Diana Thomas: Can AI Help You Lose Weight and Keep It Off?

James Hill:

Welcome to Weight Loss And, where we delve into the world of weight loss. I'm Jim Hill.

Holly Wyatt:

And I'm Holly Wyatt. We're both dedicated to helping you lose weight, keep it off, and live your best life while you're doing it.

James Hill:

Indeed, we now realize successful weight loss combines the science and art of medicine, knowing what to do and why you will do it.

Holly Wyatt:

Yes, the “And” allows us to talk about all the other stuff that makes your journey so much bigger, better, and exciting.

James Hill:

Ready for the “And” factor?

Holly Wyatt:

Let's dive in.

James Hill:

Here we go.

James Hill:

Holly, every day I hear something new about artificial intelligence or AI, how it's going to help us in so many ways, also how we have to be cautious and how we use it. Let's face it, AI, love it or not, is becoming a part of our daily lives.

Holly Wyatt:

That is so true, Jim. And I'll admit that I'm starting to use it. I use AI now when I'm kind of organizing these podcasts even to kind of get the questions in a logical order or, you know, when I'm brainstorming how to make something that's complex more clear. So I'm starting to use it a lot in the background of my daily life. And what really excites me, though, and I think what we can talk about today is the potential for AI to help someone who's trying to lose weight or maintain it. This could be a whole new level of support.

James Hill:

Yeah, Holly, and you and I know enough about AI just to be dangerous. So luckily, luckily, we have the perfect guest to help us dig into this topic. Dr. Diana Thomas really stands at the forefront of applied mathematics, uniquely bridging the gap between theoretical modeling and practical health solutions. Wow, Holly, that is amazing. She's a professor at the United States Military Academy at West Point and PI of the Artificial Intelligence Data Engineering and Machine Learning Center for the Nutrition for Precision Health Consortium. We'll talk about that a little later. She is a true leader in using AI and machine learning to transform nutrition, obesity, and personal health.

Holly Wyatt:

I love this because she really can take the pie in the sky when we talk about the pie in the sky and put the pie in the plate, right, can really give those kind of practical solutions. So, Diana, welcome. You're a mathematician and you're having real impact on public health. Tell us a little bit about your journey and how did you end up working at the intersection of AI and weight science?

Diana Thomas:

Well, it actually is because of you guys. But let me start from the beginning. I didn't start out going to grad school and thinking, I'm going to do this. And in fact, I'd like to point out that no one my age did start out that way because AI machine learning programs weren't there back in the day. So you're going to find people like us coming from computer science, traditional computer science programs. So I went to a traditional math program. I was a pure mathematician until maybe 2008-ish when I had a student approach me about a problem. But even before that, I was thinking about this because in year 2000, I had my son and I had just 1997-ish finished out at West Point where I was really active. We used to play basketball instead of actually have lunch. Then real life comes into play. You get married, you have children, and I started gaining weight. And I walked into that pregnancy at a higher weight than I'd ever been in my life.

And then I gained even more weight. So by the end of the pregnancy, I was 179 pounds. And then I started losing the weight. I went on Weight Watchers and I started losing the weight. And it took me, you know, the big dose response. You lose a lot of weight right at the beginning. And then those pounds start to decrease as you flatten out and plateau and pounds of weight loss. And so then I started thinking, like, why is it so hard? Why is this hard at the end? And I had a friend at Montclair State where I taught who worked with Toyota as a consultant in engines, thermodynamics. And he said, Di, the thing is that we as humans are the best, most efficient engine you can imagine. Let's model that. Let's think about what that looks like and why it's so hard to burn off a pair. And that's where I started thinking about this. In 2007, 8, I had an undergraduate.

Undergraduates in mathematics love real problems. They're learning how to do mathematics, and they care about the world around them, and they may not be as jazzed about doing something in pure mathematics at that age. And Ashtu Chesla came to me and said, you know, I really would like to see, to model the fact that if does our metabolism adapt to weight loss? And can we put that into models and see why that might be? And around the same time, I had also published a paper with another undergraduate on fetal growth and energy balance and fetal growth. So these things were all converging. And that's where I started thinking about weight change, modeling it.

