New: The State of Fitness, Q2 2026. Get the free report
← All conversations

Ken Ford: 40 Years in AI and Why Chatbots Aren’t Thinking

Tech and AI59 minHosted by Eric Malzone

With thanks to our partners EGYM and Perfect Gym

Open the Buzzsprout player

"The human brain is what, 20 watts?"

In a scale of one to ten, one being Skynet, right? And ten being uh this abundant utopia where we all get to do exactly what we want and life is good, right? Where where are you on that sliding scale of where do you think AI is bringing us?
Somewhere in between. But don't think AI is bringing us there. I think it is the quality of our leadership and the thoughtfulness of which we as humans employ the tools. A lot of the policymakers and politicians know literally zero about how it works. And so they they play with it and they go, oh wow, it's a solved problem. All we need is a bigger data center. We have data centers now that are under construction that will require more power than some states, entire states. Look, the human brain is what, 20 watts? These things are talking about large nuclear reactors to run one data center.
Hey friends, today's episode is proudly brought to you by EGIM. The fitness industry is entering into an exciting new era. One filled with smarter technology and clearer paths to wellness. We are finally seeing the potential for deeper connections and more effective, data-driven wellness results for everyone involved. That's operators, trainers, and members alike. EGIM is at the heart of this shift, building the ecosystem for health. By connecting their smart strength equipment to the entire fitness floor, members stay motivated and engaged. And with AI-generated training plans, trainers are free to focus on what matters most: building relationships with their clients. EGM is turning fitness into a measurable, repeatable, and actionable science. Now, what really excites me about eGym is that they are democratizing fitness. They are making gyms accessible to everyone. And eGym is listening to what the market wants. With their newest longevity training program, operators are primed to meet the evolving needs

Sponsor: EGYM

of the modern gym goers. Their mission is simple shift the healthcare landscape from repair to proactive prevention. It's time for progress that actually counts. See how they do it at eGym.com. That is egym.com. All right, here we go. Ken Ford, welcome to the future of fitness. Great to be here. Thank you. No, thank you, man. Thank you so much for doing this. I uh I've been looking forward to this since our first conversation, um, you know, prepping for this a couple weeks ago, and you know, set the table like I was out to dinner last night and with my wife and some friends, and AI came up in the conversation, of course, right? And everyone has an opinion, right? Everyone has an opinion of AI. One guy's a contractor, one does, you know, coding uh for insurance. I do what I do, whatever that is. My wife is a VP of operations, like everyone had an opinion of AI, but no one really is qualified to have a solid
opinion of AI. Now, you, you do. And I think this is what I was telling you before is I want this to be the fundamental, one of the fundamental resources for our industry to really understand what AI is and the misconceptions around it. And you are the man to do it. So before we get into all the questions, I'm just gonna introduce you who you are. So you're the co-host of STEM Talk, which is outstanding. It is such a good podcast. I think uh you guys do such a great job of uh disseminating really important information in technology, healthcare. We'll we'll we'll get into that. You are the founder of IHMC, so the Institute for Human and Machine Cognition, a premier research organization you built from scratch in 1990 in Pensacola, Florida. Um, you're a PhD in computer science, hundreds of published papers, six books. You've been doing serious AI research for 40 years. Uh, you chaired the NASA Advisory Council, served on the Defense Defense Science Board, the National Science Board, the Air Force Science Advisory Board, and the National Security Commission on Artificial Intelligence. You uh it was the homebrew computer club in California
in the 1970s. Uh, you were part of that. Pretty cool. The same club Apple came from and has been at the forefront of every major wave in computing science. IHMC works in three areas. So primarily AI, robotics, and exoskeletons, uh, health span and human performance, which is exactly where we're gonna have the convergence of the conversation here today. So you're qualified. So let's start with this opening question, Ken. You've been doing this research for obviously since the 1980s. You know, you watched to go for something nobody really understood, nobody really talked about, the something that everybody has an opinion about, right? Everybody. What is the most dangerous misconception out there right now about artificial intelligence?
Well, there are quite a few, unfortunately, but I think one of them is a an over-attribution of ability to the machine at this point. People attribute more to it than it's capable of and attribute human characteristics to it. So you hear people constantly saying things, oh, I wonder what it's thinking about, or uh stuff. It's not thinking about anything. And and it it's very easy to sort of humanize it because we say you anthropomorphize it. Right. Because you're having a dialogue with it that seems kind of natural and it it pulls you in. You know, people actually have relationships with these things and think it's their girlfriend. Now that probably tells you a lot about that gentleman, but anthropomorphizing these machines, I think, is a a big a big problem. And then of course there's the classic issue of any powerful technology, a lot of its utility or its risk comes

The most dangerous AI misconception

down to the wisdom with what which we humans use that tool. And uh that's a big factor. You know, AI is used for wonderful things, medicine, all kinds of breakthroughs, and it can be used for very uh bad things, uh surveillance, all kinds of other we see this in China, for example.

What happens when you prompt a chatbot

What do you you know, when a user a typical user like me uses Claude or Chat GPT or whatever, you name all of them, right? What do you think like I what I believe is happening when I prompt it. But what is actually happening when I prompt it?
Right. Now that that actually is the right question, and there's a whole lot going on. But what's important is what's happening at the essence, at the core. Yes. And these are essentially large language models, typically with uh a programming harness around it. But think of it as a large language model, which is a very sophisticated way of doing statistical prediction with respect to what the next most likely tokens are given your prompt. So it's not the case that it deeply understands your prompt. It's matching those squiggles in computer science we call them tokens to the average person, they look like letters, which we think of as words, but that's not how it's actually working. It's associating probabilistically this set of tokens, and then it's predicting what is the most probabilistic outcome. It's almost like form completion. Like if you use Google and you start typing something, right? And then you'll see
it starts filling it in for you. It's using a very similar method, not as large as a typical large language model, but that's what's happening.
Okay. So it's not, I mean, as we would think, like it's your point of anthropomorphic, that's a tough word. I'm gonna have a tough time with that word. It's not thinking.

