In the world do.
- [11:01]
Promotional Voice / Announcer
Go to your everydayai.com partner to get in contact with our team or you can just click on the partner section of our website will help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on Gen AI.
- [11:21]
Jordan Wilson
Yeah, and you know, you bring up some great points there. And even as I, you know, myself kind of think about the shift from, you know, prompt engineering to context engineering. Right, and what's the big step there? Right? It's bringing in more relevant context either to what you're working on, your team's working on certain data, right. That personalizes something for you or your industry. You know, what role, you know, in the prompt engineering process. How does this, you know, the human in the loop role continually change as the models change? And what should people that are maybe not just looking for the quickest answer and are looking to turn a large language model into a creative problem solver, how does their role continue to change as the human in the loop? And what should they be looking at in terms of improving on the original, original output that may or may not be right?
- [12:16]
Leslie Grande
That's the heart of the question. In the human AI partnership, AI has a very distinctive way of converging on something. And humans have divergent thinking. We're messy, we have emotion, things trigger memories that are associated with something that other people don't associate with that thing. And so what we value and what meaning we ascribe to things aren't innately in the answer you get from AI. And so being able to challenge the value of the answer as it relates to meaning and purpose is really a critical component of it, but also understanding the thought process that led there. So what's really important is to learn from how the answer was derived. What areas did you look at? Did you also look at this other area? And so asking the questions back so you get more understanding of the journey AI took to bring you those options also expands your thinking. So why did you think that? Where did you come up with that? Again, is agency over the output to make sure you find the meaning in it that it, that AI maybe unintentionally ascribed to it?
- [13:27]
Jordan Wilson
You know, I, I love that that Little phrase there, agency over the output. Right. For me, one thing I personally do when using large language models as a creative problem solver is well, I' off, you know, the memory and chat history and go into a kind of a private chat and I'll, I'll do two different ones always using a thinking model and I'll read the kind of chain of thought to better understand, you know, how the model is, is reasoning and, and how the model is tackling a problem. I'm wondering for you like what's been your kind of personal approach to this to really just turn a large language model from more than. Yeah, I'm just going to try to improve on an output and iterate to make it a right. How are you actually collaborating and what are the best practices that you're seeing to continue to collaborate with these models?
- [14:19]
Leslie Grande
Well, I, I absolutely use more than one tool almost all the time. I like to see how the answers differ and then exercises like the generic parts technique, you can run that same problem through two different tools and get two different answer sets, neither of them being right or wrong. Right. And so the idea that I'm not looking for the right answer, that I'm looking for the most expansive way to think about the problem allows me to get more opportunities because the language models don't approach the problem the same way. And so using more than one tool is, is, is kind of my standard go to method. I on the other hand, do exactly the opposite. I have one that knows me really well because I want to see the bias and then I use the other ones that don't know me that well where I have no shared memory because I do want to see to your point when it's shortcutting what it thinks I want because it knows me versus what happens when it doesn't. And it doesn't necessarily mean the one that knows me is better or worse. It just makes me realize it may have jumped over or leapt over some areas that it assumes I'm not interested in because of history. And so those areas now become more available to me if I use another tool where it doesn't have that history and memory. So using multiple tools I think is super critical. I even like to take the answer from one tool and put it in the other tool and I like to see how the other tool responds to what it saw as the answer I provided it. I'll say comment on this perspective and I'll give it the answer. And then it gives me some reason why it may not be the best perspective. So it helps open the door for more thinking.
- [15:54]
Jordan Wilson
Yeah, and, and I love that the first kind of framework we tackled, hey, just happens to be GPT, right? An easy acronym to remember in the generic parts technique. But maybe what are some other kind of creative frameworks that maybe our listeners have used or maybe that they haven't? What are some other ones that you kind of lean on?