And by 2011, I realized I'm actually working only in this space. And it's time to stop working in number theory and combinatorial game theory. I'm not doing that anymore. I'm working in the field of weight change, nutrition, obesity. There's just too many interesting problems out there and they are interesting and engaging and I love the fact that I was learning new mathematics. And so I think we also as a disciplinary depth, we don't learn new tools that easily. we don't need them but all of a sudden I need needed them. I needed to learn statistics. I never had even taken a statistics class. I'm a mathematician by training. So learning statistics, learning how to work with data, learning about functional forms and curve fitting, all that was new, different. Learning how to write and speak to another audience, all of that was new. So I started enjoying all aspects of what I was doing and dropped my pure math background and did this. And by the way, I did lose the weight.

James Hill:

Oh, great.

Holly Wyatt:

Well, I think this shows so much. You're evolving and changing and we always are, but there's a passion to it. And that's when you can hear it in your story, right? It was personal, there's passion. So going out there and learning new things is exciting. And, you know, moving forward and changing is just right where you want to be. So that's a great story.

James Hill:

Great story. So before we get started, very succinctly, what is AI?

Diana Thomas:

So let me see if I can break it down, because if you go to nuanced definitions, it becomes really hard to understand it. Because if you think about it, if you think about in general definitions, even the simplest mathematics would fall under that definition of AI. But I think when most people think of AI, they think of some few things. One of them is maybe the Terminator robots. They think of that level of AI, that something's taken over, it has knowledge on its own, It can generate knowledge. That brings to the second piece, which we are now familiar with, which is generative AI and the ability to converse with a tool that has algorithms in the background that speaks to us as if it were human. I think that's one of the ways people think of AI, that we think of it as something that's been programmed to behave like a human.

James Hill:

Wow. Okay. So let's talk about AI and weight management. What are the most exciting things in this area that you're working on right now?

Diana Thomas:

Where to start on that? So there's the personal. The personal has never left. I'm going to go to that second. I'm going to start more with maybe what companies are doing and what they found. And I'm going to focus in on one company. I don't think they're the only company doing things like this because it's a natural evolution. There's a company called Simple. It's a platform and I think it's run out of the UK but it's using AI as a weight loss coach.

James Hill:

Oh, okay.

Diana Thomas:

So, in the past we've had commercial platforms that use people as coaches and there are, you know, you can tailor. In fact, I talked to someone at Simple the other day and I was asking them if they were using people as their coaches instead of the AI avatar as their coach. How would they train people to do the same things the AI coach is doing and what happens when you work with people? People are humans, and we have our own human personalities that come, and you'd have to train someone to tailor their approach to that person to map what is being done by generative AI as a coach or an an avatar as a coach. Some of the things that Simple is doing which I think other places are doing as well is trying to tailor that coach to that person that they're helping. They have daily conversations with the person. They are pinging the person if they don't hear from the person. Hey, what's going on? It emulates some of that human behavior but as a personal coach. The user can prompt the avatar if they want to talk to the AI coach but the ai coach can also prompt the user if they have certain things like they don't hear from the person for a few days so there's prompts that reverse that and the AI coach can come out and ask questions. They have definitely seen a jump in folks losing 5% of their body weight. But the person, Roe Huntress, that I talked to, she said that they can't attribute that to just the AI coach because they're also changing other aspects of their platform at the same time. So it's not like they can isolate and say it's the AI coach. But one of the big things I've heard, not just from Simple, but on the research end, is that if you have something like an AI coach, that coach is available 24/7.

James Hill:

Yeah. And people aren't. You can't always make people available all the time.

Diana Thomas:

No. And the AI coach also is tailored to you because generative AI starts learning your habits and the way you like to hear things. So that is a big, big difference. Let's change this instead of weight loss, your math teacher. Your math teacher used to teach you math and that math teacher is the same for everybody. I'm a math teacher also so I tailor my approach to my students maybe by what I feel they need or their personality but there's only so much because the baseline of my personality is not going to change. So if someone's overwhelmed by my extroverted personality, there's nothing I can do about it. I'm an extrovert.

Diana Thomas:

But the AI coaches, they can tailor that to your personality. And in fact, I've done that on my own. I use ChatGPT for helping me code. I like my code and help a certain way. And over time, my ChatGPT has evolved to know what I like to the point that even when ChatGPT opened a new beta version of something that will help me with my programming, I actually preferred my tailored one over the new one that was made by ChatGPT because that one's tailored to me. And so those two things, being a 24/7 available and tailored to you, I think are huge pluses.

James Hill:

So we're going to see more of that. But the science, we don't have the research yet to show if they're going to be better or how much better. But I do think it makes sense that we're going to see more of these.