AI is a field, not a product

No. And it's, you know, what one could argue about what thinking is, but if we wanted to attribute thought to it, it would have to be a model of thought, unlike any biological model of thought. One of the issues for me in current AI, and let's stop here for a minute. I just said current AI. So AI is a field of intellectual inquiry first. And then there are AI applications. And the current most popular AI applications are the chatbots. And they all have a very similar method. They're all derivative. They have differences, but they're derivative of the same method and the same research. And that is just one part of the field of AI. But we use the terms very sloppily now. You know, if it's kind of like if people started, say at the end of World War II, we wanted to end the war quickly. So we dropped a physics on Japan. No, we dropped a product of physics
on Japan. And we we we talk about AI in sort of a funny way much of the time.
Well, let's let's take a huge step back then. What what is artificial intelligence, do you? And you've been doing this since the 40s. Like how have you seen this whole thing evolve, you know, uh yeah, over, you know, since the home homebrew computer club, right, to to where it is now. Has has the fundamental definition of artificial intelligence changed in your opinion, or is it just simply been the same and it's just evolving in a way that's coincides with that definition?

Why AI might call a penguin a typewriter

I'm not sure the definitions have changed a lot, but the what we think of when we think of AI and the technologies we think of when we think about AI have certainly changed. And AI is the effort to develop computational devices that display intelligence, that display behaviors that would otherwise be considered intelligent if done by a human, for example. And they can be narrow, like you could be very competent in some aspect of medicine and know nothing about French philosophy. Or you could try to make a general purpose AI like the chatbots that know, quote, know a lot about a lot of things, because they're trained on pretty much everything that's available to train them on. But one of the things I'd like to point out is there's not an ontology. So when you're thinking about something, let's say you and I are discussing primates, and and as we chat about primates,
I might bring up bisons. And then you would think, bison is not a primate, and because it doesn't fit in the category primate, where you might have orangutans, gorillas, chimpanzees. AI systems, that's not how they represent the world. And so one of the challenges in AI, particularly in counter AI, is you can inject noise into the system and knock it a little bit off the equilibrium where the large language model converged. And so it might call a penguin a typewriter because there's not an ontology that typewriters are not animals, penguins are this kind of animal. It has to do with the statistical probability that it got to that point. Interesting.
So that that would be a hallucination, as we would say, right?

Hallucinations vs. confabulations

And I hate that term because it implies too much, right? It's granting the machine too much. When a human hallucinates, this is typically a functioning human uh who's made a mistake or is having a an erroneous reasoning process, or maybe they're they have a medical issue that's causing them to hallucinate. When the AI systems make an error, and these are errors, not hallucinations, it's working perfectly. So when and they have a very high error rate, it's just most people don't see it. But if you use your AI system and it makes a mistake, it truly is functioning properly. It's a byproduct of how they're designed. It's in other words, it's not a bug, it's how they work. And when I say they, I mean the current popular versions. And this would be uncompletely unacceptable in any kind of other computer science. Imagine a spreadsheet that would make up incorrect answers
20 or 30 percent of the time. Or people talk about, well, we don't need humans in this aspect of medicine, we'll use AI. But you can imagine if there was a human at the hospital that was known to hallucinate, they would keep that human far away from patients. Right. And so we use these words in such a sloppy way. Maybe the best word for what people call hallucinations in AI might be confabulations. Okay. That it's confabulating, because confabulating doesn't imply intent, and it does imply that it's malfunctioning.
So explain that a little bit more. Like for those of us who don't use the word confabulate very often, like how does that uh how does that apply to what we're seeing on daily use of LLMs? Like mine, you know, quote unquote. I I don't like it either. I just I don't know. Like I'm trying to think of a project I was doing. I was I was writing a proposal and I read through it again. You have to check it, of course. And you know, I'm like, well, that doesn't make any sense. And then so I respond, I'm like, Are you sure you got this right? It's like, no, I didn't. You're right. And then it, you know, clicks, which is the, you know, also uh how it's always apologizing, right? Just really weird. That's a really weird character. Yeah, yeah, it's really odd. So yeah, maybe extrapolate a little bit more, like pull explain that uh what that confabulation means versus a hallucination.
Yes. So in the case of a hallucination, the human is in a is essentially malfunctioning, right? If you're hallucinating, maybe it's medications, you know, there's something going on with you that is definitely not operating correctly. Uh, with these AI systems, when they make these things we call hallucinations, they are operating perfectly. This is how they work. And it's a byproduct of the method at this point. We don't really have any solid idea of how to remove them given the method that most of these systems use. So confabulate is more like what it's done. It it doesn't have intent to be erroneous and it's not malfunctioning. It's essentially confusing, but even confusing is anthropomorphizing it. Right? When I say confusing, that so I just did it myself. It's not confused, it's not capable of being confused.
It's it's such a my it's such a mind blow. So LOMs and chatbots specifically, like that's what everyone knows. It's like the productized form of of AI. Maybe give some insights in the vast world of AI. Like where do like where do chatbots fit in? Like what percentage do you think roughly they are of the entire AI industry or AI evolution? And what what are some of the other forms that we're just simply not aware of as consumers?

