← The Good Question Podcast
The Good Question Podcast
Next-Level Human Performance Thoryn Stephens On AI, Brain Health & Longevity
00:21:06
Loading summary
Transcript51 lines
- [00:01]
A
You're listening to the Good Question podcast with Richard Jacobs. Our goal was to make each of our guests exclaim, hmm, that's a good question. I don't know the answer. Because when that happens, it means you, the listener, may be inspired to learn more beyond the interview and to ask great questions yourself that lead to new insights. In this podcast, we cover historical and current anthropology, comparative religion and history. Welcome. And let's get started.
- [00:36]
B
Hello, this is Richard Jacobs with the Good Question podcast. My guest today is Thorin Stevens. He's a scientist. He's working on adaptive AI to model and hopefully emulate human behavior. He's also a data visionary, keynote speaker. A pioneering space where adaptive AI, human behavior and biology intersect. He's the founder and CEO of Brain One. So we're going to talk about what he's doing in the world of AI and it promises to be very interesting. So, Thorin, welcome.
- [01:01]
C
Hi, it's so great to meet you and happy to be here.
- [01:03]
B
Yeah. Well, tell me, have you been around in the AI world for many years and now it's kind of a new thing for you that you're heavily engaged in, or have you been working in the trenches for a long time and now things seem to be coming together?
- [01:15]
C
That's a great question. So I began my career as a molecular biologist and focused in biotech and drug development. And this was in the 2000s, essentially when I graduated college. We just sequenced the human genome and at that time we were doing bio informatics. It was really that intersection of biology with advanced computing. I really got into AI, I'd say roughly, probably like early, you know, 2010 ish. And as I transitioned out of biotech into essentially data science, it was through that experience that I really began to, number one, have access to these massive data systems and then began to dabble into machine learning and had a number of mentors there that were very innovative and early in the space, including Dr. Galen Buckwalter. And Galen. They built, you know, they were doing production level machine learning at a company called Eharmony, you know, back in, gosh, 25 or so, you know, 2005 in forward, so about 20, 30 years. But anyway, I know. So, yeah, very much ingrained into kind of my worldview and then definitely into my most recent research.
- [02:17]
B
I have a question, why, why all of a sudden, you know, AI has been around for a long time. Perceptrons, all kinds of different models. Why all of a sudden in 2023, was this a huge breakthrough in terms of ability? Like what, what do you think caused wonderly?
- [02:30]
C
I think part of it is computational power. Price has dropped so substantially. I mean we were doing, when I, as I got into data science and I was doing more consumer research and testing and optimization, we were doing, you know, production level AI, you know, with Meta in brands like man, roughly 2015. So let's say 10 years ago. And I mean even back then, you know, it was still a bit clunky, you know, relative to the models and the data management side of this. And it's just got so commoditized and really productized that, that it's, you know, really part of our, our day to day, you know, at this stage.
- [03:11]
B
So you think the amount of computing power created this emergence of ability? There's LLMs. I mean it's weird, you know, they're, they're essentially interpreting what would be said next in a conversation. But they seem to be intelligent even though they. How could they?
- [03:24]
C
Yep, yeah, I mean a hundred percent. I mean being in the trenches you know, for at least a decade or so, I mean what we've seen is that you know, first and foremost, oftentimes when companies say they're using AI or machine learning, I mean number one, that's not necessarily even true. You, there was that implosion of that big AI company about a year ago where it turned out much of what they were doing was actually like rule based or worse like you know, human level interaction to get the things done. So I think historically, you know, it's, it's actually been misused quite a bit. It maybe wasn't truly machine learning or AI and again a rule based approach to solving a problem. But again I think just, yeah, the technology has really gotten commoditized, easier, easier to use, more consumer friendly, you know, a lot of those pieces and, and now it's, it's pretty exciting. You know, people ask me this pretty right regularly. I think it's important for humans to embrace the technology and understand it and then at least come from an informed position or opinion.
- [04:18]
B
What other emergent properties that you expect that you're postulating with even more computing power?
- [04:24]
C
Yeah, just in general or relative to health or in one area.
