Beyond LLMs: The Spatial Web, World Models, and Active Inference with Dan Mapes

What if the web didn’t just host pages—it hosted intelligence? We sit down with Dan Mapes, director of the Spatial Web Foundation and co-founder of Verses AI, to unpack how a shared protocol, digital twins, and Active Inference could turn today’s internet into a living, learning fabric that understands space, time, and context. Instead of one giant model, imagine thousands of expert AIs—cardiology, ports, power grids—each built by domain leaders, all interoperable and discoverable like websites. That’s not hype; it’s a shift from a 2D information web to a 3D web of intelligence.

We explore Karl Friston's Free Energy Principle and Active Inference with real-world stakes: energy efficiency closer to brains than data centers, autonomous decisions that don’t break when the world changes, and robotics that adapt without million-dollar pre-training. Dan explains how world models evolve through sensing and action, why protocols matter more than monoliths, and how decentralization enables hyper‑local solutions that still plug into a global network. From OODA to flow, meditation to geometry, we connect the mental models behind invention with the engineering that makes it durable.

The implications are huge: smart cities that coordinate in real time, supply chains that sense and respond, and a path from LLM content tools to embodied, autonomous systems. We also talk culture and economics—why abundance eases control, how institutions adapt like they did after the printing press, and why this next era could make us more human, not less. If you’re a founder, policymaker, or systems leader, this is a blueprint for building actionable intelligence—starting with small, truthful models that learn every day.

If this conversation sparks ideas, follow and share the show, leave a review, and tell us where you want intelligence to plug into your world next.

NWO Intro with Boyd

March 25, 2025

Flow Learning Lab

Find us on X. @NoWayOutcast
Substack: The Whirl of ReOrientation

Want to develop your organization’s capacity for free and independent action (Organic Success)? Learn more and follow us at:
https://www.aglx.com/
https://www.youtube.com/@AGLXConsulting
https://www.linkedin.com/company/aglx-consulting-llc/
https://www.linkedin.com/in/briandrivera
https://www.linkedin.com/in/markjmcgrath1
https://www.linkedin.com/in/stevemccrone

Stay in the Loop. Don't have time to listen to the podcast? Want to make some snowmobiles? Subscribe to our weekly newsletter to receive deeper insights on current and past episodes.
Recent podcasts where you’ll also find Mark and Ponch:

The No Bell Podcast Episode 24
...

Chapters

• 0:00 Framing The Vision: Spatial Web + Active Inference
• 4:31 Host’s Journey: Psychedelics, FEP, And OODA
• 9:31 Enter Dan Mapes: Foundations And Intent
• 11:51 From Neural Nets To Active Inference
• 20:26 World Models: Specific, Evolving, And Embodied
• 24:16 Protocols Over Monoliths: A Web Of AIs
• 29:06 Standards And The 3D Web: IEEE And Digital Twins
• 32:16 Performance Proofs: Robotics Without Pre‑Training
• 35:16 Flow, Meditation, And Creative Insight
• 38:16 Humans, Abundance, And The Intelligence Age
• 44:16 Energy Reality: Data Centers, Graphics, And AI
• 48:06 Autonomy At Scale: Beyond Human‑In‑The‑Loop
• 51:16 Building Organizational World Models
• 57:06 Decentralization: Hyper‑Local AIs And Control
• 1:02:16 Democracy, Institutions, And Power Shifts
• 1:05:46 Consciousness vs Intelligence: AGI Timelines

Transcript

Dan Mapes: 0:56

The spatial web combined with active inference AI, now you have an intelligent web that covers the entire surface of the earth. And that's going to allow us to do things, uh, help manage our climate and deal with supply chains and have a web of smart cities all interacting and talking to each other with ports and airports. We can now get to things that we've been dreaming about in science fiction. And now with the spatial web and active inference AI, uh, developers all over the world can build it. We don't have to build it. This is more like Apple's App Store. We're we're people can just take our SDK and they can build a cardiology AI in Australia or something. You see what I mean? So we're not like OpenAI. We're not building this giant thing. We're building AI tools to enable millions of people all over the world, just like websites and apps, to build uh whatever AI they want to build. And then the protocol allows them all to communicate with each other. And in that way, they kind of become a unified field of noise. Should I call you Punch? No, you can call me Punch. That's fine. Uh so Dan. Whatever you want, whatever you want. Punch, Punch is fine. In fact, I'll try to get you one of these I'll try to get you one of these t-shirts uh in the next few days, too. Awesome, awesome. Yeah, that'd be great. In the age of AI, uh, there is no way out. So uh bottom line here is Dan Mapes is in the studio with us today. He is the director of the Spatial Web Foundation. He's also the founder and president of Versus AI. For those of you who are not familiar with Versus, uh, that may or may not be fine depending on where you are as an investor or as a uh AI enthusiast or a leader in an organization, but that's why we're here today. So, Dan, welcome to No Way Out. Good to see you. Thanks, Flan. Great to see you. Hey, so I I'm thinking I'm gonna do something a little different with you today. And what I'd like to do is give you a little background on how we came to have this conversation today. I want to go back about four or five years, share with you some insights. And if I share anything that I shouldn't share, uh we could always edit edit that out later. Uh, you'll you'll need to tell me. Um, but I want to give you some background, and I think it's gonna start a might be a good foundation for our conversation today. So it's gonna be pretty unique. So hold on, everybody. Going back about four or five years, you know, I'm I'm a product of a culture war known as the war on drugs. So going back to the 70s, uh, we found out that uh, hey, psychedelics are bad, stay away from them. Um I went into naval aviation, I crawled into a bottle at the age of about 2020, 22, 23. Um, we drank a lot of alcohol, uh, we've had a lot of problems. We have a veteran community that has a mental health crisis. We're losing 22 to 44 veterans a day to suicide. It's unacceptable. So I want to take us to about 2019, 2020, uh maybe 2021. And I came across this book called How to Change Your Mind by Michael Pollen. And you may be asking, how does this connect to artificial intelligence? We'll get to that. I promise you that. It'll be pretty fun. So I read this book. I come across this theory by the gent by a gentleman whose name is Robin Carhartt Harris. And he talks about the snow globe that you shake up the snow globe and you get higher entropy, you get more disorder, and you get these counterfactuals, and it's pretty amazing. And he called it the entropic brain hypothesis. And he also has another one called Reduced Beliefs Under Psychedelics. Okay. So what does that have to do with anything? Well, if you dive a little bit deeper and you look at who he worked with on that and the theories he borrowed from that and built on, he built on a theory or that's excuse me, a principle called the free energy principle, right? An active inference. And that's by Carl Friston. So uh, long story longer, um, I hold space for one of my friends down in Mexico. I get to see the power of Ibogaine and Five MEO DMT, and in just a few weeks' time, we're gonna see on Netflix a movie called In Waves and War that features a lot of the folks that that are working on psychedelic assistant therapies for veterans down in Mexico. Anyway, um, I see the power of that. I see that, hey, this culture war that we had in the 70s may have been wrong, that we need to embrace this new way of thinking, potentially a new way of thinking where psychedelics are useful uh in some contexts and maybe not so much in recreational use, but that's, you know, I'm not here to tell you to do them or not to do them. I'm just saying there's value in them. Uh anyway, I see the power of Vibe Again, 5MER DMT. I dive deeper into um the entropic brain hypothesis. I start looking for organizations that understand active inference, the free energy principle, and I can find very, very few. This is about 2020. And I keep looking, I'm like, wait a minute, this lines up quite well with what we know as Observe Orient Decide Act, the real OODA loop, which is built off of physics, neuroscience, biology, and of course uh cybernetics. And it there's a lot of connections between what John Boyd was looking at, what we talk about on the show, and what Carl Friston mathematically came up with in his free energy principle. So I keep looking for organizations, and I come across two. Now, this is gonna be kind of an interesting point. The first one's on in is in financial markets. It's a company by the name of Hedgeye. Hedgeye is using John Boyd's Observer Decide Act Loop, and they're using Mandelbratz geometry. So geometry has a big part in a lot of things that we're gonna talk about here shortly, I'm sure. But I dive a little bit deeper in that. I go up to Connecticut, I get to hang out with these guys. Uh, you know, I'm like, this is these are my type of people talking about complex adaptive systems, neuroscience coaching, new processes, implicit guys to control, and I get invited up to go do a TV show with them, right? What do I talk about? I talk about the free energy principle and active inference. Well, here's the secret thing that most people may not know. One of the founders, and this may or may not be true anymore, I'm not sure where where this is at the moment, became the chairman of the board for versus AI. Michael Bloom. Yeah, yeah. Yeah. So just out of coincidence. So the second thing I was looking for it is a small world. And the second thing I was looking for is who in the world is doing this FEP active inference thing? Uh who's looking at it with regard to leadership, uh technology, whatever it may be. And I come came across your company called Versus AI and this thing called the spatial web, right? And I'm like, what is this? So uh I bought the book, The Spatial Web. I believe it's about what, six, seven years old now? Is that right? Yeah.