Diana Thomas:

You're already seeing it on the research end, Jim, as you and I know, when we go to conferences, people are talking about their research. And we do see it in presentations. These things are evolving in the research realm. And then we will see the research to go with that. I just spoke to someone who has a paper under review, but she'd already told me, we have an AI coach that prompts users and does some of these things for her research. And that paper's under review right now. So we are going to see this and we're going to see the data to see if there is a jump in improvement.

James Hill:

Okay.

Diana Thomas:

So it's coming.

Holly Wyatt:

I have a question. I've noticed, too, that the iterations, it's now knows what I like and now knows how to present things to me, especially in some of my ChatGPT type applications. Is that always good? That's the kind of question, meaning sometimes I'm thinking I push people sometimes in a way that I'm not sure ChatGPT would push me. So is there a downside, I guess, to that aspect and getting really into this is the way I like it and this is the way I'm always going to get the information?

Diana Thomas:

That's a good question. And that's something that I think could be a good research question. What are we losing when it's so tailored to us? Maybe there's certain aspects that will be great. For example, I literally told ChatGPT please don't congratulate me on every thing that I'm telling you, I'm when "Oh, this is a great idea, Diana." I actually find that annoying please don't do that. What am I losing when it does that? I don't know. And they did find that at Simple the same thing that that congratulatory piece. They asked it to temper some of those things. What do we do when we tailor it and I explicitly tell it not to do these things? I'm not sure what we're losing. Is it giving me code in the best way possible because it's so tailored to me? I don't know.

Holly Wyatt:

Yeah.

James Hill:

But it's moving very quickly, and we may be able to incorporate those sorts of things. So, Holly, you talk a lot about getting out of your comfort zone, and it may be possible to include in your coach the fact that they push you out of your comfort zone.

Holly Wyatt:

That's true. Maybe that's what I need to say. Maybe I need to tell it that and see what it does. Yeah.

Diana Thomas:

In fact, Simple has done that. I'm trying to remember what they did. Oh, yeah. Okay. So if you look at ChatGPT, one of the things it'll do. So ChatGPT is a great example of this. And it's hardwired in here because it just happens. So, I did ask it. And this is not weight loss, but it's a weight management thing. I have a thyroid problem. I'm hypothyroidic right now. And with a low thyroid production, weight gain is a consequence. And being in the obesity research field and having had high weight gain in the past and lost it, that's a fearful thing for me. So I asked ChatGPT to make me meal plans. And this is something that others have seen as well, that the one thing they love about gender of AI can make them meal plans. I asked it to present for me a seven-day meal plan. I like my comfort foods. I like pizza. I like Mexican. I'm a foodie. I love all those things. And so I said, give me those things that help my thyroid. And also at the same time, give me like low calories so I don't gain weight, all that stuff. But it always asks you next. Would you like, it gave me what it asked for, but it asked me for improvements.

That I find really annoying. And Simple has found the same thing. They found that actually to be not good. They actually do ask and query their customers and the customers don't like when they're on a weight loss plan to always be told that they need an improvement and just to end that conversation with what they have achieved instead of just keep, and now you can try this. And so they have taken that out. They've asked it to be taken out. And as Jim pointed out, we're learning this. We're not there yet. One of my colleagues, Rodolfo Barango, who's in genetics, said, act one, scene one, AI has just stepped in the room. That's where we're at.

James Hill:

We have a long way to go.

Diana Thomas:

We have a long way to go. Yes.

James Hill:

So if one of our users out there is beginning a weight loss journey, how can AI help maybe in anticipating their rate of weight loss, etc.? And then Holly and I are very big on distinguishing weight loss from weight loss maintenance. How can AI help with both phases there?

Diana Thomas:

So for number one, this is my son, actually, while he was deployed and came back, had gained a lot of weight during his deployment. And he was actually looking at, you know, using ChatGPT. And one of the things, and this is one of the failings of ChatGPT, when he asked for a rate of weight loss, what do you think it said? I'm just curious what you guys think it said when he asked about his rate of weight loss.

James Hill:

It probably gave him a linear weight loss.

Diana Thomas:

It did. It gave him the 3,500 calorie rule.

James Hill:

And we know that doesn't work.

Diana Thomas:

And so he said something along the lines, hey, it looks like you're using the 3,500 calorie rule. Hasn't that been out? Isn't that outdated? And it said, oh, sorry, you're right. And it started using thermodynamic models.