- [04:27]
B
Oh, from here, from this point, let's say computing power 10X's. What do you expect that will be possible or emergent at that level?
- [04:35]
C
Yeah, I can tell you what we're seeing. So we have a member of the team, his name is Dr. Galen Buckwalter. He's a chief chief brain futurist and so he's really on the cutting edge and has been. I've known Galen for decades and he is in, you know, he's a, an AI scientist, behavioral scientist, and his research just spans decades. And so in Galen's case, he's also at quadriplegic and he has six neural implants called BCIs, or brain computer interfaces. And he's in the middle of a clinical trial at Caltech. And so we will go to Caltech. And I've gone with him a few times and you know, he's, he's sitting there and essentially he has these six chips that are actually on his neural tissue measuring down to a single neuron firing. Like never in human history have we had this depth of brain resolution. And essentially what's happening is these arrays are measuring his electroactivity of his brain and then it's using AI to then, you know, essentially allow him to control a robotic arm is one of the, you know, the immediate exercises, essentially. And so in that instance, you know, right now we're seeing the intersection of biology and machine. And it's our generation, literally right now. Galen is a pioneer there. And again, that's also, you know, both on the ability to measure the brain, to excite the brain, and then also to drive AI to, you know, allow you to control an artifact of the brain.
- [05:55]
B
But again, like, let's say I know you're not him, but just based on your experience, computing power like a hundred x, what do you think would be possible generically that is not possible now? What would you notice or what do you think could be done? A thousand X?
- [06:07]
C
I mean, a thousand x? Yeah. I mean, again, you know, computational probability just goes through the roof. You know, one of the most. I'm a biologist by training. Molecular biology and cell biology were research for a number of years. And so where we're seeing that applied are areas like, you know, protein folding. And so what could that do? Well, that advanced processing power of a thousand X could allow us to understand interactions on a, on a protein amino acid level that have never really been able to be modeled before. And so theoretically that could also help us in the drop drug development cycle, you know, to really understand if we have the right specificity of this molecule delivered, you know, in this very particular way, it could, you know, ultimately help curb a disease.
- [06:46]
B
Okay, so what is Brain dot one? You know, co founder and CEO of it. What's the premise of the company? What does it do?
- [06:52]
C
Yeah, so Brain one, we are, we are focused in the concept of, you know, the mission first is to reach a billion humans and help them optimize their brain and their biology. And the vision of how we get there is through the idea of a hyper personalized health protocol. And so what is a health protocol? So I come from the world of advanced or I'd say endurance sports and you know, things like triathlon and Ironmans. And so in those worlds you're following a protocol to achieve a goal which could be, you know, racing an Ironman as an example and you're following a plan. And so a protocol basically is just a breakdown of a plan. It includes microhabits and then it includes the non negotiables in health. So things like nutrition, exercise, sleep, stress connection, those are some of the major pillars and then it breaks it down into microhabits that you know, a human can follow to ultimately improve their health or well being. And that's the, that's been the, you know, really the, where we've come from, the genesis of the company we started off in Brand Brain. An example could be, you know, dementia is preventable if you make certain lifestyle modifications. And there's a paper written called the Lancet 2024. We will take that paper, we will ingest it into our AI, develop editorial, and then ultimately a protocol that ideally our grandparents could follow, our parents could follow and ideally again reduce the, the possibility of getting dementia through these modifiable lifestyle changes. That's the problem.
- [08:14]
B
Well, let's say you're running a marathon and you have this pump connected to you and it's looking at the levels of a hundred different biomarke occurs every second and it senses that you're low in X, Y and Z and too high in A, B and C. And it selectively releases cofactors or enzymes or whatever it may be to keep you in balance on a like literal second by second or minute by minute rate. I wonder how much you can improve someone's run or performance. You know, this isn't like real short term instead of long term. What do you think something like that could do?