- [16:16]
Leslie Grande
Absolutely. Let me just say before I go into that, that I did actually create a GPT for GPT so you can look in ChatGPT and explore the GPTs and find a generic parts technique GPT that will actually walk you through the process and teach you the process. So that's one of the other great things about it is you can build a GPT to teach somebody one of these frameworks. One of the most popular frameworks that I like to use is actually Scamper and it was developed decades ago by the O in bbdo, the advertising agency that was big and popular during the madman era. And it gives you seven specific moves to make in order to consider how you might adjust your thinking around a problem space. So SCAMPER is an acronym and it stands for substitute, combine, adapt, modify, put to another use, eliminate or reverse. And so when you think of that, a really great example would be, hey, I want to, I want to think of another way to make shopping convenient for people. And of course Instacart realized that you could shop online, somebody else could do it, you could pay for it and that it could get delivered. But curbside pickup became the really big revolution where somebody else did the shopping and then I went to the store, but I just picked everything up already paid for. Right. So these modifications or reversing the steps really can open the door for a whole new way of thinking about a problem space.
- [17:46]
Jordan Wilson
So yeah, the Scamper one is really interesting and I love, you know, using, whether it's, you know, copywriting techniques or problem solving, you know, acronyms in, in this case from pre generative AI that works great with today's latest technology. You know what does using something like scamper. Right. So let's say someone is trying to solve one of their business problems. I don't know, maybe they're in software and their churn is, is too high. Right. What is using something like Scamper, what might that help? You know, a decision maker stumble upon that maybe if they weren't using a frame that. But we're still using a large language model. Right. What is that maybe going to lead to that it might not lead to? If you weren't using a more structured kind of creative problem solving process.
- [18:41]
Leslie Grande
It's such a great question because I think one of the things this framework really does, it makes you rethink the problem space. Because I think we get so functionally fixated on how workflows and how people need to get from A to B in a process, or how people use your product as intended, and yet there's friction and people don't always behave the way you want them to. And so what really helps with these technologies is to ask the questions not in any linear sequence, but just to randomly pick these letters, like using the S in scamper. What could be substituted to remove friction around this step? A great example of another problem space is if you're looking at trying to make pins on debit cards more secure because everybody could watch you enter your PIN number, Scamper is a great way to say what could be eliminated to make that problem less secure, more secure, less insecure. What could I do that would be a modification of the flow that could remove the PIN altogether? What could I eliminate? So now I'm looking at very specific part of the problem, but I'm asking to use a technique for things I could explore that really smooth out the friction or add some behavior in that is more normalized for how consumers want to work. And so it pops up these ideas for you to imagine that you might not have thought about because you're so functionally fixated on how the process works or the product is intended to be used.
- [20:13]
Jordan Wilson
You know, one thing that I've always personally held a strong belief on is when you should use a large language model, and for what reason? Right. I feel very early on.
- [20:23]
Leslie Grande
Right.
- [20:24]
Jordan Wilson
Maybe in 2023, early 2024, it seemed like large language models were just kind of a content creation machine. Right. Just, you know, making short blog posts longer or, you know, helping you rewrite an email or something like that. Very, very small.
- [20:41]
Leslie Grande
Right.
- [20:41]
Jordan Wilson
Very small outputs. Right. And. And as we shift to more agentic models that can research and go back and iterate on their own while you sit there and wait. Right. So what? Maybe for the average business leader who is trying to also maybe re incorporate or better incorporate large language models to help with problem solving, strategy, etc. What are some of those kind of maybe agency unlocks or agency rewiring that we all need to do? Because it's not easy.
- [21:17]
Leslie Grande
Right.
- [21:17]
Jordan Wilson
Because sometimes it can be time consuming to go through these type of frameworks.
- [21:21]
Leslie Grande
Yeah. I think the things that I see people do that narrow their focus. And that LLMs can unlock one is you have cognitive biases. The way you think about solving problems is just your natural tendency. It's not a bias against a solution per se. It's a bias that may come from your expertise or your personal experience. And I think the psychological distance of LLMs is what they that's so critical in creative thinking because you get really attached to the problem space, the way you think about it, or you get really attached to the first solution that you came up with. But you at the absence of any other solutions, you move forward with something that may work but may not have considered all of the possible ways that solution could go wrong. And so testing your own logic and not being tied to your own solution is one of the best ways to use AI is to really say, I thought about doing this problem this way. What are other ways I could think about it? Which helps you break the bias for how you approach problem solving. Another cognitive bias that we often have is that we think we know who the customer is because we are the customer. And so sometimes you want to look at another domain where you're not the customer to see how they solved a similar problem because there may be a better solution that works in another place. And a great example of this is, right, the military often learns from how nature works, right? How do insects swarm? Right. And if you want to look at drone warfare, they learned a lot from how insects swarm, right? And so looking outside of your domain for the answer or outside of the competition for an answer takes a little bit of comfort with ambiguity because there may or may not be something there. But unless you ask, you don't know.