James Hill:

Oh, wow.

Diana Thomas:

But he had to point it out. And he was he happens to be my son. So he's heard this incessantly for the past decade. There's still bumps in it.

Holly Wyatt:

Well, I think there's lots of bumps in it. And that, to me, is the problem because when I'm in a subject I know, I see it. I'm like, wait a minute, that's not right. And it'll say, yes, you're right. Sorry, da-da-da. And I'm like, but wait a minute. If it's a subject I don't know, I would have potentially taken that. So that, to me, is a big problem. And what do we do with that?

Diana Thomas:

I think at the education level, people are extremely fearful of AI. It's a new tool. In my field, we had graphing calculators. We had computers that can solve algebra problems. This is a tornado in the education field in comparison. It is wiping through. And so what should we teach our next generation of students in all fields? It's the sniff check. It's being very careful and being not taking, you know, we had this with Wikipedia. Wikipedia is, yeah, I don't know if you had it in your classes, but should Wikipedia be ground truth? Well, no, it's an initial place for getting something. It should not be ground truth. Teaching our students in K-12, in undergraduate and graduate programs that ChatGPT or generative AI still needs that sniff check. You should look at that as the first place, but not the end-all be-all. And the idea that this all goes away one day, like it's going to get so good that it's going to know all those things. I also need to know what question to ask.

You need some domain knowledge to ask the right questions. So I've heard, for example, that we won't need programmers. It's going to replace us. Well, no, that's not quite true because I need to know what Python can do, what it can do, like what kind of questions to ask. If I don't even know what question to ask, well, as a user, the most dangerous student is the one that thinks they know. And so I think reformulating how we teach students, teaching them with deliberate exercises of sniff checks, and maybe even our end users that are using generative AI to help with weight loss. Hey, caution. This is not the end-all be-all. You have to actually sniff check some of these things. Come to us with questions.

James Hill:

So, Di, that's a good point. And it may be, I mean, people are out there saying, well, these coaches are going to replace dietitians and behaviors and so forth. Maybe a better way to think about it is this is a tool that can actually help them do their job better. So it's not replacing them. It's giving them a more effective tool to help their clients.

Diana Thomas:

Yes. I think that in general, when we look at AI, we think of it as replacements. And in fact, some of the things that I've heard about AI, it can do not just generative AI. Think about one thing that we do on a daily basis in our research fields. We look up information. And Holly, you had mentioned already that you look up information, right? Andit's at your fingertips. And maybe one day ChatGPT won't hallucinate. So there's bumps that may be fixed. But, you know, we can write what we call a review paper pretty quickly now. And it's not generative AI, it's actually just AI. I can write code to scrape PubMed and pull every paper on a topic, put it into nice files. I can have those abstracts summarized, organized. And then you can ask yourself, does this replace how someone does a review?

Well, no. First of all, that's a very unimaginative use of AI that is replacing what we've done before. That's one part of it. You still have to do the sniff checks. The second part of it is, you wouldn't ask the all-important question, what can we do that's beyond what we ever started doing? Which is, is there a way to use that information now to ask what reviews haven't been done? That's a question I couldn't ask before. And so I think the excitement behind AI can only be there if you're willing to be a little bit imaginative of what it could do.

James Hill:

I want to get back to weight loss maintenance, Di. And in weight loss, you can go and temporarily do a lot of things and lose weight. And there are many different diets that allow you to lose weight. But weight loss maintenance is figuring out how to live your life long term. And things like physical activity become much more important there. How can AI help? And how can it understand the fact that everybody's a little different in how they live their lives?

Diana Thomas:

So I think, first of all, like, I'll say this second, but I'll say this first. So having your cell phone with you and having AI at your fingertips with you, you have to use this tool over and over to see what it can do. And then you can use it in all kinds of ways. So here's one of the questions I asked it that had to do with physical activity. I was in this room and I wanted to know if the room had, it had a sled it had the weights, I wanted to see if it has the same dimensions as what was required for the army combat fitness test. And so i took a picture of the room and I asked is this the right distance, it's the same distance if i was going to do the sled drag? I think the sled drag is a phenomenal exercise, by the way. What ChatGPT told me is it looks like it's this many yards across and blah blah blah. It gave me some dimensions and it said I think you're good. And then I asked it, sniff check, how did you come up with that? And it said, "Oh, in there there was a box that people used to jump. The jump box you jump up and jump down, and these boxes are usually this height so I used that box as my my tool to figure out what the dimensions of the rest of the room were."