- [08:41]
C
Oh, I think that's an awesome hypothesis, Richard. I'd love to test that. I mean we're, you know, we're doing that now to a degree. But of course it's the feedback loops, it's not in real time yet, you know, meaning, you know, when I started doing travel 20 years ago, you know, it is big bulky Garmin watches and they've only gotten better and then, you know, have additional biomarkers or biometrics, you know, out of the box basically. And, and so now as we're training and you know, doing these types of races, you're getting feedback. The piece that's missing though is like the, you know, the real time optimization, like to your point, oh, your carbohydrates are down or, oh, you're, you know, getting dehydrated or you need to improve your electrolytes so that level of feedback's not there yet. But, you know, I, I think it could be coming. Another example that we're seeing pretty prevalent are things like continuous glucose monitoring, you know, those devices and tell you in real time what your glucose is looking at. Like, that's a good example where you can immediately, you know, ultimately titrate up or down, you know, depending. But yeah, I mean, I think it's exciting. And what you're proposing, you know, we're not quite there yet, but you know, I think we're, we're moving in that direction.
- [09:45]
B
Yeah. And I guess for the long term, again, it would do the same thing at a much lower sampling rate. But let's say you have a biological pathway where you can't process says B vitamins unless they're in the methylated form. It would be tailored to you and it could give you recommendations on the fly. And with your unique biology, I guess it would craft an ongoing changing plan for you based on what it's seeing so that you could, you really optimize quickly.
- [10:10]
C
Yeah, so we're, we've built parts of that now, it's not yet in real time, but you know, the concept of adaptive AI, are you familiar with, with that term?
- [10:20]
B
I was going to ask you because that's the main thing that you, you have in your bio. So what does that mean for listeners?
- [10:24]
C
Yeah, so it's a, it's considered a, of artificial intelligence and it continually learned and ultimately evolves in real time. And you know, that's based on different data inputs. In our, our case, it could be biometric, it could be our biomarkers, and then it could also be clinical assessments. And, and what that allows us to start to understand again is the, it's the composition of the human. And then further, we can use that data to then adapt to their health protocol, you know, to them at that moment in time based on their data. And so an example could be you wake up in the morning and we're agnostic to the wearable. We integrate into over 300. But let's, you have an Oura ring and your aura ring shows, you know, ultimately, Richard, your HRV is down and you're like, okay, why Is my HRV down? Well, I had a glass of wine last night and, you know, that'll certainly do it. But then the adaptive part of this would be like, okay, your HRV is down 25%. You know, you didn't sleep as well as you normally do. Normally we would maybe recommend a microhabit, like cold plunging, but you know, maybe since your parasympathetic is a little bit more, more taxed, why don't we recommend something like breath work? And so that's what we've built again, an adaptive AI in the context of a health protocol that's an integral part of the, the brain. One solution. And I think there'll be, you know, a, a very near point in the future where again, people are utilizing, you know, these tools to, to understand their health in real time and then it's adapting to them based on their biology.
- [11:42]
B
Ultimately, no, that's really cool. I'm sure, again, the performance would be a lot better with, with that, I imagine. You know, I was imagining like two MMA guys. And you know, what if you had a series of, I don't know, sensors so that you can see what, when one guy is going to punch you and you can counteract it and, and you know, I guess you could probably do it on a microsecond or even faster level and you could block any punch or kick and just, it would just be really interesting to have something like that, I guess. There's, there's so many applications for this, you know.
- [12:08]
C
Yeah, a hundred percent. And again, you know, the one that's like, I would say out there in real time would be, you know, the glucose example where, you know, humans are managing their glucose in, in near real time. But yeah, this will absolutely expand to other areas of, you know, one, on the measurement side, different types of biomarker measurements. But then potentially, which you were alluding to earlier, you know, the optimization of the biolog, you know, based on whatever the, you know, upregulation, downregulation might be of that, you know, particular gene or molecule, you know, whatever it might be for that human at the time. So that's certainly the goal.
- [12:40]
B
It'll be interesting if you track, you know, again, a whole bunch of biomarkers, let's say, on a, you know, second by second basis, and you look at it over a month's time. Yeah, it would be interesting to see as these increase and these decrease, it would show patterns that you would never have seen before, you know. Oh, I didn't know that. These, these seven seems always respond in this way, when these three change and you know, what do they look diurnally or in response to stress or response to exercise and. Yeah, there's so much learning, I guess, hidden in there, we just can't see it.