Brian "Ponch' Rivera: 7:19

Um you go there.

Dan Mapes: 7:20

It held up well. It held up well.

Brian "Ponch' Rivera: 7:22

Yeah, yeah. And I started reading it and it starts talking about geometry, right? Spatial relationships. And I'm like, wait a minute, three dimensions. Where in the world have I seen three dimensions before? Dog fighting, air combat maneuvering. In fact, you go back and you look at the origins of John Boyd's work in the Oodaloop, it started with spatial relationships, understanding that to it's not just three dimensions, it's four dimensions, right? But it's it's time. So we start getting into geometry. Uh, I start looking around some more. I find these folks that talk about sacred geometry and, of course, the connection with the psychedelic assistant therapy on what's going on there. There's some connections to the brain, there's ideas that as above, so below, all these things. And I'm like, wait a minute, this is all about flow, flow states, right? Peak performance and things like that. So I came across your work, what you're doing at the Versa's, and I knew I had to have you on the show. It took some time. Accidentally, our our ninth guest on the show was uh Dr. Hippolyto. She came on to talk about the free energy principle and active inference. We've had several people from neuroscience with neuroscience backgrounds come on to talk about that in their context for markets, one of them being Patrick Scotanis, who worked with Carl Friston as well. I've done a little bit of work with uh Stephen Cotler, who's done work with Carl Friston on flow, intuition, type zero, and type one cognition. So that's a lot of me spewing information out to you and the world. So I want to I want to have you take that insights or those insights and kind of take us back to where the spatial web came from, what it is, and of course, what you're doing with Carl Friston and active inference.