And so this is just one use of physical activity for this. But you could also put in some of the other things like I did with diet. Like, I like my pizza. I like my things. Give me some recipes that mimic that feeling of comfort food that I like, but keep the calories low and are thyroid enhancing. You could do the same thing. I don't like to run really fast. I don't really like doing those things but I want to be physically active. How can I do? Or I'm in a confined room or I'm traveling. What can I do to get max heart rate some other way? Or I like my endorphins when I run. How can I do it another way? And it will give you ideas that you never thought of before.

Holly Wyatt:

I think the key is the question.

Diana Thomas:

The question, yeah.

Holly Wyatt:

To have the question. And so it's almost learning how to ask the question, I think, is the first step really in getting good at this. And that's what I've noticed. I've gotten better at asking questions and I did the same thing. I'm like, why do you say this? How did you come up with this?

Diana Thomas:

Sniff check.

Holly Wyatt:

And I do it all the time and it'll give you a long list of how it came up with it. So one of the things that my state-of-slimmers are using this a lot for is they'll go into their kitchen, have certain number of ingredients and say, what can I have for dinner? I want it to be this many calories or set some kind of macros and it's amazing what it can come up with. What would you say if you're just getting started? How could you use it to help manage your weight? What would be an easy way to dip your toe in the water?

Diana Thomas:

The easy way to dip your toe in the water is just to do what I did with the thyroid problem, which is tell it what foods you have, ingredients you have, just like you did. Tell it what you like. I think telling it what you like is a huge part of things. And my son gave me the idea. I was like, oh, he said he was tired of eating the things that it told him to eat. And so he asked it, you know, to get me some recipes that make me feel like just had a ramen bowl. Get me a recipe of something that is the things that I like but tell me what the calories are in the end and how you estimated it and then...

Holly Wyatt:

When we say it, I know some of the listeners, they're starting out. What is it? Are we saying ChatGPT? Do you need to buy a subscription? Is it the very basic? How do they get there?

Diana Thomas:

I believe the subscription of ChatGPT is like the best one. It's bang for the buck because it's not limited to just asking questions about weight management. It's for anything. My first part of this would be asking just questions about developing that meal for you if you're really sick of whatever meal that keeps you in calories. I'm supposed to have 300 calories a meal.

Holly Wyatt:

Yeah.

Diana Thomas:

Give me that meal. That's where I'd start. And I use ChatGPT.

Holly Wyatt:

ChatGPT where you pay for this. There's a free ChatGPT. And then there's a subscription. I think it's like, is it, I cannot now remember if it's $40 a month or what it is per month.

Diana Thomas:

I would start out with the free one.

Holly Wyatt:

Okay.

Diana Thomas:

And see if you can use it for other things. And why I say that is definitely my family is, my siblings are all big users now. But the one person I showed this to that I thought could really use this is my mother. And she was watching me use it for something. And I had shown her, she wanted something filled by the side of her house with concrete. And I asked ChatGPT, how much would this, what are the dimensions. So I had learned that from my physical fitness question like oh tell me the dimensions of the place she needed filled with concrete and give me an estimate. I live in this town in Pennsylvania, give me an estimate of how much it should cost to do this. Give me some rough ideas so that when she gets quotes she has some idea. She watched me do that. And I said, "Mom, this is actually a great tool for you. Let me show you how it can help you." So I showed her a few examples, like, oh, she wanted to go tour around Philadelphia with her sister. They're both elderly. They needed access to bathrooms. So I asked her to do that, and it gave me some options for things she could do that were low cost or no cost. She was sold. I bought her the prescription. She came up just recently, and she was watching us all use it. And she goes, I need to use that. And so for her, the paid subscription's not doing it. I would start with the unpaid, the free version, and then get used to using it first. Even if you're using it like Google, for every Google search you do, go into ChatGPT, the free version, and ask ChatGPT to do it for you.

James Hill:

Interesting.

Diana Thomas:

And get used to getting started using it. Because AI is like a big landmass. You're stepping out into this landmass and you have no idea what it can do for you yet, and each of us are forging our narrow path on this and there's hedges up. We don't see what the other person's doing unless we're talking about it like my son and I talk about it. But, you know, until you start navigating that path and using it for your own things, you're not going to see what power it has. And they do have classes in prompt engineering now. Worthless. Because you have to ask those questions yourself. Until I know what I'm going to ask, what it can do for me, I'm not going to really know what to ask it.

