Welcome to Real Talk on AI and I care with your host, Dr. Scott Morris. And I'm co host, Dr. Rayhan Ahmed.
In each episode, we discuss current news, adding a little AI education, and debate innovative AI tools and topics that will change the industry. Hey, everybody. It's Scott Rayhan. Welcome to this week's Real Talk.
I can't believe it. Rayhan, we're on Real Talk number 46. I never thought when we started, we'd get this far, but here we are. So what do you got for knowledge bite for the day, my friend?
Well, what I wanted to talk about is basically part two of prompting. We talked about how to craft a good prompt. And in November, it seems like a long, long time ago, but we talked about how you can craft a prompt and how you prompt engineer, which is really a whole topic in itself. And I want to extend that to a phrase that I'm actually calling prompt pushback or kind of like the socratic method of prompting.
And I'll tell you why it's because there's a lot of AI slopp that's going around. And I think we all sort of recognize it when we get an email or see a post on LinkedIn or just an article that we're like, hey, did AI write that? And the reason we're seeing that is because we're not interrogating the prompts enough. I was listening to the radio today and and I it was a guy called Cal Newport who recently wrote an op-ed in New York Times is called there's a good reason you can't concentrate.
And basically his argument is that we're not distracted because of lack of willpower. It's that the digital environment, you know, all the Instagram feeds has have really cut down on our cognitive fitness. We are not we are sort of letting things because we're just we're humans. We just do what comes easy and AI is especially good at reducing friction, but friction is exactly where thinking happens.
And so what happens is you know, you have something to write or something to talk about and you just ask the AI, hey, write this for me. Right. And it comes back with a beautifully very confident couple paragraphs and you could just copy paste and that's where AI saw comes from. So in this idea called prompt pushback or the socratic method of prompting.
It's it's asking a ton of questions after you get that do not take what is given to you at face value. There is an interrogation. There's a back and forth. So you could ask questions like what assumptions am I making?
What could be wrong with this? Think about being a skeptic. What you could literally say, you know, what would a skeptical expert say to what you just wrote or what evidence would change your conclusion? Where are you being too confident or where are you being too confident and what's missing?
You know, Scott, you know, I don't know if you know, I was a philosophy major and so I didn't know that. Yeah. Before I decided to go to med school and this is this is what in philosophy call it's after the philosopher's opportunities and play the call the socratic method where you just ask a whole bunch of questions. And people plush answers you ask questions you expose week assumptions you make people examine what they actually understood and what they really think this is what how we should actually approach AI.
So this is really not about prompt engineering and this sort of knowledge about this is stuff we should be doing with everything and you should treat your AI colleague the same way. They're different than probably what we do when somebody gets up and says, I have this great idea about this procedure or drug. We're like, well, why? Yeah, as I care providers, we're already really good at this stuff, but sometimes when a computer saying it and it says it's so beautifully and confidently, we sort of like take it at face value.
But prompt pushback is really about asking second order questions and forcing yourself to better understand the thinking and the assumptions under underneath that beautifully written paragraph. And so that's that's the knowledge by today I encourage people to really interact and interrogate and try different elements to get the, you know, the best possible use case as we're all learning and exploring these fascinating new technologies. Scott, well, what do you think? I mean, is that something you're you've been working on yourself as well?
Yeah, it is, you know, I mean, I've really been focusing on prompting and doing some of things and ask acting treating it more like a team member of, well, why are we doing this? What are we doing it for? It goes all back to the same questions, right? The what, where, why, who, when, and how?
And I think if you get an S you, you say I'm going to prompt somebody, you get an answer. I think if we have an answer, those five or six questions. And sometimes what you find out what your AI comes out with isn't really all that strong, you know, so who's your audience? Why are you doing this?
What's the reason behind it? I agree with I never really thought about calling it the surprising method, but that really is that's a great idea for what it is. I think I think that brings us into our fireside today, you know, because for the audience, we're talking a little bit about it before this, you know, I feel like emotionally, there's some changes happening. And last week we talked about, well, what is what are the reasons AI is going to fail?
And it was first time I really kind of started to think, maybe I have a little disillusionment. And it brought us back to one of our first podcasts, we talked about the hype curve. And we said, hey, you know, we're in the rapid rise of the hype curve back then, you know, oh my gosh, that was what almost 14 months ago, that was in the dark ages of of AI. Now we're looking at going, I think as I read LinkedIn, I read the posts.
And of course, now there's a thousand AI experts that are out there that say their I experts. And I started wonder, I like, is hasn't delivered on the promise or hasn't delivered on a promise we thought was being offered. And I kind of feel like, kind of feel like there's four big shifts going on, Rayhan, in terms of where are we in the trough. I think we're on the downhill slide.
I think that the trough of disillusionment is we're entering that quagmire. And we could be there for six months, 12 months, 18 months. But this is a hype curve, right? And the hype curve has a rapid rapid increase, a rapid decrease, a stabilization, and then we have growth.
