Transcription
You might have seen Google, Amazon, Shopify say that, hey, AI usage at work is now baseline expectations. Everyone is scrambling to future-proof themselves and try to keep up with the exponential change that's happening every single day. And we all find the problem of, okay, I've learned this, but now ChatGPT rolls out a new model, and all of that I've learned now can now be done with a click of a button. So how do I make sure I keep up? How do I make sure I know which tools are useful, which ones are just gimmick?
And as a framework thinker, because you're on this side of the internet, we're looking for that 80/20 way of understanding how we can incorporate AI into our knowledge work life so that we stay sane and stand out at work. It's not about the fancy syntax and the symbols you have to use, even though if you want the basics, here they are. Take a screenshot. It's also not about, do I use ChatGPT or do I use Claude? Which one's better? Should I use both? Should I use all of them? Those things don't matter. What matters is how you use AI to help you improve the amount of signal you have amongst all the noise.
And where people get tripped up is this fake notion of AI is here to improve efficiency. So everything needs to be quick. Here's what I mean. We currently see two types of people when it comes to using LLMs. There are those who want to put 10 words in and get a thousand words out. They think that is such an efficient use of their time; things just come out in seconds, and you can move on to the next thing. And then there are other people who put a thousand words in and get a thousand words back. On the surface, it sounds like, well, that doesn't sound very great, right? You might as well just write the 10,000 words yourself. But the result that you get from ChatGPT on this will be at such a higher quality than what you can write by yourself within that kind of time. And most knowledge workers are not using AI this way.
People are over-obsessing about the concept of time that, hey, I have this task; I need to get it done. And as a knowledge worker, you know that your leverage doesn't come from time, right? You're not being paid by the hour; what your value is is the quality of the idea, how far you can take something that others can't see. The fundamental of any company it comes down to differentiation: how does your company differentiate from a competitor? How do you, as one of the employees, differentiate yourself from someone who's similar? So you don't want to be thinking about time; you want to be thinking about the quality of your output. I also talked about it in this video here where I said that you can use AI to increase the quality of your input in order to help you improve the quality of your output. 99% of the people currently are stuck at this 10-word-for-a-thousand-word level.
And you see two extremes, right? One are experts who really know their stuff, and when they start to use ChatGPT, they see that, hey, you know, the quality of what it gives me is terrible. So I'm not going to use this; this is a waste of my time. I might as well just write it myself, and you're not leveraging ChatGPT the right way. Or there's the other side of the coin where you say, "Okay, well, you know, I don't really care about the quality. Let me just put 10 words in, and I get a thousand words out, and I check the box and move on to the next thing." Again, not a long-term strategy you want to be using, especially if you're a knowledge worker. The problem is most people are looking at ChatGPT as a source for answers, but actually what the 1% does is they look at ChatGPT; they look at AI as a way to refine the problem. Einstein famously said, if I had an hour to work on a problem, I would spend 55 minutes on the problem and 5 minutes on the solution, and most of us don't do that; we do exactly the opposite.
I recently did a guest lecture at a university, and I saw how smart these students are. They learn fast. They have great ideas. They are aware, and a lot of them are really articulate. The professor brought me in to talk about mental models precisely because the students are good students, and so they are fantastic at taking a problem and solving it. But they don't have experience coming up with problems, identifying problems, reframing problems to then get to the right answer. And you see this is exactly the problem we see at work as well because the boss gives us the problem. We say, okay, I'm going to tell you exactly how to do this. We don't even think whether the problem brought to us is wrong. Most employees are good students. So now that AI is here, how you are going to stand out as a knowledge worker is not going to be around how quickly you can give a good answer but rather how can you define the problem, reframe the problem, and make sure everyone is working on the right problem.