Dan Mapes: 8:54

Wow, that's a lot. Let's see, where to begin here. I mean, obviously, uh let's let's take a minute and just honor uh Dr. Carl Friston. I mean, uh, he's going to be a go-down in history as one of the great minds of uh the 21st century. And Carl um is a professor at University College London. And a lot of people don't realize uh that that's where uh Deep Mind came from, that was also at University College London, uh, because uh UCL is a big neuroscience center. So uh Deep Mind uh spun out from uh UCL, uh a bunch of PhDs at uh UCL that were working in uh neural nets, spun out and launched uh a company called Deep Mind back in, I don't know, 2012, somewhere in there. And then Google, I think, bought them in 2014 for uh half a billion dollars. And so that's where Google DeepMind came from. Uh so that was from the computer science side of the house. But the neuroscience side of the house is actually looking at uh wetware, actually looking at the brain itself, not just computer chips. And so neural nets are uh pattern recognition machines. They're machines, they're not intelligent. They're a machine that uh mimics intelligence, but and it's a wonderful labor-saving device. Uh a document that might take me 20 hours to write using GPT 4 or 5. I can write it in five minutes. I may want to edit it for a couple of hours because it will hallucinate and uh have errors, but in general, it's a huge, huge time-saving device. But Dr. Friston was down the hall and he's kind of uh developing uh his understanding of how the uh how the mind is developing, and he's noticing a couple of things. One, let's use 100,000 watts to answer a simple question, whereas the human brain uses 20 watts. Wow, that's something different. Secondly, how is it that a child can go from the ages of one to five to ten to fifteen, evolving its intelligence without ever having to go into the hospital and get a new chip upgrade? So there's no GPT-2, GPT-3, GPT grow every minute, every day. So um, so uh, as he attacked that problem uh back in the early 2000s, he uh discovered the free energy principle. And um the free energy principle is uh you could argue that um there there's two terms that you'll hear a lot when you're looking at advanced artificial intelligence. And really you can call it the next generation of AI after neural nets. So everything OpenAI, Grok, all these people are running on 40-year-old technology that uh Dr. Jeffrey Hinton uh really uh helped develop in the um in the 80s and 90s, along with Jan Lakun. Hinton then became head of Google Brain, and Jan Lakon became head of AI at uh Meta. And so they were looking at how to take these uh pattern recognition machines and get them more and more accurate. And they realized by making them bigger with more chips and more data, then they get more accurate. And maybe you can just do that till the end of time, and then you'll have AGI. You'll have a smarter-than-a-human machine. But Dr. Carl Friston was going like, hey, that doesn't seem like the right approach. Why don't we why don't we learn from nature uh how intelligence is evolving through species and within a person? And so you could argue that intelligence has evolved even in the in the human tradition from Stone Age to now, uh, but uh certainly uh through the whole hominid line and to uh uh Neanderthals to us and that kind of thing. But then uh uh but then even myself, uh being a young boy uh growing up and now uh you know going to college and into adulthood and then uh being an inventor and traveling the world and things. I mean, how is it that our intelligence is growing larger over time? So I still knew, I still know the things I knew when I was five, but now I know things that I didn't know when I was five. And even some of the things I thought I knew when I was five were incorrect. I've now adapted them. So, secondly, not only does the brain use a lot less energy, but it self-evolves. You don't need to reprogram it, you don't need to put training data into it. It's using in senses, it's called embodied intelligence. So it's not just a uh a machine in a room that we're loading with information and then asking the big machine to answer a question. It's more like an organism living in the world and looking at the world in real time. And so if I ask uh ChatGPT, uh, where's the teacup in this room? No idea what I'm talking about. Right. It might go, well, teacups are generally uh 87% in the kitchen, you know. But actually the teacup's sitting right in front of me, and I just opened my eyes and look. Suddenly it takes me 20 watts to answer the question, I just look around the room with my eyes and no, there's the there's the teacup right there, you know. Uh T has to look through all of this data about houses and teacups and everything to answer a question like that. So Fritzson immediately attacked the problem, and what he discovered was is that an organism living in the world has to survive. And if it makes mistakes, it will probably be eaten. An antelope can't just walk over to the pond in Africa and have a drink of water very casually and walk away. It's gone in a second. So uh antelopes have learned over time that uh if they go very early in the morning when the lions are sleeping, they can probably get a nice long drink that'll hold them. And then if they get really desperate later in the day, they'll still go to get a drink, but they're very careful. They're looking right around all the time, looking for any danger, and then they just dip their head, drink, look, and then off they go. And so they're learning techniques to survive. And uh, so they're building an understanding of the world. The lions come in the middle of the day, and the analysts can come early in the morning, and we won't be bothered. And so that is called a world model. You're building an understanding of the world around you. And a one-year-old kind of models the house, you know, two-year-old goes to the yard. Uh, once they get their bike at six or whatever, they can ride around the neighborhood. They're modeling, their world model is growing. And the model isn't only geospatial, it's also psychological. If I do this, it makes mom mad, or things like that. And so uh so they're we're we're modeling the world constantly every day. Even this conversation will cause your world model to shift a little bit and my world model to shift a little bit, and people listening to it will do it. And so that's how we're evolving our consciousness, our understanding of the world. So I want to make it how would we do how would we do that mathematically? Well, let me let me just finish the thing there. So then his question was well, then how would we do that mathematically? And so he came up with two things. One's called the free energy principle, and the other is called active inference intelligence. And so you could say that the free energy principle is the why. Why do we, how does this work? What what's going on? And active inference is the how. How do we restructure our thinking and our world model so that it grows? Well, if you understand the why and the how, now you can mathematically model that. And once you can mathematically model that, then you can turn it into code, and that's what we've just done. We just spent 140 million dollars. And Carl's chief scientist and all of his uh top PhD students are the AI department, and we've just spent the last uh five, six years uh developing this thing and evolving it, and uh now it's done. We just finished it in uh in uh May of this year, and uh we're slowly uh rolling it out, and uh, so it's been an exciting journey. Next generation of AI, uh big deal, and it does everything that the older generation could do with one one-hundredth of the energy, hundred times the accuracy, and then it can do all kinds of things the old generation can't do, like real-time data and things like that. So it's a nice moment in history, and uh, we owe a great debt of gratitude to Carl for uh coming up with the free energy principle and active inference, and then what a great team we have at Versus that's been able to take that and turn it into a commercial product.

Brian "Ponch' Rivera: 17:58

Well, I have so many questions. Uh I want to start with a world model, and and uh I'll I'll give you my point of view. A world model is not does not mean I understand everything in the world, right? And I think some people take it as that. It's it's just a uh I think an agent can have a world model and just understand this room that I'm standing in at the moment. Uh is that true? Is that how it kind of works?

Dan Mapes: 18:16

So a world model is any understanding you have. Uh a two-year-old has this big of a world model, and an eight-year-old has this big of a world model. I mean, uh, so yeah, you can have and the and and as you said, you can have very specific world models, like if you're a cardiologist, you might have a very specific knowledge of cardiology that I don't have. You know, I have a very, maybe very lightweight thing about cardiology and heart health, but a cardiologist is they can they could build a complete server for that. So you're right. So uh world models, so there's a general world model that we have of understanding of the world with gravity and light and the sun comes up in the east and things like that. But uh but that world model is evolving constantly, and science is causing it to change, our experiences are causing it to change, travel is causing it to change, conversations like this are causing it to change. So that that's a way to think about it. But you're I think you put your finger on something very specific, and that is OpenAI and Chat GPT and uh and uh Grok and uh Meta and all these others that are building these uh uh that I mean these yellow limbs are building a giant data set of like every word that's ever been put on the internet and magazines and other kind of thing. And then you get to query it, but it doesn't actually have a world model. It just has words about the world in there. So we don't need to have, as you point out, we don't need to have a complete model of the world ourselves. Science is building models of the world, we're learning from it as we need, but there's specific models of the world that are being used for different things in medicine, space exploration, manufacturing, whatever it is, pharmaceuticals. And so these are very specific world models. And so that's another thing that Carl noticed is that why build a giant thing that requires billions of watts to run? Why not build a number of specific things and then have them very specialized and curated around cardiology, around space exploration, around education, whatever subject matter, and then let them talk to each other. And so that's what the spatial web is. So what we've learned on the internet is I if we make a protocol, uh, the first protocol was TCP IP that enabled email, it allowed every computer in the world to talk to every other computer in the world by just having an IP address and um and really got cool new approaches to sending messages across the network. And then Tim came along with HTTP and HTML and said, hey, let's do fully formatted pages with HTML so we can locate things on the page and HTTP to link them all. And then uh the spatial web says, hey, why don't we so now we have a web of communication first, then a web of information. Why don't we have a web of intelligence? And so that's really what the spatial web is. It brings in a web of intelligence with instead of 2D pages, three-dimensional geospatial environments. So that the spatial web combined with active inference AI, now you have an intelligent web that covers the entire surface of the earth, you know, and that's gonna allow us to do things, help manage our climate and deal with supply chains and have a web of smart cities all interacting and talking to each other with ports and airports. So that's kind of the exciting moment here. We can now get to things that we've been dreaming about in science fiction, and now with the spatial web and active inference AI, developers all over the world can build it. We don't have to build it. This is more like Apple's App Store. We're we're delivering a platform. People can just take our SDK and they can build a cardiology AI in Australia or something. You see what I mean? So it's not, we're not like OpenAI, we're not building this giant thing. We're building AI tools to enable millions of people all over the world, just like websites and apps, to build uh whatever AI they want to build. And then the protocol allows them all to communicate with each other, and in that way they kind of become a unified field of knowledge. So that's kind of a quick way of looking at the overall thing and why this is such an important moment.