Chatbots are 99% of what the public sees

Oh, they're they're 99% of the public's experience with AI because really it's the only AI that the general public has access to. You know, it's it's it's what is is available and is very captivating. It sort of uh pulls people in to thinking they're having a conversation with an intelligent entity. And you know, uh you could squint and think, well, it's like some kind of intelligence, maybe, but it's very different than than human intelligence. But in terms of the field, the field is incredibly broad. Machine vision, machine learning, which is used in chatbots heavily, uh, but machine vision, um all you look, even your car is full of AI. If you have a modern car, your transmission has AI in it, the control uh modules for the engine, modern cars have AI, but they're not chatbots. You know, none of the AI I worked on uh my career were chatbots. So it's a broad, broad
field, includes almost a big chunk of what you think of as computer science, but particularly the parts having to do with intelligent behavior. And not all chatbots, you know, GPS systems, if you're using your phone, modern phones are chock a block with AI.
So you just said something interesting to me. You said intelligent behavior, which is different than intelligence itself, right? I mean, when we look at is is this the right term? Is artificial intelligence the actual right term, do you think, for what's developing? Or is it just we're displaying intelligent behavior through technology, through you know, these mixtures of machine learning and LLMs and and all and all of those pieces?
Yeah, I dislike the term artificial intelligence and have complained about that for decades. You know, if you think about it, it's it's a very strange choice of terms. It they came up with this in the late 50s and it became more popular in the 60s. And the idea artificial intelligence, what a strange term. Heavier than air flight at the turn of the century, not this last century, 1903, 1902, and that era. Heavier than air flight was often called AF, artificial flight. And these aircraft actually were flying, but they didn't land in trees, they didn't poop on your car, they were not very bird-like, they didn't have beaks, though some people put beaks on them. I see a similar thing with with artificial intelligence. It's an unfortunate term in in in so many ways. A better term would be something like augmented intelligence, because it puts the human in the center. So

Why "artificial intelligence" is the wrong name

the AI at IGMC always has the human in the center, not off to the side, but in the center of the subject. So we like terms like augmented intelligence because it's augmenting my intelligence. I'm I'm I'm the reason that it exists. And uh I I think that perspective is too often lacking in discussions about AI.
Yeah, well, I think that brings us to a good point because in our uh prep call for this, you had you had a really good uh way to per to to present this, like it's orthotic versus prosthetic, right? So maybe explain because that that's that you finally got through to me on that one, like my dense head. I was like, okay, that makes sense. Thank you, Ken. So maybe explain that to the audience. Like, what did what do you mean by that?
So for much of what I think about AI and others here, we tend to think of it as a cognitive orthotic. Uh you could think of my eyeglasses as an ocular orthotic, or many of us have a knee brace. There are all kinds of orthotics. It improves the human performance, it braces me, it makes me more capable. A prosthetic is a very different thing. And the last thing I would like to have is a cognitive prosthetic, because that would imply I didn't have the functioning brain. If you have a prosthetic leg, you're missing a leg. So that's a bad vision for AI, in my view. This was brought to my attention, and I'm a little bit ashamed to admit it, it was brought to my attention maybe 30 years ago or something like that, 25 or 30 years ago, and I had a talk called Toward Toward Cognitive Prosthetics. And I hadn't really thought through the use of that word. I I I meant what we've been discussing, I meant a cognitive tool that

Cognitive orthotic, not prosthetic

extends my performance. But I uh and it was a popular talk, and I'm on the break, and these two uh physicians were in the talk in the room, and they came up to me and said, That's a great talk, but I don't think you mean prosthetic. And and I thought about it and said, Yeah. And so that prompted me to change that to orthotic, cognitive orthotic.
Okay. And and what is that? Like, how would someone use a cognitive orthotic effectively?

From user to teammate: AI autonomy

Well, the the more seamlessly it connects to the human, and chatting with it probably isn't the most seamless way, but it's the way we do it now, the better. So any orthotic, the better it fits you is the key. Right? It's the goodness of fit, has a very large degree of determination on its performance. Uh, that's true for eyeglasses, it's true for knee brace, whatever orthotic you have. So I'm very much in favor of human-centered approaches to AI where the human is at the center of the topic, and we have to study the human and understand how we could extend human performance through these computational devices rather than be so focused on building a standalone intellect that we somehow have to prove to people that there's no human involved. You know, it's sort of smart by itself. Okay, that that's great. And now we're into an era in AI
where we're really talking frequently about autonomy. And this is a different perspective where you're not so much trying to extend the cognitive performance of a single human or team of humans, but rather with autonomy, these this has a scale to it and a time dynamic that you can't keep the human in the loop. So whether it's a self-driving car or a self-flying airplane and hundreds of other kinds of autonomy, you have to change the relationship. The human isn't so much the user anymore of the device. That relationship switches to something much more like a teammate. Okay. And that changes everything. And when you make that change, then you're looking at interdependency relations between you and your machine. How are you interdependent? So if you think of a self-driving car, you can build lots of intelligence into the self-driving car and a lot of capability. But if you still have the relationship of
user to the car, then you will never actually have the best functioning outcome where both the human and the machine are functioning as a team to accomplish the task.
So the term agentic is thrown around a lot over the last year, right? I think there is what was the uh the the claw incident, right, where that got loose allegedly, but then maybe it wasn't even AI, maybe it was just people. But what is maybe help me explain agentic? Because I used it the other night in a conversation. I'm like, I doubt I used that correctly. So when we're when people are talking about agentic now, is that similar to autonomy or is it different? Is it related?
Maybe it's a little related, but mostly different. Basically, well, you're using an AI system to perform some tasks that you don't have to be sitting there waiting for it to do this task and then this task. It'll take a role or it'll take a task. It its performance is typically vastly exaggerated currently. And so the it's interesting, the AI companies themselves, particularly OpenAI with ChatGPT and also uh Anthropic uh with Claude, they delight in frightening the public. It's the strangest behavior of a CEO I've ever seen. Can you imagine a CEO, say, of a drug company, and say, look, I don't think these drugs should even be legal. And this drug just got away from us. We don't know how it did it. Yeah. Or an automobile manufacturer. It's unthinkable. But somehow in this field, this grossly irresponsible behavior of the

Agentic AI and CEOs who sell fear

executive teams has become almost normalized. And, you know, and these things don't escape. And you know, it it's incredibly uh it's like hype-driven because if you give the image that your AI, your new model, is so advanced, the world might not be ready for it. And then if the government rolls in, which they did with one of Anthropics products and says, oh, it's too dangerous, well, then no foreign dissemination, you can't have any foreign workers, and that quickly ended that discussion. But you know, that that's an annoyance to most of the technical people in AI. These stories that are made up and the media picks them up. Probably the worst is the New York Times. Uh, that probably would surprise you. But they come up with just these fanciful tales that actually never happened.