- [13:09]
C
Oh, yeah, I mean, a hundred percent. Yeah. We look at studies in a couple different ways. You know, one is we can, you know, have a study where we will look at biometrics and then we'll give a human a protocol essentially of different types of lifestyle modifications and microhabits. You know, so again, things like circadian regulation and direct sun in the morning and hydration and, you know, cold plunging as an example. And then we have, you know, then we have a measurement of, you know, their, let's say, their, their core biometrics over a time series. But what's also interesting is that we can also capture what are the other microhabits they're doing as part of their day to day. And so it actually fills in this bit of a void, you know, relative to the other things that, you know, point could really impact their, you know, their biology.
- [13:49]
B
Yeah. What are, what are some examples of microhabits? Like how long is the habit? How many times?
- [13:54]
C
Yeah, sure. Yeah. So, you know, when people ask me like, okay, well, what is a protocol? What's a microhabit? So the way we think about the world is the microhabit is the smallest action that really has compound impact. You know, it might seem small. And so a couple examples would be if you're familiar with Andrew Huberman, the Stanford neurobiologist, you know, we've analyzed every protocol on longevity ever written or every protocol that's ever been published, you know, by Huberman as an example. And one of the number one things he evangelizes is the concept of circadian regulation and specifically getting direct sunlight in the morning. Are you familiar with that microhabit, Richard?
- [14:30]
B
I've heard it's good for you. But, you know, what is the, what do the studies show? What does it affect now?
- [14:34]
C
Yeah, I mean, so it's, what's interesting is again this concept of the circadian clock, which essentially is a biological clock that helps jurisdict when you should be going to bed. And what's a little bit ironic is that, you know, that clock, it also, it starts first thing in the morning, basically. So if you want to work on your sleep, you know, which is in whatever 12 to 14 hours from the time you wake up, it actually starts with that moment that you wake up. And so getting direct sunlight in the morning and it needs to Be, I mean ideally it's, it's direct, no sunglasses, no filters, you know, and so forth. But what that does is to your, essentially your nervous system, it helps regulate it such that you're again setting the, you know, your circadian regulation for what will happen that evening. And so that's an example of a microhabit. 15 minutes in the morning, as soon as you wake up, as soon as you have direct sunlight and just getting outside, you don't need to do anything. I mean you could go on a walk, but that would be one very specific microhabit. And then, and then you get into things like, you know, duration and frequency. Another one could be cold plunging. I'm a big fan, I live here in Colorado. We go in the rivers, you know, right now they're about negative 2 Celsius and we'll go cold plunge. And so in the winters, you know, what are the variables you're looking at? You're looking at the temperature of the water, which obviously you can't control, but you can control how long you're in the water for. And at that temperature I can, I can maybe do 90 seconds. You know, it's, it's literally below freezing, but it's moving. And then you also have things like frequency, so temperature, duration, frequency. And those are some of the variables that we tune within the microhabits. And the microhabits then form a protocol which is again is just a structure, just a list of, you know, tasks, things to do. And then the protocols form a program. And a program usually is a time bound, essentially objective, you know, you're trying to reach. Like it could be, could be mental health, it could be weight loss, it, you know, could be anything under the sun, better sleep. But that's how we generally think of the world with the, the microhabit is, you know, the nucleus.
- [16:28]
B
I think the big win will be when you do the microhabits for people. So like let's say you sell a sleep mask and you know the different frequencies of, you know, morning sunlight. And at a certain time it's programmed to emit that light directly onto your closed eyes while you're sleep, you know, in the last five minutes before you wake up or something and prepare you for the day. So you don't have to go outside, not out of laziness, but just, you don't have to do anything in order to accomplish that microhabit or you know, it is just the temperature of the room at certain times to optimize your sleep or. Yeah, you know, just does all Kinds of stuff like cools your bed down to 55. Well, city late, the cold plunging.
- [17:03]
C
Exactly. And they actually have that right now. Richard, like do you know that mattress sleep 8? Have you seen that?
- [17:08]
B
Well, I used to know the Uler and then there was a few other ones. I guess the newest one's the eight, right?
