How to Use AI Without Letting It Think for You transcript

Hello everyone. Uh welcome back to my Code to Care video series. Uh today I thought I would do kind of a different sort of topic, which is how to personally use LLMs. Uh and I don't want to be too preachy, but I'll at least tell you how I use LLMs in the process of of writing something or thinking about a particular uh topic. And I have a strong belief that writing is thinking. And by that I mean I'm careful to not use an LLM to think for me.

And so I'm careful not to use an LLM to write for me either. But I do use it in the thinking uh process in a way I'll describe. But first of all, I'll give you a few dangers, a few studies that have shown the LLM can cause some problems with um uh with your thinking. There was a a pretty good study that showed students when they used I when they used AI, their accuracy on homework improved 48% and when they stopped using it, it declined it was 17% worse than the study group of students that didn't use AI at all.

So AI kind of eroded your ability to do your uh homework when they when you used it too much. There's another MIT study which was interesting. Uh and it said that 83% of students after they wrote something couldn't quote their own work when asked uh to give us quote from an essay that they had just written. Uh and you might say, well, everybody does that. It's hard to remember what you just did. Um but that was in contrast to 11% uh which was the group that didn't use AI and then just wrote their their essays by hand.

Uh and another uh a in the healthcare industry was with uh They were using AI to identify particular problems with patients and their identification rate was 28%. Before and then while using AI and then when AI was taken away that dropped down to 22%. And the term that sometimes used is intuition rust. Which is the more you use AI, the more your own sense of intuition about a situation can erode a little bit. And I think it's important not to let this happen.

So one of the things that I do is use AI use LLMs and AI very consciously and strategically in my work. And so so I'll tell you how I do that. So whenever I do any writing, I think about it in three phases. The first phase is perhaps the pre-writing phase and this I use LLMs quite a bit. And I use LLMs mainly for two things. So one you might refer to it as landscaping. So let's say one of the things I'm working on today is a pricing strategy document.

So landscaping would be using an LLM to say, well, how are the competitors doing it? Or are there different pricing strategies in Europe versus Asia versus the US? Or that how has pricing evolved over time? Those kinds of things. So really understanding the landscape, doing kind of quick market research for you, I think is a great use of LLMs. And then I would say mind expanding activities. So you might throw out a few ideas and say, what else could I think of?

What would be kind of the universe of ideas? It's sort of like asking a doctor, "What else could this be?" You're sort of forcing an expansion of the possibilities before you start to converge. So, the second step when I write something is just all me, I would say. And in this case, it's just plain writing, which I do once, and then editing, which I do once. So, I never let an LLM write the first draft of something for me.

And I know that's debatable, but that's just my choice. So, when I get to the stage, I write it beginning to end without any help. And I print it out and I edit it once without any help. And then lastly, the last stage is I use an LLM for grammar and spelling only. Okay? So, So basically, I'll take the whole thing that I wrote, throw it in there, and I say, "I want you to give me advice and edits for grammar and spelling only." So, cuz I want my words to be in the ending document and I don't want the ideas changed, but sometimes I need help with grammar and spelling.

So, So that's the final stage. So, this is kind of the way I think about it and the way I use LLMs when doing something like like this. And one way you might think about it is if you're familiar with decision theory, in decision theory, in the beginning, you want to be divergent. You want to get all ideas on the table. This might be a group discussion or debate or what have you. And then, and then you want to be convergent.

So, okay, let's finally decide. Let's narrow this down. Let's decide what we're going to do. So, you kind of want to narrow the playing field and then pick one option in the end. And so, I think LLMs are good for divergence and I think you need to do the convergence of thinking to come up with your own idea in the end. Uh so, so that's basically the way I do it. Now, I realize not every job is like mine. In some cases, you might have a writing job and LLMs help you be more productive.

Um but in my case, I have a senior leadership job. So, it's more about decision accuracy and strategy and those kinds of things versus just churning out a lot of writing. So, I'm careful not to let LLMs help me, but help me in a very strategic way. So, I hope that was interesting and maybe uh useful to you in the way you think about your use of of GenAI and how it helps you think and do your job. Okay, until next time.

Bye.