Sponsor: Perfect Gym

Hey friends, artificial intelligence is moving faster than most of us can keep up with. That's no secret. We can all feel it. The pressure to keep up is very, very real. And no, I'll get to that next week is not a strategy. So I am proud to say that our presenting sponsor is Perfect Gym. The people building tomorrow's infrastructure for gym and health club operators. I have sat down to talk to thousands of leaders across the fitness industry, and one theme keeps coming up. The right tech can make or break your business. Perfect Gym isn't just another management system. They are part of the Sports Alliance Group that is Europe's leading fitness software company, and it's now all in on the US market. So in the age of AI, being data ready isn't just a buzzword. It is critical for survival. This, my friends, is the disruption we have been waiting for. Member management, billing, scheduling, reporting, one platform, zero duct
tape, single studio or multi-location franchise that scales with you, zero growing pains, and their open API connects to over 50 technology and payment partners. No closed ecosystem, no getting boxed in, just better data, better decisions. Perfect GM is trusted by international fitness chains because they focus on what actually matters. That's operations, retention, and growth. The European market has spoken. Now it's America's turn. Check it out at us.perfectgm.com. Which, you know, at the core of it, man, I think the the period of time we're in right now is so the technology is advancing so quickly. And the stories that we built around it, the beliefs that we have, the uh differing opinions, like it's it's it's a crazy time to be alive because we all feel like we're rushing towards the edge of a cliff and we can't stop it. Right. And no matter what, but all we can do is just
argue about it over dinner tables or in conference rooms. And, you know, it doesn't really matter. Like this is just happening anyway. It's this, it's this strange. I mean, I I grew up in the Cold War too, right? Which was like the same thing. It's like, well, you know, if the siren goes off hide underneath your desk and hope for the best, right? Like, like that's gonna help.
Yeah.
But I feel like it's it's very similar in eeriness where it's like, we all just feel, yeah, we just feel helpless, but we have an opinion about it, yet we're using it all the time. Like we like it, but we also are extremely scared of it. I mean, when you look at this, how it's affecting human psychology at this point, like, do you have any advice for us? Because you know this stuff better than most people. Is this is this founded?

When AI does your thinking for you

Yeah, it's kind of worrisome in a way. I mean, I have friends, I don't mean AI researchers, but other friends in my life who are generally well-educated, sophisticated people, you know, at the top of their career, and they anthropomorphize the AI like crazy. And I'll point it out to them nicely, I hope. And uh but they keep doing it and they keep vacillating between a love affair where basically the AI has become their prosthetic. You know, I have a lot of friends that don't write a thing without AI writing it. Yeah. You know, and and and of course you can tell if AI wrote it. I mean, I can look at some text and tell you what the role of AI was in that text, and then there are programs you can run it through which are much better at that than I am. And of course you can defeat them, you know, these programs as well. It's like a cat and mouse game. But, you know, it doesn't even bother me that much if people have AI help them write something if it if it's their thoughts
and they wrote a draft or they and the AI is sort of working with them and and they're thoughtful about it, and they think of it as a tool. Like it doesn't bother me that engineers use calculators, right? It's a tool. But a lot of people have sort of made it into a prosthetic. It's it's doing the thinking. Thinking's the wrong term. They think it's thinking, it's producing plausible text, right? And the text is plausible. When you see the text, unless you know a lot about the topic, it often is that's probably true. You know, it they'll have references, but some of the references are irrelevant or to papers that don't exist.

A 45-minute proposal and the guilt that came with it

You know, it's it's a I I wrote a post yesterday, I can on LinkedIn about how I had done, I did a sponsorship renewal proposal in about 45 minutes. I pulled in all the data, all of the numbers for the year, proposed some things, you know, and I sat there and I talked to Claude and we worked it out. And I fed in information. And it took me 45. That would have taken me four or five hours two years ago, right? To collect all that and put it together in that format. You see. What I what I wrote about was like, I don't know why this, I feel like the four hours would have made me feel better than the 45 minutes, right? Because I felt feeling unaccomplished, beautiful, you know, way better than what I could have done two years ago. But there was something to it about like, and I wrote as like, is this, do I why do I feel guilty about this? Like, is that what it is? Like, and I was trying to figure it out, and I feel like is what a lot of people are feeling. And then I'm like, well, no, because if it was a calculator or a spreadsheet, is the examples I use. Like, I no, it's my work. And you can't, you can't take nine years of running interviews and podcasts and 10 years of running a gym and all that and put it all in
L like that's that was the value of it. But the point was, I just feel like a lot of people are probably sensing that too. Is like, is this my work? Am I allowing my thinking to degrade, my writing skills to degrade? Am I gonna regret this someday because I just gave it away to an LLM?
So it's a very strange time. It's important how you use the tool. So people, too many people, as I mentioned earlier, are using it as a prosthetic. It's it's doing the bulk of the work, and then they sort of tweak it a little and polish it up. And it's better to think of it as a tool that's supposed to increase your performance and amplify your abilities. And if that's how you're using the tool, then I I'm not too worried about it. But I am concerned about you see a large swath of people that use the current AI, which, as we discussed, has issues, but it's very powerful. They use it as sort of a substitute for their effort, for their deep thought. I mean, a good interesting example is you know, you see pundits write about AI, and some of them are AI critics, or they're people who are thoughtful people writing about AI, like the current Pope, and
you realize a lot of that seemed to be generated by AI. And so you think, wow, that's interesting. Um and it doesn't bother me if it reflects the actual perspectives and thoughts of the person writing it. But it it does bother me when people say, write me a story about, you know, and AI writes it and then they tweak it a little, and then they'll say, Oh, yeah, that's that's the story.
Yeah.
Writing's not great.
Yeah, it's not. And uh, I think you have to, when you do publish something that has been like I wouldn't consider it a priority piece of mine. Like if I'm gonna have it right or multi-purpose some content for whatever, something I like Instagram. If I put it on Instagram, I don't really care what happens on Instagram. So it's like that. But if it's not, it's one of my fundamental p pillars of content or my thought, you know, I'm gonna want to make sure that it actually sounds like me. It has my character in it, right? Uses the same words I do. Like one of the words it uses all the time that bothers me is actually. It always says what actually works, right? Or what you know, that it it's such a telltale scene.
It uses quiet a lot. But the the different chat bots have different favorite word usage. A lot of them use a lot of long dashes, you know, M dashes.
The dash, yeah, yeah, yeah. Yeah, yeah. Yeah. Well, I I want to get in a lot more into like the health span and humor performance at IHMC because that's something that's you know, obviously very I can talk to you about AI all day. That's I feel like that's selfish. So a third of IHMC's work is health span, resilience, and performance. So you guys are at the cutting edge. Like, what does that work actually look like within the lab? Like, what are you guys what are you guys doing there?
Well, there's a wide range of work in that area, and I'll tell you about some of it. Uh the large building you can see behind me, uh that entire building is health span, resilience, and performance research. I would say uh a good chunk of it is aimed at the military. And so the military is one of our big research customers, DARPA, the Air Force research labs, Navy research labs, the military more generally, and also in particular the special operations world, and because they're very interested in performance, right? I would imagine. So that's a big chunk of the work, and that ranges from physical performance to cognitive performance to everything in between, including resilience, which is very important to that community. Another big sponsor of work in that building is the National Institutes of Health, the NIH. And uh there we're interested in a lot of things with NIH, but in particular, we're very interested in the aging population. And everyone