This is such good news, though, because what AI doesn't have is your experience, your judgment, your intuition that you've built over the years. And so knowing where the problem is, how to phrase it are exactly how it's going to help you get good answers out of AI tools like ChatGPT. Let me show you the three requisites for refining the problem. And then we'll use them in ChatGPT in order to go from noise to signal. And don't get drowned out in all of the things that people are saying. Know for yourself exactly what you're looking for. So the three things we want to look for when we refine the problem is number one, the assumptions. These are usually left unsaid, and when you bring them to the surface, it becomes a lot easier to figure out what is actually the problem. Second one is the five Ws, the consulting way of going from one level to the next; go from superficial to something a lot more fundamental. And then the third one is the alternatives that we usually think in one narrow, laser-focused way, but that prevents us from considering other alternatives. And when we're thinking about the problem, potentially reframing it, redefining it, we want to know what are the alternatives. So let me show you what this means in ChatGPT.
All right, before I show you, I want to just quickly mention the dictation tool on ChatGPT is so accurate, 99% accurate. So if you haven't been using this to type, highly recommend you do it because it helps you get into the mood of having a conversation with ChatGPT. And this is where you turn the 10 words to a thousand words into a thousand words for a thousand words because you get to do that back and forth, that iteration. Let me just show you. Hello, it's Vicki. Now, the only annoying thing is it uses the most common spelling and things like that. So sometimes it misses words like my name. So another tool that I used is called Whisper Flow. It's also a dictation tool, but it doesn't just work in ChatGPT; it works everywhere. Shout out to Terry for introducing me to this. And I use it now to just dictate everything. So let me use that as we go through this example. I'll put the link down below if you want to check it out. Highly recommend.
Let me use the example of my guest lecture. The problem that the professor shared with me was: share mental models with them because the students are so good at taking a problem and giving it the best answer they can, but they're not great at dealing with uncertainty, complexity, when things are not known, and when they are dealing with risk, how can they think through it when not all the information is there? So let's give this a try. I'll give some context text to ChatGPT so it knows what's going on, and then we'll use the three frameworks to see how we can go further with this problem. I'm giving a guest lecture to university students on how to use mental models to navigate complexity and deal with uncertainty. This group of students are smart. They are quick to learn. They, a lot of them are honor students. But the problem they face is that as they're studying in finance and risk management, in the real world, they won't have predefined problems. In the real world, they have to think short-term and long-term where things are uncertain, things are unpredictable. So how can I introduce to them mental models so that they have the tools to be able to think clearly even when things are uncertain, even when they don't know the answer. Now I hope you can help me better understand the problem that the students are facing. All right, let's put this in and see what ChatGPT comes back with first. All right. They're trained on the academic mind; they're entering a world that robots adaptive minds. Okay, that's pretty good. And the key problem they face is not a lack of intelligence, but but it doesn't have a "but." There's no lack of intelligence. It's the over-reliance on an environment where the right answer is known. Okay, great. So ChatGPT goes on to break down some things and tell me, okay, these are the things that perhaps we can share. Great. But what I wanted to do is before getting into all of this, I wanted to refine the problem first. So let's talk about some assumptions. We are assuming that mental models are great tools to help them deal with complexity. But there are a lot of unspoken assumptions that exist when we talk about that. Can you help me list out 10 assumptions that these students might not know because they are probably not exposed to mental models at this stage? Let's see what ChatGPT comes up with. [Music] So good, right? Mental models are not answers; they are lenses. You're already using mental models; you just don't know it. I love this. And this is actually how I started the lecture with is I gave them the $5 challenge. If you don't know what I'm talking about, I talked about it way back in this video here. And I show them, hey, you already have mental models on how to make money, how to how to understand the value of time, of work. So it's great to be able to articulate the assumptions so that we can meet our audience wherever it is they are. Mental models are not universal; they're situational. Interesting. I would reward this, but all right, it's helpful to have a first draft. The difference between simplifying and dumbing it down, also something we talked about in the actual guest lecture I gave as well, because everyone always worries, especially people who love to think deeply, they're like, oh my gosh, but if we simplify, isn't it oversimplifying? Well, there is a line, and just like we're saying right, how do we figure out signal