And I think my personal opinion, I'll be interested to see what you think. I think we're entering the trough of disillusionment. I think that's because generative AI people are going, yeah, I've done that. And it really hasn't changed medicine.
Like, it's changed the way we do some things. But I don't think it's changed the systems that medicine operates on. So I think that there was a little, it's going to do this and it never really did it, or at least it hasn't done it yet. I don't know.
I think that's number one. Number two, I think that, you know, I think we had high hopes that ambient intelligence, ambient scribes, documentation tools, we're going to change our life. But kind of like we talked about when we did the scribe, real talk a while ago, was that the problem with that is that the scribes are really good. But what they're feeding their data into isn't.
And so we still have a data problem. Yeah, and I think part of the entering of the trough of disillusionment is the economic uncertainty that's occurring. Maybe, and it not totally clear on whether the job displacements, job losses are being, you know, people are blaming AI, though it's not totally clear that AI is actually causing that. Or is it just sort of a negative economic cycles, particularly in the in the software industry.
I know friends or kids of friends are really having a hard time. Recent graduates are having a really hard time finding jobs, really science and majors in particular. And you know, is it due to AI? Is it due to just a tech, you know, a boom and bust type of cycle?
Not, I mean, I don't think anyone is totally sure, but certainly, you know, AI in software engineering is probably the canary in a cool mind and a lot of experts are out there saying that what's happening to software is going to be happening to all our industries as AI sort of enters there. Totally agree. So I think there there may be something about that in terms of the hot, you know, us getting disillusioned. Yeah, I think I said this before.
We oftentimes overestimate what's going to happen in one to two years, but underestimate the ten years and I think that's where we're at. The uptake in who's using AI is very uneven. I talked to a bunch of friends and of course more and more people are using AI to this survey every year and I would say most people now have interacted with chat to BT or Gemini or whatever, but using it. Like a genetic AI or any of the stuff that's just yeah, scribe being preauthorizations probably because technology solutions are sort of.
They're not fully baked into workflow yet. And that's exactly what's going to say right now and I think, you know, when my business classes was a long time ago, you know, they called it last mile problems. And I think we kind of talked a little bit last week, right? I think that our systems.
And which if there's a lot, but our systems are hitting this last mile problem in that. We don't have solid systems that the AI can learn from. And thus there's a technical issue, there's a data issue, there's a cultural issue. There's an implementation issue.
I think that the tech not a lot of the tech is there, but the real challenges, how do we actually use it successfully? I don't think we've answered those questions yet. Yeah, when you have a new technology that sort of a foundational technology like electricity. Historically, the gains you see sort of happen very, very slow at the beginning and then amazingly fast.
And I think electricity is a good example when it sort of when it first sort of invented and introduced factories are trying to retrofit electricity into their workflow, which to now in like modern ears sounds crazy like how do you retrofit some things of foundational. And there's only when new factories were built that assumed electricity would just be available. Which is how we view electricity. So if you were to build a new planet, a new workflow, assuming AI is going to be touching every single piece, that's when I think the big gains are going to come.
We are very much in this transitional period of retrofitting AI into workflow and it won't be until this is my theory, but it won't be until. And AI is embedded in every single workflow that when we come up with entirely new designs, we assume AI is going to be part of it just as a matter of course. And that's when those workflows are really going to. Yeah, the foundation of the workflows instead of the tool trying to make it better.
Yeah, it's going to be in the same way as soon electricity is going to be part of the workflow that there's going to be an AI element to the workflow. Yeah, we are very much. So I think we are in the, I think we are entering a trough for a variety of reasons, the economic reasons and for the lack of really big wow technologies. But we're going to come out when we have much more foundational use cases and sort of what I described and then.
That five to 10 year mark is just going to be really. And it's going to be amazing in ways that I don't think anyone can really predict. I love what you said, though, Rayhan, about the fact that we it's going to require a change in how our systems work. And our industry, the I care industry is not exactly.
Super progressive and changing workflow. So there's going to be some challenges, but I totally agree with you. Something I hope maybe this one got you a little thing a little bit thinking about where you each individually sit in your journey to change your systems and change your workflow and where AI is going to fit into that. We want to be the voice of reason and not always, you know, we're both I think AI's cheerleaders, but we're also realists.
I think I think my son calls me a realist, a realistic optimist. So I'm not sure exactly what that is, but I think that's what I am. But you know, so we try to be realistic and be optimistic at the same time. So if I ever learn something, we got you to think a little bit today.
Yeah, appreciate everyone's listening in comments and in the same way we want you to push back on your prompts, push back on us too. You can contact us anytime on LinkedIn or at rayhan our ehan at a i and I care dot AI or Scott that's one T as COT at AI and I care dot AI. And find us at AI and I care dot com where you listen to all our podcasts and read all our material. Thank you so much for listening and we will talk to you next time.