Inside IHMC's health span lab

is aging if they're not deceased, but the process of aging and how to age better. So some of the studies for NIH have to do with resistance training, for example, for older folks. You know, uh, even for young people, the responsiveness to a well-designed resistance training program has a high degree of variability from person X to person Y. As a trainer, you've seen that. You do the same thing. This person responds incredibly well. And this person with the same level of effort, less. I mean, everyone responds, but less. And in the older population, it's been noted that that difference may even be more, you know, due to lots of issues. And the government would like to know what those issues are. How much of it is it hormonal? How much, you know, what's going on with these people? And are there biomarkers that would predict a responsive, a positive response as opposed to a mild response? So those studies are large studies
with hundreds of people and two other institutions. We we're the lead and we coordinate it, but there are two other institutions in that particular study where we're trying to give them a really intense resistance training program and for a long period of time and measure the changes, the epigenetic changes, any kinds of biomarker changes, you know, we look at the whole package. So that would be an example of NIH work. Another example of NIH work would be metabolism related mostly to mitochondria. So mitochondria becoming more and more important and linked to more and more of the hallmarks of aging and disease, and performance, of course. And so from our point of view, our mitochondria lab will become increasingly busy. It's busy now, but it will become increasingly busy. The first floor is mostly cognitive performance or human machine teaming, that kind of thing, mostly for the special operations world. And then the rest
of the first floor is a large intervention center, which looks like an extremely well-equipped uh gym. Uh, and then uh on the sides there would be things for like cardiac evaluation, VO2 max evaluation, gate evaluation, that kind of thing. And the core of the floor looks like what it is, a large fitness facility.
Uh I would love to see it someday. That that sounds really, really cool. You know, one of the things that you and I talked about too was like longevity, right? So that category is blown up. And, you know, I think it's you know, when I think of longevity, I kind of think of old rich white guys, right, paying a lot of money for to just extend their life a little bit longer. My words not anyone else's, but that's what I think most people think of. Uh I think the term is interesting. I like the way it repositions what we do as an industry here into like this is no longer a fat loss one-year endeavor. This is a lifelong journey, right? Like you have to continue to do these things for a long time. This is critical for your health. On the other side, it's damn good marketing, right? Like it gets people to buy a lot of different things that, you know, are really unproven as far as science. When you look at the field of longevity, and I think you and I both appreciate health span a little bit better. I think that's a better term. But what what are some of the misguided components of this world of longevity? Uh, much of it.

Living to 180 and longevity hype

The the top quality researchers related to longevity are are fine and they don't make ridiculous claims. But most of the work isn't so much about longevity. It's about what you said. It's about extending health span, the period of time where you're functional and can really do most of the things that you need to be able to do physically and cognitively in your life. You know, there are in podcast land, it's really bad. There's this one guy that talks about living to 180, and he's sure he will. Now, I I actually find it hard to believe he thinks that, but but if he does, boy.
Right. And and would I mean I don't know. There there's a whole philosophical question around that, like, would you want to live in 180? Like No, and but I don't need to worry about it. Almost no one does. What is the long well, how much what's the longest recorded lifespan of someone? Is it like a hundred years?
It's controversial because a lot of times it's unclear that the numbers are accurate. But there's quite a few that I think are in like the 120s, but that's extraordinarily rare. When I say quite a few, I mean like a handful, you know. There was a French woman, I can't remember her name. She might be the person lived the longest with a clear historical record. In other words, we're pretty sure when she was born.
For our industry here, you know, let's let's just say Avatar like uh gym and health club operators, which is probably the constituency mostly represented by my bio, it's like what can we learn from the research you're doing? Like we've always known fitness is good for you. Right? We it's I think it's pretty undeniable at this point. You know, but what I'm seeing is, you know, and what I'm pushing for and helping trying to help it happen is like this convergence of fitness, health, wellness, all those terms, like it's all coming together. And I think health clubs and gyms are very well positioned to be at the forefront of that. So if you were gonna give us some advice as an industry, how would you tell us to pursue that that goal?
Sounds like you have a excellent handle on it from my perspective. You know, I I think too often each of the key aspects of health span and generally wellness are sort of stovepiped in into separate categories. And it's opportunity, I think, for fitness centers in particular to play a more integrative role. Not a medical role. That's a quag bar. And uh I what I mean though is a a role broader than just it's time this is your kettle bell day, this is or this is your leg day, a deeper relationship with the customer that goes beyond includes all that, but it goes beyond that. And I I do see a lot of people talking about that and thinking about it now. And it's a little tricky because you very quickly start sliding into topics that are essentially medical related, and that's that's an area to be cautious moving into for regulatory