Holly Wyatt:

But I did, I actually, I think I bought one of those books because when I first started out, it was like how to prompt. And what it did show me is examples of types of questions and it kind of got me thinking, oh, I didn't realize, you know, because I had no idea the different types of questions to even start to use. So I don't know that I'd do a class, but I think, you know, kind of getting out there and saying, okay, maybe use ChatGPT. What types of questions, you know, would be useful, et cetera, you know, on a certain topic would be something to start with.

Diana Thomas:

I would say you're better off on your own. Use it as Google at first because the best AI machine still is you.

Holly Wyatt:

Yeah.

Diana Thomas:

Right? So you are also learning.

James Hill:

So play around with it. Try it. Try asking questions. Di, I have a question for you. We talked about the ways it can help with coaching and so forth. One of the biggest barriers sometimes is providing the right motivation for making behavior changes. Can AI help with that?

Diana Thomas:

So I'm not a behavioral psychologist but I think it can from the work I've done with Corby Martin that having frequent check-ins. Well, ChatGPT and things like that can do that for you. Having immediate measures. These are things that Corby told me are really important for behavior change. So I can do my calorie. What's my total calorie count right now for today? I ate your meal plan thing for breakfast. I ate it for lunch. I've not eaten anything in between. Tell me where I'm at?

James Hill:

Okay.

Diana Thomas:

I think that one more thing it brings to you is the same things that Google does, which is not completely private, but usually that makes it so that Google searches tell you who you are. Because we are very open with Google of what we search.

James Hill:

Yeah.

Diana Thomas:

And because we feel much more at ease asking Google, you know, you might search for what the ex-classmate of yours is doing today, which you never would tell, hey, I'm really curious. I was really jealous of so-and-so during high school, and I want to know what they're up to today. You would be much more open with Google than you would chit-chatting about that with your friends.

Holly Wyatt:

That brings a good question, because I think about this, and I think some of our listeners would too, kind of what are the ethical concerns about putting sensitive data, maybe sensitive health data, since we're talking about weight loss, you know, in the AI applications? That data, where's it going? When I give it something about myself, can other people have access to it? Where's it going? What's being done with that? So ethically, how do you think about that?

Diana Thomas:

That's an important thing for me to ask also, because sometimes I put my ideas. The thing about generative AI is that it's interacting with you. So not only are you putting information in, it's giving you information out and you're interacting with this. And after some time, it's hard to dissect apart who is DI and who is generative AI, because we've kind of melded together like the bionic woman back in the day, like interacting like this. So I did some experiments. One of the best things about ChatGPT or something like that is you can ask ChatGPT, what are you doing with the questions I'm asking you? And it will tell you. And we did a little experiment.

I had given ChatGPT an idea because it was helping me code something and it was complex. So I gave it an idea. I said, can you, instead of doing what you're doing, can you try this? And it's a great idea. And it gave me the revamped code. So I ran an experiment with my son who's in Colorado and I'm in Jersey. And I said, why don't you ask it the same question and see if it will give you my idea? And it didn't. Not right away. He had to narrow it down and slide to that. And then he asked it, well, you know, my mom asked you this question in Jersey. And it said, oh, I'm not able to take her idea and give it to you. That comes after the next version of ChatGPT comes. And even then, I will not give you her idea. I will blend that idea with other ideas of other users so it doesn't come out as one idea.

James Hill:

Interesting.

Diana Thomas:

And so what we learned from that is I think most people are afraid that, hey, you shouldn't put your social security number into ChatGPT because there's some guy with spectacles on inside ChatGPT that's going to use it now to open credit cards. That's not really what it's doing. It's doing something quite different. That would be the reason you don't put your social security number in Google. Actually, the privacy things that they have in ChatGPT, that's not what it's doing. It's doing something even more complicated and more scary. Right? Because as someone who generates ideas, the thought of having my ideas merged with somebody I don't even know that might be asking the same questions somewhere else, and blending them and then spitting them out in ChatGPT5 somewhere, we don't have any love for that.

James Hill:

That's interesting.

Diana Thomas:

That's uncharted territory.

James Hill:

Di, I want to ask you one more question here. In the weight loss area, the big deal now are the new versions of weight loss medications that are based on GLP-1 receptor agonists. How can AI play a role in people that may be taking or considering to take the meds?