Brian "Ponch' Rivera: 22:19

So let me try this on you. Uh in complex adaptive systems thinking, we we learn that it's the interactions between the agents that matter. The quality of the agents is matters less than the protocols, the interactions. In fighter aviation and commercial aviation, we learn that protocols, what we call crew resource management, teaming skills matter more than the quality of the agent. Same thing is true in sport. You can have um an average group of players beat phenomenal players if they're interacting at a higher level, if they have more flow, if you will. So what we're seeing is that the protocol that you're talking about is kind of like connecting these, I'll call them agents with oodle loops for now, but they're they're agents with a world model, which we call orientation. They have sensory states that access the external world, and they have active states where they can act on that world, which is the same as the oodaloop, where you have observation and you have action. So those represent the boundary states, and you put a boundary around them. So so what I think the way I look at this is you're copying nature. You're just saying, hey, if we do what nature does ecosystems. Yeah. And we may be wrong with the, you know, FEP may change over the next 20, 30 years, but what we know today is the best we know right now, right? And that's all we're talking about.

Dan Mapes: 23:32

So even if the even if FEP doesn't change, a child doesn't have to change its FEP. Just by us taking in more data, it just becomes more aware and uh, you know, change turns into an adult. So uh FEP doesn't have to change. Equals MC squared doesn't have to change, but you know, that this is a principle in nature. Uh so um so that's the great thing. I mean, think about it from an inventor's point of view. There's two ways to invent. One is to hack something together. Hey, uh, here's a wheel, here's a wheel, let's make a bicycle. Yep. There's another way to invent, which is uh let science do its magic and then use that result of that science to invent. So uh the Bernoulli principle is a good example. Uh Bernoulli principle, we used to, you would you ask any normal person, they think an airplane flies through the air because the jet engines push the wings this way and it surfs the air. No, it's because you know it's the shape of the wing and creates a negative lift. But we didn't know the Bernoulli principle for a long time, and then we did. And then now you can apply, once you know the math of the Bernoulli principle and you know the weight of the plane, the actual wings and everything are almost designed by the Bernoulli principle. This is how big the wing has to be. This is how much thrust you have to have to lift that off the ground. So if you have a science foundation and then on top of it, then you build your engineering applications, that's a really good solid approach. And that's what we've done here. Carl uh represents uh kind of the peak of European uh scientific thought. And uh, it's odd too that so much of AI comes out of England. I mean, Alan Turing was in England at Bleshley Park when they uh built the uh computer to crack the uh German cipher. And he he's the first, really the first person to kind of coin uh artificial intelligence. And then uh Tim Berners-Lee, the English guy, you know, developed the World Wide Web and Carl now uh with active inference. And so uh, of course, uh uh even DeepMind is based in England. And so we really we really have a great uh tradition in in England around uh advanced uh computing uh uh architectures. And uh and then we put that together with uh Silicon Valley uh entrepreneurship and engineering. So it's a nice combination of uh a good scientific foundation with good engineering on top of it, really what's allowed versus to uh have the breakthroughs that it's having. And and if you go to our website, you'll see the breakthroughs that we're beating everybody at the Atari games. The latest one was on uh robotics. They have a test for robots called Habitat. And uh probably the best score on Habitat might have been, I don't know, 50 something prior to us. But the way that neural nets work is you have to pre-train them. You have to give them millions and millions of instances of possible things they might encounter. If they then encounter something they don't have in their training data, they're kind of flummoxed. They don't have a reasoning ability to adapt. So all the different uh robotic systems up until now have been based on uh neural nets. And so uh and they they don't do very well, but they do better and better over time. But uh, we just took the test uh here a couple of months ago, and we just wanted to make a point. So we didn't do any pre-training at all. No, no pre-training. So very costly to pre-train, millions of dollars to build all these millions of instances and train the neural nets for the robot and all that. No, no, no, no pre-training. Just pure active inference. Robot comes in the situation, solves the thing, scores something like a 75, and beats everybody with no cost, no pre-training, nothing. So that's so radical this new technology is. It really is a tiny fraction of the cost and of multiple capability, and then the ability to do things that we've never been able to do before because it just unlocks new possibilities.

Brian "Ponch' Rivera: 27:16

So you brought up uh a couple places where your creativity, creativity or novelty emerges from. One is is uh your analogy of breaking things apart, like snowmobiles, you know, putting parts together of different things and science. Let me ask you this is there a third one, potentially access to maybe uh a grand consciousness or the Akashic Records or Akashic Field, or or is there another way to get and this this might be a leading question because I'm really curious. How did you come up with this idea of the spatial web? What were you what was going through your mind you know, 15, 20 years ago?

Dan Mapes: 27:46

Honestly, honestly, I thought I thought I thought a thousand people would attack it as so obvious. It's so obvious. So here we are in 1995, and the World Wide Web is blowing up. Yeah. And so it's a 2D web. It's a web of web pages. That's why it's Facebook and PageRank, it's the pages, right? HTML, uh markup language for pages. This is bold, this is italic, put the picture here. It's a markup language for pages. But at the same time, I'm a computer game designer at that time. I'm working on a Tony Hawk Pro Skater and other big games, and they're all in 3D. And so we've got now 25 years of 3D computer gaming on Xboxes and PlayStations, and 25 years of a 2D web because the web can't handle 3D, it's too heavy. But it's pretty obvious to a technologist that bandwidth is getting bigger, the chips are getting faster, a 3D web. Well, it was so obvious, even in 1995. Oh, we only have a 2D web because we can't build a 3D web. So you could argue that from 1970, when the internet just first began, we've always wanted the spatial web. We just couldn't get there because the ships couldn't handle it. And so the only thing we could do in 1970 was unformatted little text files to each other called email. And then as soon as we had more bandwidth, Tim said, hey, now we can download pages. But if you were around in 1995, I mean you could actually watch a page download with really slow internet. Now, of course, we're watching Netflix and playing Fortnite and everything online. And so we had the bandwidth for 3D. So, Billy, the 3D web is just the what we always wanted from the very beginning. Now we have the web of the internet of everything, because now every building can be online. So that gives you smart cities, supply chains, forests, monitoring the coral reefs, I don't know, ocean temperature, well, but what's called a digital twin. Right now, everybody's working to build a digital twin of the entire planet. And once you've got that digital twin of the planet with satellites updating it in real time, now you can put active inference in there and it can help manage things. So that let's just say that that's been the dream of uh humanity for a very long time. And now, because of the changes in chips and internet speeds and things like that, it's realizable. You know, it was so to me, it was so obvious we were going to need a 3D web. And my partner Gabriel uh Renee and I have been uh working on projects together for a long time. So we just looked at each other and went, well, you know, here we are in 2018, seems like a good time to start. It won't really have the bandwidth until the early 2020s, but you you don't wait. It's like a surfer, you know, you don't wait till the wave is here to get paddling. You start paddling and catch the wave, and so we could see the wave coming. We got, oh, early 20s, we're gonna have enough bandwidth, so let's start coding now. And uh so we started the uh Spatial Web project in 2018, and now it's just been ratified by the IEEE as a new global standard in um in May, just uh four months ago.