from noise? We are simplifying, but we don't want to dumb things down. So as you can see already with just one prompt, we got 10 assumptions. Now the key here is you don't stop here, right? As you saw me go through them, some of them I already like as they are; some of them I want to tweak, and what I can do is continue to have a conversation about the things that I want to tweak, and I'll give it direction as to where to go. Okay. So that is framework number one, right? Getting ChatGPT to give you the assumptions. Okay. So now what we're going to do is now based on these assumptions, can you also help me go five levels deep as to why the students are not naturally thinking in this way? Use the five Ws and dig deep as to what is the fundamental reason for them not thinking in this way so that we can better structure the guest lectures to help them use mental models and shift their thinking behavior. Let's see. You're doing exactly what you're teaching, reframing the problem. Correct. All right. Why? Number one, because they haven't been taught to use them. Okay. Because their learning environment rewards correctness, not framing. Yeah, we got that far. Because uncertain feels like failure, not opportunity. Okay, this is already getting interesting, right? That this point of if I don't know that means I'm incompetent is so crucial to understand, and this is why if you've ever worked with me or attended any of my workshops, you know I play games; it's exactly to tackle this because uncertainty is not at all failure, and actually we, by playing games, can learn better. So I love the way it's phrased here, though I want to explore that more, right? Oh, because they're protecting their identity as smart students, right? I'm sure a lot of people are getting goosebumps looking at this one, right? Even when we are at work, we still have that identity so deeply ingrained in ourselves like I was a good student, so how come I'm not performing, I'm not a performer in the team? This is so good. Now, because they were trained to optimize for performance, not judgment. This is actually so this is a light bulb moment for me in the sense that, ex how do I phrase this, you know all of us are focused or overfocusing on performance, but how that judgment in the end, being able to say, okay, out of all of these options, I think this one is the best one to go with, even though I don't have a perfect answer, is what separates the top performers from the top doers, like people who are good at doing, operate like ChatGPT says in low-complexity environments. Okay, again, right? Just by using these frameworks already, we're going so much further with this idea, with this problem, and we're starting to see that actually, you know, mental models are tools, but they need a shift in mindset in order to be able to do these things. So very exciting. The next thing is about alternatives. So let me ask ChatGPT: this is a fantastic analysis, but can you give me some alternatives, as in what else could be true? Let's say yes, we're using inversion; good identification. Now I do have to say that I use ChatGPT a lot, so it knows me. It knows my frameworks. It knows my mental models. That's why it's giving these inversions and red teaming. So if your ChatGPT doesn't do that yet, it's it's not a problem; it's just because it doesn't have the memory of talking about frameworks a lot, and I talk a lot about frameworks in my All right. Here we go. Oh my gosh, I love this one already. They do use mental models but lack the vocabulary to name them, and so a few of you who have attended my workshops or worked with me in my accelerator mentioned this, right? Is that, oh, now I know what to name it, and because I can name it, it becomes something easily recalled, and I can build on it. Before it was just this was my process; I wasn't really aware of it. But now that I am aware of it, I can do so much more with it. Fantastic. All right. They resist mental models because they associate them with consulting jargon or elite circles. Not real not super relevant for this group is my hunch because they are in finance; they are in risk management. But okay, I'll keep that in mind. They've never seen real-world ambiguity modeled by authority figures. This is so interesting. Okay, this is something I don't know. Now I'm just getting so excited. I am veering off not just about talking about reframing problems but also the things that I think every knowledge worker should know, which is when it comes to articulating complexity. Everyone thinks that smart people are just naturally articulate because they think fast. But actually, as I keep on saying this, the best authors when they give interviews is not because they can think on the spot; it's because they spent years writing out these ideas, refining them. So by the time they're sitting in that interviewing seat, they already had years of thinking about this. So of course they can pull out anecdotes and ideas and concepts really easily, right? They put in the work; it's just we don't see it. Oh, this is another really good one and a reason why people give up. So I'm not I'm getting too over-excited, but you get my point, right? Just by not focusing on the answer but focusing on understanding, refining, reframing the problem, you will get so much more depth out of ChatGPT. It really is a good student. Depending on what kind of question you throw at it, the answers it gives will be very different. If you got excited by this, I know you will enjoy my 17-minute AI workflow on how to improve your input with AI, so let your output improve.