- [17:12]
C
Yeah, it means, you know, similar. It's circul cold water but to your point and it's, it's really about the feedback mechanism and then how real time is that feedback. So. And I don't. I want to sleep eight yet. I, I pretty much own every wearable you can imagine. But I haven't bought one of these yet. But I have a friend that has one and it's, it's honestly pretty incredible because my understanding is that it is self regulating so you know, if it feels temperature going up, I believe it will like cool the mattress or a certain version. Well and it's that again that idea of the, you know, the feedback loop. Another example actually is a, there's a device called the Apollo Neuro. Have you ever seen that thing?
- [17:45]
B
No, it's a.
- [17:46]
C
Essentially we at C. About the size of an Apple watch. And I know Catherine and Dr. Dave Rabin, he, they're the co founders, it came out of some of his research at University of Pittsburgh. But the concept is, it's a, it's a vibrational device and it absolutely, it supports parasympathetic nervous system and as an example what it will do is it will see if you are getting stress, it will send a signal and you have the device on and the device automatically plays a frequency and it's like just a vibrational frequency they call it, it's a vibrational song. Song. But they've actually run a number of clinical trials and they've seen really positive data around this. But that concept of vibration again helping you know, maybe you know, improve your heart rate variability or just your stress at that moment as an example. So. So moreover this stuff is becoming real, not quite down to like a biomarker or you know, delivery of a molecule maybe outside of glucose. But you know, we're starting to see more and more of this every day. I think it's pretty exciting.
- [18:42]
B
Yeah, no, that's really cool. So what's a way for people to keep tabs on all the innovation, you know, that you recommend? What's a good central repository or place?
- [18:50]
C
You know, that's a big question. I, I mean it's something we're building at Brand one. We really, we have A system for taking peer reviewed articles and then summarizing them. And that's something that we have built in the platform. You know, I'm a big fan of Huberman. I think he, he has done an incredible job for science and helping evangelize, you know, three hours of content. I think a lot of the, the tools that he's talking about, about, I think on the, the women's health side, there's a woman named Kayla Barnes. She's doing some really innovative stuff around women's longevity. Yeah, I mean, I think these are some of the people that are, you know, on the kind of the cutting edge of, of the technology and it's changing so quickly, you know, on a nearly day by day basis, but very excited.
- [19:28]
B
It would be interesting too if you took all the Huberman's podcasts, put them in, and then, you know, had it suggest which topics he hasn't covered or barely covered. You know, so you cover everything, you know, or paste that against what's trending, et cetera. And I'm sure that'd be useful to him.
- [19:41]
C
Yeah, I don me that's, you know, I mean, does he, is he following trends or leading them? I guess I would ask.
- [19:47]
B
I don't know.
- [19:47]
C
But you know, he definitely seems, yeah. Ahead of the trends on those things.
- [19:51]
B
You want to know about a certain condition and you could have, you know, I don't know, a hundred of the top podcasts in that subject reviewed and all their content assessed to see who speaks about what. And then a summary report prepared for you. That would be pretty cool too.
- [20:04]
C
Yeah, we could, we could build that, Richard.
- [20:07]
B
And it just takes a ton of data.
- [20:08]
C
Yeah, you know, the first question, it's like we have more data and computational power than in, you know, ever else in human history. So yeah, we could literally do that exercise, I think fairly, fairly easily. It's a good idea.
- [20:19]
B
Well, excellent. Well, Thor, thanks so much for coming. I appreciate it. I know you're doing quite a few podcasts, so I hope this one was a good one and, you know, thanks for your time.
- [20:27]
C
No, great, great questions, Richard. I really appreciate it.
- [20:29]
B
If you like this podcast, please click the link in the description to subscribe and review us on itunes.
- [20:37]
A
Thank you for listening to the Good Question podcast. Please email support the Good Question podcast if you have any referrals to great guests for us to interview, visit thegoodquestionpodcast.com to hear more interviews. And please help us spread the word by rating and reviewing us on Apple podcasts, iTunes, Spotify, YouTube or wherever you listen to this podcast?