Brian "Ponch' Rivera: 30:53

So speaking about or talking about uh surfboards, you you're a fan of peak performance, and I'm pretty sure you're a fan of flow and me, I chicksand me I's work and understanding flow states and access to novelty. And I brought up uh Stephen Cotler's work with uh Carl Friston on intuition. So I gotta I want to I want to go a little bit deeper with you if you don't mind. Um so you're and you talk about consciousness, you talked about that many years ago. There's a there's a um a YouTube video out there of you talking about it and how things are connected and reality and how the sidewalk is may or may not be real, and you know, once you once you start starting down to the atomic level and things like that. So I know you I know you know flow. The the answer you gave me was conscious was so obvious. No, I think you were surfing, I think you were in a flow state, I think you had something going on when you came up with this, right? Do you remember where that was or or uh the conversations that started or what you

Dan Mapes: 31:50

I mean, I I meditate. I mean, there's no doubt. Uh things like that really help. Uh keep your mind clear because normally we get uh too involved in uh the day-to-day uh dramas of the default reality. And things like meditation and uh other things like that to help you uh step back from it all and uh just kind of get a broader view of life, of time, of reality. And I think that probably does help in the creative process, no doubt about it. But I can't imagine uh a life where without meditation was such a saving grace because you know, daily life is so hypnotic, all the dramas, all the problems, and you can just get caught in them, and next thing, you know, you're just like inside the video game instead of playing it, you know. Certainly uh meditation practices and I love Advaita Vedanta and Buddhism and all the great uh traditions that uh help us step free from that. Rumi's great work in the Sufi uh tradition and these kinds of things. So I think those are those are very, very helpful. They're they're kind of uh the right brain to our left brain. The left brain, you know, is all the university type education stuff and daily problem solving, and then the right brain would be more like uh meditation and and uh poetry and music and things like that. So I think having a balance between those things is really great. I mean, I just kind of intuitively went for that. We have a right and left hand, why not uh use them both?

Brian "Ponch' Rivera: 33:10

So humanness, getting back to being humans, I think the direction that our current AI is taking us is removing some of our capability. I believe that AI that you're talking about could actually bring us back to becoming more human and allowing us to do things that we normally wouldn't be able to do without technology. So I'll give you an example. In sports, I see great application for this for improving individual performance, uh creating agents that look at individuals, understanding their genetics, their past, uh, nutrition, how much they sweat when they work out, you know, during a game or whatever it may be, the way they shoot a basketball. That's gonna change the nature of the game here soon. That's how I'm kind of looking at at uh your AI, active inference AI, anyway. But that's just one piece of it. The other piece is in organizations, I just read that maybe LLMs aren't having the impact that many people thought they had. There's something known as work slop. There's, you know, people are creating work for others by by using LLMs. That's getting away from being human. You know, that's just kind of outsourcing your work to a machine to have somebody else do the work and have your machine talk to my machine and we'll meet up for lunch sometime virtually, right? We we need to get back to being social creatures. And and I think what I what I like about FEP and Active Inference and what you're doing is I believe it's gonna take us back to being more human, our our humanness. Any thoughts on that?

Dan Mapes: 34:25

So it's a really big idea, Punch. I mean, you know, no doubt about it. Uh it will change. Uh if we look back at human history, uh, the latest findings are that uh humans like us have been on the planet about 200,000 years. And so 90% of that time, up until 10,000 years ago, we were hunter-gatherers. I mean, our our primary tool was a spear, and uh, you know, and um the you know, we're very simple lives basically chasing uh protein around and uh and then uh picking up uh finding and picking uh unpoisonous berries and other things to eat. And so that was really our life. And that's uh the deep background of human beings living in small tribes of maybe 30, 50, 80 people. You couldn't have a very big tribe because you'd eat all the food. And then you kind of had to follow the animals too as well, so you were pretty nomadic usually. And so um, and then we uh discovered uh agriculture uh really after the last ice age uh around 10, 12,000 years ago. Um and suddenly instead of having food scarcity, we had uh food security, or at least uh vastly in ways we never could have before. And that allowed the growth of people can stay in one place. Suddenly they needed uh mathematics to count how many jars of olive oil you had and how many bushes of wheat I had. Next thing you knew, you got civilization really popping and uh huge empires growing, the Persian Empire, Greek, Roman, all these things. And uh, all just on the back of that one breakthrough into understanding that we could grow our food and we could uh breed animals, and suddenly we would have an abundance of them. And then um the next big breakthrough was uh 500 years ago with the uh dawn of the industrial age or the machine age. I mean, Gutenberg invented the printing press and used to take a scribe three years to handwrite a Bible. And so they would be commissioned. That's $150,000 in minimum wage in today's money. I mean, they would be commissioned by a royal family or something, and not many people read, and you know, life was pretty simple. You were generally just apprenticed uh for your work and worked on a farm or you were apprenticed to the blacksmith or whatever. And uh suddenly uh the printing press came out in 1457. There were hardly any books in circulation in Europe at the time, and by 1500, less than 50 years, there were 20 million books in circulation in Europe. And then the 1500s was the Renaissance, just an explosion of ideas and philosophy and art and science and new governance models and the Reformation. I mean, it's just like huge change in the next hundred or so years. Great age of discovery, people will start sailing ships all around the world, and we live in the world where we live in that world now. That's the industrial age, we call it, or the machine age, or whatever you like. But now that's ending right now. It's ending right now, and everybody can feel it. It's like, oh my God, what are we stubbing into? There's a whole new thing coming in here with computers and networks and and virtual worlds and artificial intelligence, and well, well, it's as radical as the machine age was to the ag age. Now this new intelligence age is that radical to the industrial age. And so um, I think everybody's feeling that, and you know, it's exciting and terrifying because we're creatures of habit. For 250,000 years, we just hunted animals every day and ate food and survived. And that we like, we like things to be predictable. And when we go through these big renaissance moments, and we're definitely in a renaissance moment right now, the pace of change is so fast that it kind of overwhelms us a little bit with change. But it's like what happened after the last Renaissance. Oh my God. A world opened up. You can bring anybody here from 1400, and they're going like, you guys are living in science fiction, airplanes and electric lights, so what is all this stuff? And so we're about to, we're those people, we're like the people in 1400. If we could go out to 2100 and see what's a what we're gonna create, we'd be going like, oh my God, our our best science fiction wasn't as good as this. Probably lifespans will go up to what 150. I mean, uh uh mass abundance, uh, super wealth, uh disease pretty much wiped out, uh, probably interplanetary at that point. Uh-huh. So, so you know, I think uh we're on the verge of something really, really extraordinary. And uh, we don't really have to overcome the limitations of the industrial age. It just it just gets solved by this explosion of all this intelligence, and there'll be a billion robots running around on the planet here in five or ten years. Wow. Robots right now. I'm amazed at the price of robots. Humanoid robots right now, you can buy them for under $10,000. That's today. What's it gonna be like in 2030? I mean, billions of robots, and they're gonna just be able to every house will probably have a robot just like they have a car now, you know.