Diana Thomas:

Ooh I think the first place someone might go is if they're using generative AI to ask questions about GLP-1s, I did see people using trying to identify like how people are using questions about GLP-1s inside generative AI, but you run the risk here that Holly already brought up that it is not a scientist on the other end. Like 3500, which has been outdated for decades now. If you're not the person that's going to be able to do the sniff check, I would be really hesitant to take that as the final say. No one should stop you from putting something in and asking questions and having a dialogue with generative AI on this and we're talking about generative AI. But you should always go back to your card-carrying healthcare provider for this.

James Hill:

Okay.

Holly Wyatt:

Yeah, I agree. I find mistakes all the time when I know the subject really, really well. And so that makes me then cautious. That doesn't mean I don't use it, but I definitely, I like your sniff test and I like being smart about this. And I think when it comes to something like that, your health care provider is going to know your whole situation a lot better.

Diana Thomas:

Yes.

Holly Wyatt:

Than AI.

Diana Thomas:

And the only way to get really good at these sniff checks is to just use the tool and know what the limitations of the tool are and keep playing with the tool on all kinds of questions, not just here.

Holly Wyatt:

Yeah.

Diana Thomas:

Like, how did you know the dimensions of this room were this? Tell me. And then you'll get the sniff checks. But we can build those into classrooms too.

James Hill:

Cool. Holly, I think it's time for some listener questions.

Holly Wyatt:

Yeah, we have some. We can try to do these rapid fire because I know that this is a popular topic. People have been asking for this topic. So let's see if we can get through a few of these. First one, there are so many weight loss apps claiming to use AI or using that buzzword. How can I actually tell if the one that I'm looking at is helpful. Is there any way to kind of say, is this a good one? We've been talking about ChatGPT, but there's lots of different ways to use AI and different programs using AI.

Diana Thomas:

Right. I would look at customer reviews. I looked up Simple's customer reviews today and look for the objective customer reviews. How do you know which ones are objective? Well, Amazon actually has. They show you how they use their customer reviews to weed out things that are not real and bots and all that stuff. So I would find a place that's using it that has some kind of vetting of their customer reviews and maybe a third party customer review and find out how it's using it just because someone's using it. So, for example, my brother and I are coding something up where he is, we are using AI, but the user never sees it. So can I say it's an AI driven tool? Sure, you'll never see it because it's part of the way we combine data inside there. So it's not like really something that the user will see. Here's another example. There was a company I spoke to, and they market themselves as AI, AI models of meal plans. When I really dug into it, like, tell me about the math. Well, it turned out it wasn't what we think of as an AI model. It was actually an operations research model. But yeah, under the strict definition of AI, regression falls under AI. So it wasn't really what people think of AI. That's why I didn't give a formal definition of AI, because then you start getting into the weeds of this. But under the formal definition of AI, yeah, that was AI.

Holly Wyatt:

All right, Jim, do another one.

James Hill:

Yeah, I've got another one here. I get overwhelmed by all the nutrition advice out there. Can AI actually simplify things instead of making them more complicated?

Diana Thomas:

That I think it does. With sniff checks, you know, never take it as ground truth. But that it does. So it can run those OR models, the operations research models, with a blink of an eye. And for example, the meal plan becomes really easy. the calculation, the estimation of calories becomes really easy. Keeping a running total becomes really easy. That nutritional advice can be condensed and tailored to you. That's where its strength is. So yes, it can cut to the chase. And also it's the questions you ask. You know, are you asking those questions that will help? And that just comes from being familiar, more and more familiar with the tool.

Holly Wyatt:

So I've got one more, and we kind of already talked on this, but since we had several people who talked about this, I just want to make sure we cover this. They say, I want something personalized, but I'm worried about my privacy. How do I know if AI tools are safe to use?

Diana Thomas:

So I really recommend this book. It's called Mindmasters by Sandra Matz, I think her name is. And they have a whole chapter on privacy and how things are used in social media platforms all over the place. For the most part, we have opt-in. We have to actually verbally say opt-out for a lot of things. That's hefty. There's a lot of fine print to it. And you might want to get to the platform quickly. And so you might want to look at it. I would say if you're really concerned about the issue, and you should be, read that fine print. In fact, you can take that fine print, throw it into ChatGPT, and tell it to summarize for you what that fine print is. That's a great use of AI against itself.