Where gyms fit in health span

reasons and other reasons. But I think it has to happen, and it is happening. You know, you the amount of people that are interested in this topic now compared to 20 years ago, it's incredible change. You know, I I meet people just they don't even know, and they'll like on an airplane or something, and they'll start saying, Oh, I listened to the episode in STEM talk about mitochondria, blah, blah, blah, blah, blah. You know, that never would have happened. Yeah. People are really fired up these days about health and wellness and health spend. So I think it's it is happening, you know.
It is. It's it's a really good time of, you know, uh, God, who said this? Uh, maybe it was Stephen Powell, who I think you you and I both know. Um, you know, consumer-driven healthcare, which was the first time I've heard those words put together. And when he said it, and maybe he made it up, maybe he didn't. But I was like, oh, that's that's really good, because the consumer is demanding it now. You know, and I look back like at my career as an operator, gym operator, was I would get frustrated because, you know, I wanted most of the people who came in, there was health issues, right? And I felt so constrained. I'm like, all I can do is write a fitness program here, you know, and I I tried to educate myself and got into, you know, areas that I I soon realized, I'm like, I'm way out of scope. I should not be here. And then I try to work with medical professionals and try to have this relationship, right? And it was very difficult because nay, we don't speak the same language. Um, there's not the same level of respect for the fitness industry as, you know, a medical professional and a fitness instructor are just not going to get the same level of respect. We can go into a million reasons why, but I think that's actually justified. But now I feel like we have an opportunity because, you know, I think
the the language that needs to go back and forth can now actually be data. I mean, you look at like wearable data, you look at a lot of things that could be tracked. And I know at a high level, like what's done in the lab, we all know that fitness creates better health outcomes, right? If people stick to it, this is exactly the research you're doing with strength training and things like that. But if we could do it as like an N equals one level of like, okay, this particular person is coming into the gym and following a routine, and that gets back to the medical side that says, oh wow, this has like dramatically improved their markers and across X, Y, and Z. That to me is a beautiful world. And I'm wondering, do you think that is possible in the near future that something like that could actually take place?

The two kinds of motivated patients

I do. I mean, there will be resistance to it, of course, and a lot will depend on the nature of the medical provider that is involved. And a lot of my physician friends tell me now that they have two kinds of highly motivated patients. So they have a whole slew of patients that are not motivated at all. But among those that are highly motivated, they have one group that comes in with tons of data, you know, from their uh wearables and uh all the kinds of tests the physician wouldn't order. And uh they're almost almost hard for the physician to deal with in the 10 or 15 minutes the physician has, because these people are typically on insurance through where they work. And the conversation is short, and doc doesn't really want to get into most of this, and maybe he's not equipped to get into a lot of this. And then, you know, then and then you have the other kind of highly motivated patient that just he the docs feel like, yeah, that person is gonna do what they want to do. They're doing all kinds of things they
don't tell me, and it's easy to do it in 10 minutes.
Right.
Neither is ideal.
Yeah. I mean, you know, with uh I'm I guess you could say I'm a biohacker, right? I don't know. I don't like that term.
No, I don't know.
I'm very I'm proactive with with my health. And you know, when I go to see my my primary, which really I just see once a year for 15 minutes, like he asks me questions. I don't tell him anything. I'm not gonna tell him that I tried, you know, BCP 157 when my knee hurt, right? Or like anything else.
Uh because I just don't want to You really can't, and he doesn't want you to. Yes. Uh believe me. Uh it's a very rare doc that would want to hear about that and that would take the time to have a meaningful discussion about something like what you just mentioned. And this this is causing a huge disparity in healthcare where people that can of bigger disparity than people realize, people that are highly motivated and can afford a physician where they spend, say, three hours every time they see a physician, and they see their primary maybe three or four times a year for three hours each time. They have a very different outlook and it and a real advantage compared to the person who only goes once a year, they get 15 minutes, they don't want to tell that doc much if it's innovative what they're doing. He might freak out or she might freak out. They will not like hearing it in most cases. And the doc really can't spend enough

The 15-minute doctor visit gap

time to explain whatever it is that say that let's say the physician was uncomfortable when you told them about the B BC 157. They don't have time to have an in-depth conversation with a well-educated person like yourself about that peptide. It just can't happen for most people's insurance coverage.
Yeah. Yeah. And I think that for this reason, you know, kind of circling back into AI, I feel like, you know, and we're seeing a lot. This is wealthcare. Like to do it right, to have that three-hour consultation, it costs a lot of money. It's expensive. And, you know, I'm seeing a lot of really cool models where like medical is kind of coming under, it's all coming under one roof where it's like, okay, you have the healthcare professional and you have the the gym and the trainer and the PT and the, you know, all the stuff, all the cool stuff. But it's expensive, right? It's like you're looking $1,000 a month or more, and most people can't afford that. And obviously, if we want to see this thing kind of disseminate to more people and become more accessible, I feel like one of the biggest promises of AI is that is that like it can be the health doctor in your pocket that you can talk to about it already is for most people, right? Most people who are using ChatGPT are probably asking it health-related questions. I mean, that's just happening. So, where do you see artificial intelligence role in the future of healthcare and maybe how you know do you see fitness specifically working into that vision?
Yeah, it's important to remember that when ChatGPT analyzes your medical record or your latest, say, quest blood tests, it is often in error, and it's hard for a lay person to know what is error here, or people like to call it hallucinations, as we discussed, and what is correct. Uh and so it is a little bit tricky. I know people that actually do that because they don't have any time with their doc and they can't afford like $5,000 a month for the other model. And so what they do, and they're smart people, so what they do is they're they they have this same conversation with multiple AIs and they basically have them check each other. It doesn't get rid of the hallucination problem, but it does help it. And a lot of how well AI performs, particularly in something like medicine, is how well you prompt it. So if somebody is very used to AI