Brian "Ponch' Rivera: 39:23

Just like the Jetsons when we were growing up, right? Rosie.

Dan Mapes: 39:26

Just like the Jetsons. We're heading, we're heading into the Jetson world, you know, and and it's gonna cause us to change pretty much everything, just like the last Renaissance. Government will change, education will change, medicine will change, social relations will change. Maybe we won't have to live in big cities so much concentrated because you want to be near your job, you don't want to have to commute for an hour. And if there are no jobs, or if you're remote working, you can live wherever you like. And so we might get more of a web of communities rather than these giant cities that uh that we're living in now. So if you go out to 2100 and look back, uh you'll be able to see the change really clearly. Right now, we can we kind of can intuit it, but we can't see it really clearly. But clearly, massive, massive change is coming. And the great thing about human beings is they're the most adaptable animal ever created on the planet. That's why there's 8 billion of us. So we'll adapt and we'll learn and we'll understand how to use various new uh technologies and tools to take care of everybody. And because of mass abundance, probably there'll be so much uh abundance around that uh no one will get left behind in this next round.

Brian "Ponch' Rivera: 40:35

So let's talk about current assumptions in AI, namely massive data centers and high energy requirements. And we've already talked about if you look at the brain being the brain is an expensive organ, it just reduces its energy spend by creating predictions, right? So that's what makes that's what makes your technology powerful is you follow the ecological approach, the natural intelligence approach, and you see that uh if we create predictions, we can actually, in these world models, we can uh reduce the energy required for an agent or a system. Now, current assumptions are we need tons of nuclear power, we need lots more energy, we need all these things, we need big data centers. Does your technology and your approach crush those assumptions?

Dan Mapes: 41:13

No, uh it doesn't. It does it, it does, it does, and it doesn't. There's a yes and a no to that. Uh so if we look back uh at uh what happened in the late 90s, uh we had this same exact thing we're happening right now with AI. We had the dot-com explosion. People were just throwing money at any website, hey, test.com, weddingplanner.com, you get a million, you get a million. Yeah. Not you get a billion, you get a billion. And then the it reached a bubble and then it like deflated. In the process, they were throwing in fiber optic cables and all kinds of things. So when the thing crashed, the best survived Amazon and Google and great companies that uh around today. But what's great is the infrastructure also was needed for these other things going forward. So all the fiber optic got laid and everything. We used it. I think the same thing's gonna happen. We're gonna have a huge uh LLM crash. There's only gonna be three or four LLM companies uh here in uh in the next uh three years or four years, probably, because you know, OpenAI's got $100 billion to play with. So they can survive the winter, and uh, many of the smaller companies will get weeded out. People like uh Meta will survive because they have money from Facebook, and Google will survive because they have money from Google, and the Chinese, Deep Seek, and people like that will survive. But right now we have over a thousand AI companies, and probably 95% of them will all be gone in three or four years. But uh because they're building these giant data centers, um, they won't be used for AI. You don't need them if you're using active inference because you have millions of little uh data models all over the world, and they're all just communicating with each other like the World Wide Web. So it's just an extension of the web. It can run on the cloud, you can run, you can run an AI on a smartphone with active inference. So it won't be the AI that's using those data centers. It'll be computer graphics. Now you've got to remember NVIDIA started as a game chip. You really bought NVIDIA boards to make your games run faster. Well, what's the what's the spatial web about? Digital twinning of the world, metaverses, digital twins, smart cities. I mean, all the all the great IP in the world. I mean, instead of going to see the movie Avatar in five years, you'll just go hang out on Avatar and with your headset on and live with the knobby and learn how to speak knobby and fly the fly the pterodactyls and do all the things. So so the the the metaverse and uh the digital twin world is coming in with AI, and uh that uses a lot of horsepower to do all that computer graphics. Okay. So I think those data centers will be used for that, not for AI. But we uh we we cut down the uh we cut down the need for we don't need data centers at all for what we're doing.

Brian "Ponch' Rivera: 43:57

Okay, so so there is a little bit of uh truth in that. Now let me ask you this. What why is what's why is it so challenging for people to wrap their head around this paradigm shift between, I'll call it a closed system LLM to what this open system active inference is, right? I mean, what what what when you have to talk when you talk to leaders about this, do you see them struggling to understand that the the difference between what things things are ha- what's happening today and what is on the horizon?

Dan Mapes: 44:23

Well, I there's a couple ways to look at it. I mean, one way is of course LLMs have been amazing. I mean, I use them every day. They're they're incredible. I mean, just put a prompt in and you can get a full video out. You can it's remarkable. So um so that's so exciting for most people. They don't need more than that. So they're not looking for be beyond that, you know. Now, NASA and other big corporations and uh smart cities, they need more than that. They're very aware of what we're doing. But also what's happened in the last year is we're starting to see the limitations of the large language models. They're good for making content, but you probably need a human in the loop to edit it. So when I get a 20-page report back on um the history of Los Angeles uh or the design of the new airport or whatever, uh I still have to read through it, and it's still got errors in it. They're called hallucinations or just factual errors, even. So I've got to have a human in the loop to edit it before I publish it. And the same thing with a video. I might want to edit the video a bit. No, that didn't come out. The person has six fingers, whatever. And so um, so as long as you have a human in the loop, LLMs are fine and they're great uh productivity tools. But 95% of what we want AI to do is fully autonomous. It's moving things in smart cities. Think about that. There's 40 layers of data in a smart city: electric grid, traffic, uh, emergency services, uh, you name it. And they all have to be coordinated. And so there's just billions of bits flying back and forth. What about global banking? What about global supply chains and transfers of title and all kinds of stuff? Ports, airports, I mean, these you can't have a human in the loop. Things are moving too fast. So we're the we're the next generation that does the accurate, uh embodied, uh, real-time AI that doesn't require a human editor.