- [23:12]
Jordan Wilson
You know, you shared a little bit about your background, you know, working on the product side for companies like, you know, Amazon, Apple T Mobile. I'm wondering, right, if you had today's technology, how might, how might have your decision making or problem solving changed earlier in your career?
- [23:33]
Leslie Grande
Oh, it's so rich in so many ways, that question. Because I think about how experts guided my thinking in places that I went for the first time. So when I first worked in mobile, I thought everyone else knew wireless better than me, and so they were more likely to have a better answer than I would. Right? And so I fell prey to that kind of expertise because I was new there when in fact the reason I was hired was because I wasn't that person. And so having to get the confidence that being different in my, in my approach to problems was something I had to bring to the table. It might have been Easier if this neutral third party with psychological distance known as ChatGPT or Claude, would be the person in the room saying it instead of me. Because I think people might have heard it differently and might not have thought I was being irrational when I suggested something different than what had the way it had worked all along. And I think working in environments where it's worked a certain way for a very long time, retail, wireless, whatever. As an outsider, you seem a little less sensitive to why things work that way, and you're asking questions that people feel defensive about. But the thing that's so great again about AI now is that confront those conversations for me. I don't have to look like it's a personal comment. It's what a large language model with access to lots of data was able to come up with solutions that were ones that maybe you hadn't thought about. Should we just dismiss them all? And by the way, if you think they're all going to fail, let's ask AI what could cause these ideas to fail. And let's solve those problems too. And I think that would have been a huge difference.
- [25:23]
Jordan Wilson
I love that.
- [25:23]
Leslie Grande
I love that.
- [25:24]
Jordan Wilson
A little front end metacognition there. So, Leslie, we've covered a lot on today's show from, you know, how humans can still flex their creative agency in problem solving, prompt engineering, context engineering, and went over some of your trusted problem solving frameworks. But as we wrap up, what is the one most important takeaway that you have for our audience on how to best use creative frameworks for problem solving in the age of AI?
- [25:53]
Leslie Grande
Well, I think it's critical to use them to begin with because I think we're tempted to outsource creativity in the interest of efficiency and speed. And I think it's exactly the opposite that you need to do. You need to develop that creative confidence that then gives you the agency over the output to make it. To make it more meaningful and valuable and relevant to whoever it is that is going to experience that output. Right. And so whether it's for you and making a recipe out of what's in your refrigerator that you actually might enjoy eating and making that relevant to you, or whether it's solving a really big, big problem that your customers have in the workflow that you have in the back end that shows it sort of it's dirty laundry to customers in the front end, that problem needs to be rethought. And so taking agency over what those suggested options are and making them relevant and meaningful for your audience is critical. So building the confidence using these frameworks and then applying them in a way where you're actually challenging the output and making it better before running forward with it.
- [26:58]
Jordan Wilson
All right, well, some some great advice on how we can hopefully all start to rethink and use large language models a little bit better for problem solving. So, Leslie, thank you so much for taking time out of your day to join the Everyday AI Show. We really appreciate it.
- [27:13]
Leslie Grande
Thanks for having me, Jordan.
- [27:14]
Jordan Wilson
All right, and as a reminder, if you missed anything, any of those frameworks that she shared, it's all going to be recapped in today's newsletter. So if you haven't already, please make sure to go to your everydayai.com Sign up for that free daily newsletter. We'll see you back tomorrow and every day for more Everyday AI. Thanks, y'.
- [27:28]
Leslie Grande
All.
- [27:29]
Podcast Host
And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going for a little more AI magic. Visit your everydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.