Using AI on your lab results

and knows how it works, they can prompt the system that will get a better, more accurate answer than a person that's not sure how the thing works. They're asking the same thing, but in a slightly different way. You know, it's like the uh the the old story, uh, I think it was last year or two years ago. Um a publication was shocked when they asked, I think it was, I forget which AI they asked, but they said, you know, uh, car wash is just right there. It's 50 meters away. It's a nice day. It's a really nice day. Should I should I walk to the car wash or drive? Right. AI repeatedly said to walk. That's good for you. But then you don't have your car at the car wash. And that that kind of error is quite common, and that was driven by how it was prompted. Because in the prompt, the thing that stood out to the AI was the 50 meters. And you could ask the same question in a slightly different
way, and the AI would have given the right answer. So prompt engineering is real. You know, it's a real, it's a real thing, and people are being clever about how they're using these AIs. They're sort of bouncing them off each other and checking things. The other thing I like about it is when they do go to the doc with their 15 minutes, the smart people are loaded for bear. And they don't necessarily tell the physician every uh quirky thing they did. They might not share BPC-157, but they have a very carefully orchestrated printout, and they bring it and they just go through it quickly. You get a lot more done that way than sort of an open, unstructured, well, how do you feel? I feel good, you know, that kind of thing.
Yeah. You know, it's interesting. I am I'm formulating a new business entity and I met with a business attorney yesterday. And I've already obviously inputted everything into Claude. And when I showed up, you know, I'm like, hey, here's a two-page summary of everything I want to do. And we went through it. And what probably would have cost me five, six hours of billing got done in an hour.
Great example. Yeah. People do that in medicine, I think it's wise. They they shouldn't just assume the AI is right, but that's why you have the physician and you bounce it off several AIs, and then you bring a nice, neat, efficient document. You can just blast through the document. Yeah.
I mean, do you think it's gonna get better? Yes. Right? It's gonna get better and better. Do you do you feel that AI can transform the way we do healthcare? I mean, is it problem? I mean, there's a lot of fighting, a lot of forces that will resist a lot of change in the current healthcare system. For I'm not saying, you know, it's nefarious or not. I I don't know. I'm not in those rooms, but like, do you think it fundamentally can shift the way because we got problems here with healthcare. It's it's a problem. We're really good at some things, really bad at other things.
Yeah, so when you test AI, and I don't like it when we do this because it's like John Henry in the steam engine. You know the story about the uh most powerful railroad worker. I think he was a super strong guy, and he was 100 years or 150 years ago, and he was in a race against a steam engine who could drive the most stakes, I think, into the or pilings. And poor John Henry, at some point, I think he had a heart attack, and that was into John Henry, is he was racing the steam engine. And too often with AI, we do a version of John Henry and the steam engine. So we're usually comparing it to a human physician or that that's really a daft approach. You know, it i it'd be like if uh, you know, we compared calculators to mathematicians. Most of what mathematicians do is not arithmetic. You know, it's it is a different thing. So when you do the comparison to human physicians with high-end AI, the AI usually

AI vs. doctors: John Henry and the steam engine

does better than they do on diagnosis on all of that kind of aspect. But then when you dig deeper, well, and licensing exams, the AI gets almost a perfect score. And uh the docs struggle with this, as many will confess. However, differential diagnosis, the humans do way better. So the humans will hold in their mind it could be this, it's more likely this, but if it is this, it's really bad. I hope it's this. And the human doc will hold, they call it differential diagnosis, they'll hold several options, three or four at times, open. And then they seek evidence in support or refutation of one of those. The AI tends to jump on one, the one that's statistically most probable, jump on one, and that's it. So in test taking, that does pretty well, really well. But if you were its patient,
you probably wouldn't do so well. So the in medicine, it's a great example of the AI, AI ought to be a cognitive tool for the physician, an enhancer. You know, I I had a Zoom recently with people that head the United States federal programs uh in medicine, you know, and they were asking me about what would be the right test. And they were thinking of things like John Henry and the steam engine. Well, human physicians make mistakes as well. It's better to think of it as a a tool. The test for success isn't the AI against the human. The test for success is do the docs do a better job with the AI or without the AI? Right? It's like my eyeglasses, right? The test test is on my head. Right. What we're doing with AI is akin to if we took my eyeglasses and put them in front of the eye chart, and then I had a contest against my eyeglasses. It makes no sense, right?
These tools exist for us.
That's a that's a great that's a great comparison. Okay, so I know I I don't have you for too much longer here, but I I gotta ask this question. In a scale of one to ten, one being Skynet, right? And ten being uh this abundant utopia where we all get to do exactly what we want and life is good, right? And great, even it's it's perfect. Where where are you on that sliding scale of where do you think AI is bringing us?
Somewhere in between. But I don't think AI is bringing us there. I think it is the quality of our leadership and the thoughtfulness of which we as humans employ the tools. You know, uh, it's a very powerful tool and uh it's often attributed more performance than it actually has, but I don't think AI is finished. If if what we have now is what we have is just a slightly better version in 10 years, I'll be very depressed. But this has actually become a problem for AI research because the a lot of the policymakers and politicians know literally zero about how it works. And so they think they play with it, you know, and they go, oh wow, it's a solved problem. All we need is a bigger data center because the performance increase in the last 10 years has been driven almost entirely by scale, not by cleverness, by scale. Well, this is unsustainable. So currently