Brian "Ponch' Rivera: 46:13

So I'm kind of curious. When it comes to building those world models, and let me just go back a moment. We talk about strategy. When we help organizations with their strategy, we ask them to take a model of their world. You know, there's mapping techniques that allow us to do that. Generally, most corporations don't want to do that. They're like, well, I just want a mission statement, I want a vision statement, I want a 40-page PowerPoint presentation, and that's what I want to execute on. And we're our we argue that, hey, look, the best way to build a strategy is to understand your external environment. How do you map it? So my question to you is are there solid mapping techniques that you recommend, or how do you recommend organizations get ready for creating these world models that look at the functions of things and the interactions between them as they develop um both human and agent uh models where humans and agents can work together? Is it are you looking at any of that? You're you're really good, Panich.

Dan Mapes: 46:59

I mean, it's exactly the it's now the new skill, exactly right there. And so uh what's great about active inference and the free energy principle is you don't have to have the most accurate world model when you start. You just need a good, solid, small world model, and it will learn and grow itself, self-evolve forever after that. So uh so that's one good piece of news. Uh, if you look at LLMs, they're static models. You build them and they're static, they're frozen. And then that's why you have to wait when they finish GPT-2, then you got to wait two years for GPT-3, two years for GPT-4, then two years for GPT-5, because they're rebuilding the model. Whereas uh with active inference, it's just kind of self-evolving every day. And you never do this retraining and rebuilding a thing. And so um, so that's one good news right there. And secondly, uh then uh you let uh people that have domain knowledge, your specialist in cardiology, let's say you're in the clinic uh is gonna build a cardiology uh AI. Well, now you've got access, they're like websites. So now your your your meta AI, your personal assistant AI, which is like your Jarvis, uh has access to every AI on the internet, right? It's a meta-AI, it's aware of everything out there. And so when you uh when you have a question around the diet uh related to heart health, it might query the AI at the Cleveland Clinic. And uh, it might have a uh validity rating higher than any other uh cardiology AI, almost the way websites do now or apps. So um so the great thing about the way the World Wide Web works is it's decentralized. And so we generally have a swing from central to decentral, from central to decentral. So computers started out with big mainframe computers, IBM was the biggest company, and you plugged everything into the mainframe. If the mainframe went down, all the terminals went down. And that's why you'd go to the bank, they go, sorry, the computer's down at the moment, you know. So uh the uh the internet uh was the uh a way of uh solving for that. Let's get rid of the central computer and let's just create a web of computers, and we'll give every computer an IP address, and then they can send messages to each other all over the world. And then the same thing, uh, when uh the same thing happened in the 90s, AOL got really huge. They got so big they bought Time Warner. And it was a big central information hub with the New York Times, Wall Street Journal, stock quotes, mail, everything, all on your AOL uh uh CD that you paid a monthly subscription for. Then uh uh Tim Berners-Lee released the uh web protocols, and uh one of the people at the New York Times raised his hand and said, Hey, uh we're on AOL, but we don't see the customer, they're AOL's customer, even though they're reading the New York Times. If we have our own website, we can run our own ads and have our own subscribers and do everything that we like to do uh as well, and we'll and we'll have a direct contact with our uh with our users. And so everybody migrated away from the central system to a decentral system. That's now I think we have four billion websites all over the world, built by probably a hundred million people or more, you know. Well, no company has a hundred million people, and I mean I think Microsoft as big as they are, under a hundred thousand. So um, so to say uh to have a company like OpenAI, which has got like, I don't know, four or five thousand people, and we're gonna build an AI for the whole world. No, no, no, no. Get out of here. Yeah. No, no, no. Decentralize. Here's a tool, people. People in Malaysia are building AIs for Malaysia. I mean, AIs for farming in Japan are going to be very different than AIs for farming in Kenya. They've got to be done hyper-local. So now we've got a hyper-local AI. We're not building the damn thing. We're building the tools like Apple in the App Store or like WordPress for or Shopify for the website. You just grab Shopify and you can make your own store and load it up and go. And so um, so we build SDKs that allow uh people all over the world to then download their knowledge into these uh AI uh domain models and then share them with the world. And they can charge for them and uh whatever. It's just very similar to the World Wide Web. So it's a worldwide web of intelligence now instead of a worldwide web of information.

Brian "Ponch' Rivera: 51:17

Do they have to share? Let's say uh there's some something that's proprietary. Let's say I want to build uh a way to understand the stock market, I don't want to share it with the world.

Dan Mapes: 51:26

It's just like a web. You you you you you have a website. Uh you you have to be a member, maybe, you know. You you you you have total control over your over your AI.

Brian "Ponch' Rivera: 51:35

Okay. Okay. Yeah.

Dan Mapes: 51:36

So you can either make it available, you can put it make it available for a price, you can only make it have to be over 21 or whatever it is. You see what I mean? You have all those rules. It's just like the World Wide Web.

Brian "Ponch' Rivera: 51:46

Okay. So the idea of a distributed or federated approach is really like the concept of team of teams that everybody got excited about 10 years ago, which is actually built upon something we learned 30 years ago in the military, and that is you want to have decisions made locally. Uh, you want to push information to the decision makers. The same thing is true when it comes to building in a high performance organization. You don't want to centralize all that, you want to decentralize it. So it's interesting. I was just having a quick conversation with Dr. Bray about this, that I think that this approach is actually going to allow organizations to become more agile, more adaptive, because they have to learn from the technology. It's basically a mirror that says, here's what we're borrowing from human you human. You ought to learn how to do this too. You ought to learn how to work in a distributed approach with a mission command mindset where leaders uh you act like gardeners and provide the context, right? So I I think that that's my hope, is is uh I I think I think you put your finger on it really nicely.