James Hill:

Cool.

Holly Wyatt:

I love it.

James Hill:

So one of the things we like to ask our guests, Diana, is what's one thing you got wrong in your career as you went from mathematics to public health? What's one thing that you might have mispredicted the future or got wrong?

Diana Thomas:

I don't know if this has got wrong per se, but weight loss is much more complicated than I ever, ever thought.

Holly Wyatt:

Yes.

Diana Thomas:

Weight loss and weight management. I am continually reminded of that. So I mentioned the thyroid problem. Yes, it's calories in minus calories out. But how I process those calories now with a thyroid problem, I think I made a joke on LinkedIn recently. Where someone said, it's just calories in minus calories out. And I made a joke and said, then you've never had a thyroid problem. It is so complicated. And we're all looking for it to be easy. And it's just as complicated as models for space travel. It's just as complicated if not more so than models of the weather. And we've tackled those things, but we haven't mastered them. Otherwise, the weather people could just tell us what the weather is going to be like a month from now, and then they can't. So I would say all the things we have, they give us some insight and approximation of our world in the world of weight management, but there's so many dimensions to it.

Holly Wyatt:

We agree. I think this idea that it's simple in one solution, in one diet, in one way has really, we talk about this all the time, is a disservice to the field. We now need to understand the complexity and really work together. And that's a problem I think science has, you know. It's about combining expertise to really get to move the needle, in my opinion.

Diana Thomas:

Even if we get there, we haven't got there because you're always going to learn one more layer that's more complicated. So I would say the solution is an asymptote. We're approaching maybe be that asymptote, but we're never going to get there because it's just so complicated.

Holly Wyatt:

What do you think is your biggest challenge or maybe just the biggest challenge in making AI useful for real people with real lives?

Diana Thomas:

I think it's the data access and the tools to make that access usable for an average user. We've had a lot of time to think about something like predicting the weather. So now I have an app where it shows me a happy, sunny face. It took those complicated models, and then it gives me that way of making that complicated model useful for me. We've had a lot of time to do that since the beginning of computers, right? But we haven't. If that's a tornado, this is a class five hurricane of data that we have right now. And I would say it's a challenge combined with what Rodolfo Barango said is act one, scene one, AI has just stepped in the room. All of these things are just stepping in the room, and they're stepping in the room together, and we're writing that play, and it's so complicated and hard. That's the single challenge. Like, how do you make that easy, not just for the end user, but the researcher? It's complicated math, and it's complicated and hard for me.

James Hill:

All right, Holly, let's see if we can sort of do a takeaway here. The one thing I think that our listeners should understand is AI, while it's just getting started, is something that can be useful now. And it's something that our listeners should start playing around with. And I think Diana's advice is use one of the free versions of ChatGPT and start asking questions. Look at the answers with a healthy dose of skepticism. But as you learn to use it, and again, this tool is going to get better and better and better, but it's a tool that's here to stay. So start using it. Start getting familiar with it. Again, be a little skeptical, but it is here. It is a great tool. It's going to get better. And it's something that I think can be very, very helpful to people who are working to manage their weight.

Holly Wyatt:

Yeah, I would agree. Give it a try. It's not going away. I don't think ignoring it is the solution. Get in and just play with it. Have some fun with it. Ask some questions that don't matter, you know, that aren't private or that don't matter and just see what it can do.

James Hill:

It's fun. It's fun to play around with it. You get started, man. You get all these questions and you ask and anyway, give it a try. Di, thank you so much for joining us today. We really appreciate your time. And I think your advice is going to be very helpful to our listeners.

Diana Thomas:

Thank you for having me.

Holly Wyatt:

Yes, thank you. Bye, everybody. We'll see you on the next episode.

James Hill:

See you next time. Bye. And that's a wrap for today's episode of Weight Loss And. We hope you enjoy diving into the world of weight loss with us.

Holly Wyatt:

If you want to stay connected and continue exploring the “Ands” of weight loss, be sure to follow our podcast on your favorite platform.

James Hill:

We'd also love to hear from you. Share your thoughts, questions, or topic suggestions by reaching out at [weightlossand.com](http://weightlossand.com/). Your feedback helps us tailor future episodes to your needs.

Holly Wyatt:

And remember, the journey doesn't end here. Keep applying the knowledge and strategies you've learned and embrace the power of the “And” in your own weight loss journey.