Skynet or utopia: Ford's verdict

the data center power requirements are growing at half an order of magnitude a year. Half an order of magnitude a year. So that's roughly 3.2 times 3.2 a year. So we have data centers now that are under construction that will require more power than some states, entire states. And the water demands are incredible. So you have to think about look, the human brain is what, 20 watts? You know, these things are talking about large nuclear reactors to run one data center. I think what we need to be doing is research, fundamental research in AI that will reduce methods that are more controllable, more efficient, and that scale in a way that you don't need to build these giant data centers. But the approach we're using, we're locked into this. And it's a problem because when you talk to the political leaders, and I often do, they're dismayed to hear they're like, really? I thought this was solved. I thought all
we had to do is build bigger data centers. No. We need it, uh we need new methods, clever new methods that would allow us to be more efficient. And it would be nice if there were less quote hallucinations.
Yeah.
Wow.
Ken, this has been so great, man. I really appreciate you taking the time to you know specifically kind of talk to this industry. I think it's it's really important that we get experts like you to coming in and STEM Talk, amazing podcast. Maybe take us tell us a little bit about that one, uh, why you do it. You seem to have a really good time with it. At least that's from a listener, that's what it sounds like. And then where else would you like people to go to kind of follow or or be in touch?
Or yeah. Well, IHMC has a website that's open to the public. Uh, in terms of STEM talk, I really do enjoy doing it. And the gentlemen that have been helping me with the technical work with your podcast here are very involved in STEM talk. The idea in STEM talk was deep dives, you know, hour and a half kind of deep dive, not talking down to the audience, explaining things if they need explaining, but a lot of the listeners email and say, you know, I use AI and I use Google 25 times in that podcast, and that's okay, right? I don't want them to feel that a lot of science podcasts are sort of not scientists talking to each other. It's sort of a watered-down version where the subtlety is lost. And so we're hoping for audiences that appreciate that and that are thoughtful audiences. And so that's why why we we do it.

Why STEM-Talk goes deep

That's great. I I lie. One more question. What what can we help you with, Ken, as an industry when you look at this fitness and wellness industry and the work that we're doing? Is there anything we can we can help you with with the work that you do or the concepts that you have that you're pushing forward?
Well, I'm I'm sure there is. A lot of the health and wellness industry, particularly on the fitness side, which is sort of merging, as you said, with other aspects, uh, is very informative to what we do. And a lot of times we learn first about things from actual trainers, and but they don't have a large enough N, right, for the for a study. They'll say, look, I've had this experience with a lot of clients. You ought to investigate that. Or I don't know why. Maybe there's a biomarker, maybe maybe you can look at epigenetic factors. And so we can do all those things, but a lot of times you get the ideas from experience, from people who've sort of had boots on the ground, so to speak. Because if you just stay in an academic place, you can very easily get disconnected from what people actually are doing in the field. So many of the people here that work in health spending and resilience performance, you know, we have an incredible facility, but a lot of us actually

What trainers teach the research lab

like to go to other places to see what uh people are doing in the real world, you know, and outside of that our facility. So we do learn a lot from that community.
Yeah, I mean, fundamentally what our industry does, which I think has been misrepresented so many times, is we're we are in the business of behavior change. We can have the best laid out plans, but it's that human-to-human interaction, the community, all those things that we value as an industry that that changes one's behavior and almost helps them re-identify as this new person that they're becoming. And that's equal parts art and science. And you guys have the hard science, but I feel like our industry has a lot of the art, right? That we can totally agree.
Totally agree. And it's one of the best and most meaningful changes a person can make in their life is to take some responsibility for their physical condition, no matter what the age.
Amen. All right, man. Ken, thank you so much for doing this. It's been an absolute pleasure. Yeah, this has been great. I'm super satisfied, and I'll be reaching out. I have some suggestions for people you can keep an eye on to.
Great. Thank you very much. I enjoyed it.
Yeah. Ladies and gentlemen, Dr. Ken Ford. Thank you.

Why this conversation matters

Show notes by Dani

Dr. Ken Ford has spent 40 years in AI research, and he says chatbots aren't thinking at all. His test for when AI helps comes down to a pair of eyeglasses.

FREE REPORT: Get the Future of Fitness quarterly industry report → https://q2-report.future-of-fitness.info

Eric Malzone sits down with Dr. Ken Ford (founder and CEO emeritus of the Florida Institute for Human and Machine Cognition, co-host of STEM-Talk). Inside: why chatbot errors come straight from how large language models are built, why AI scores near perfect on medical licensing exams but falls short on differential diagnosis, and IHMC's NIH resistance training study searching for biomarkers that predict who responds best in older adults. Ford also explains why data center power demand growing half an order of magnitude a year is a dead end, and what gyms can teach his lab.

WHAT YOU'LL LEARN
• Why AI "hallucinations" are the system working as designed
• How to use AI as a cognitive orthotic instead of a crutch
• Why AI aces licensing exams but misses on differential diagnosis
• What to bring to a 15-minute doctor visit to get real answers
• Where gyms fit in health span without stepping into medicine

"The test for success isn't the AI against the human. The test for success is do the docs do a better job with the AI or without the AI?"
(Dr. Ken Ford)

"The last thing I would like to have is a cognitive prosthetic, because that would imply I didn't have the functioning brain."
(Dr. Ken Ford)

GUESTS
Dr. Ken Ford, Founder and CEO Emeritus (Florida Institute for Human and Machine Cognition)
ihmc.us

Thanks to Our Sponsors
EGYM: Connecting smart strength equipment across the fitness floor with personalized, AI-powered training experiences: engage.egym.com
Perfect Gym: A leading fitness club management platform, part of Sport Alliance, now expanding across the U.S. market: PerfectGym

Get the State of Fitness quarterly report: https://q2-report.future-of-fitness.info
Partner with Future of Fitness: futureoffitness.co
Follow Eric on LinkedIn: linkedin.com

ABOUT FUTURE OF FITNESS
Future of Fitness is the business podcast for gym owners, franchise operators, health-tech executives and wellness entrepreneurs, hosted by Eric Malzone. Subscribe for conversations with the people building the business of fitness.

From Insights

Enjoying the show? Rate us on Apple Podcasts ↗ It helps other operators find these conversations.

Up next

Continue the conversation