Dan Mapes: 52:39

I mean, when the printing press came out, as I said, hardly anybody was literate in uh Europe at that time. But by but by 1500, there was a lot of literacy. There were 20 million books in circulation, and then we had an explosion of uh uh ideas and uh new thought all through the 1500s and beyond up until now. And so uh this is going to do exactly the same thing. We co-evolved with with our tools. And so now we're gonna have an explosion of intelligence, call it natural or artificial intelligence, or Friston likes to call it natural intelligence. It's more like digital natural intelligence, the same, same FEP that we're running is now running on the internet. We're gonna have access now to, I mean, every every eight-year-old girl in the world is gonna have access to as much information as the top PhD has today. And so we're gonna have uh uh generations of smarter, younger people coming up, and uh it's distributed widely. You can be in Bangladesh, you can be in Alaska, you can be in Greenland, you can be in in Brazil. It doesn't really matter. You're gonna have access to the same thing. Uh the thing I like to remind people of is you you look at somebody like Bill Gates or Elon Musk, they're using the same damn smartphone that a uh a 12-year-old in Morocco is using. I mean, yeah, because really the phone is simply an a uh a window into the real value, the trillions of dollars of value of the World Wide Web. So the uh the value isn't in the little thing you're holding, it's what it accesses. And so now we're gonna have an intelligence web. Oh my God. So you're gonna access it with glasses, with headsets, with other kinds of things. And so everybody's gonna get. I mean, you come back here at 2100 and talk to a 12-year-old girl, and she's gonna be like smarter than Carl Briston. I mean, you know, there's just this, we're we're we're co-evolving with our with our AI, you know.

Brian "Ponch' Rivera: 54:35

So this has got to be a threat to some institutions where they they're dependent on centralized control. And maybe maybe some governments out there that that may not want to see this because you lose power, right? If you come up if you come co if you become closer to a real democracy, which this really enables you to do, right, is is create that uh true democracy. So there's got to be some evil in the world, some pressure uh from institutions or organizations that may not want this. The question to you, are are you seeing any of it seeing any of that or nothing at the moment?

Dan Mapes: 55:02

No, really. No, no, no, because uh the the what the World Wide Web is exactly the same thing. I mean, it uh democratized uh information all over the world, and people use it every day all over the world, and and all the government agencies realize, oh, we're more efficient when we use this. So it really wasn't uh it turned out not to be a threat. And maybe if you describe it, uh uh, you know, here's what's coming. Uh, we're gonna decentralize information, it's gonna be locked in these big libraries at the top of universities. We're gonna make it available to every man, woman, and child in the world. Oh my gosh, we're gonna lose our power for Oxford or Berkeley or MIT or whatever. But that wasn't the case. And I think it's the same with this. I think um the AI comes in, it makes everybody just so much more efficient, so much more capable. And with superabundance, really the need for control starts to lessen because it's all around scarcity, is uh where all these control functions come in. So we can even loosen up a little bit because the world generally is getting more abundant. If you go, you bring anybody here from 1400, they're going, like, how did you guys get so rich? I mean, my God, what's a car? I mean, everybody has a car. I mean, you know, it would be the most amazing thing in the world in 1400. And so um I think uh we got probably half the planet into um what we would consider uh historical wealth over the last 500 years. Now with the age of intelligence, we'll get the other half in as well. And by 2100, we've got a really uh uh Star Trek civilization, you know.

Brian "Ponch' Rivera: 56:34

Wow. I want to shift the gears a little bit and talk about consciousness. I know that there's a there's a lot of discussion about that around uh active inference and and stuff that Andy Clark talks about as well, and and others are talking about it. Um, in your view, based on what you know, will this active inference uh in next generation AI be conscious? Or can it be?

Dan Mapes: 56:54

So you're you're into an area that's uh really impossible to give a uh a definitive answer to because for one reason, we don't really honestly have a good definition of consciousness. And uh and there's two schools of thought right now. One is that consciousness emerged out of matter, and the other is that there is no matter. Uh there's only consciousness. So those two schools now are having a really big discussion because when we when we keep looking at matter closer and closer and closer, it keeps disappearing. When I look at wood really closely under a microscope, I can't find wood. I only see molecules. And then people go, oh, well, the whole world is made out of molecules. I got it. Oh, no, I look at molecules really closely, and they're all atoms. Oh, wait a minute. Oh, the world's all made out of atoms. No, we look at atoms really closely. They're subatomic particles. Then what are subatomic particles? Well, they're magnetic uh vibrational vibrations, yeah. So yeah, so we we're kind of the the materialists and the idealists are having a really interesting conversation right now around that. But what we can talk about is intelligence. Okay. And so we don't have to talk about consciousness. And so uh we can talk about intelligence and self-awareness, uh, that kind of thing. And then we don't get into the argument, is it conscious or not conscious? But uh certainly uh by 2030, looking like in that region, plus or minus three or four years, uh, but you know, it could be somewhere between uh 27 and 2035, we will probably uh uh get to a level of uh of artificial intelligence that everybody would go, like, you know, just smarter than human. Yeah. Yeah. Generally uh AGI is called, you know, and uh and if that if when that happens, and if it happens, uh most uh most people in the game uh believe it's about to happen. That's a big historical moment uh because it's not stopping there, it's gonna continue to grow and take us with it. And so uh so uh our level of intelligence. Uh we can if we talk to a Stonish person, I mean, even a six-year-old child has more knowledge than a Stonish man or woman, you know. And so um so it may be that uh we are the children of our own future parenting. I mean, uh that the adults of 2100 may look at us with total compassion and go like, oh my gosh, those poor people. How did they handle all of their problems with the level of low intelligence they had at that time? We think we're so smart, but relative, it's all relative, right? Come back every 10,000 years and take a check, the average level of IQ keeps rising because the IQ score is all based on the bell curve. So it isn't based on some absolute thing. And so it's like half the people have above a 100 and half the people have a below 100. I mean, but but that that bell curve middle is moving forward. So what would have been uh uh a 150 and uh in 1800 is probably a 90 today. You see what I mean?

Brian "Ponch' Rivera: 59:52

So I've got you brought up vibrations, and you know, we've looked at Cymatics, we looked at sacred geometry on this pop on this podcast. Uh, we know a lot of people that look at that. There's maybe some ideas out there that vibrations and geometry can help us heal. My question to you about this is with your background and how you look at things and with this active inference AI, will it be able to look at these wooish things and help us understand them a little bit better?

Dan Mapes: 1:00:15

Well, AI has been helpful in that in generally. I mean, uh it can look and it can find patterns that we can't see. And so I would imagine that, as you point out, cymatics alone is just an amazing uh the forms come out of vibrational states and just change the vibration a little bit to sand on a on a plate with vibrations on it, and it takes all these uh amazing organic forms. I mean, uh they look a lot like nature. Uh so uh but I would imagine AI, again, like I said, the this new level of AI, LMs are generally taking what we already know and putting it into a thing, and then we could query it. But what active inference does, it's got a curiosity quality. It's kind of constantly probing and looking forward. Uh I think uh we may it may help us discover things that you're alluding to here that are beyond our current science right now. And so the science we have today is uh relative to the science of 2100 as the science of uh 1800 would be to the science of today. I mean, again, we're really cutting edge right now compared to anything we've ever been, but I mean, we still don't know what time and space are. I mean, Dr. Donald Hoffman's work at uh University of California Irvine on uh his great book is called The Case Against Reality. I mean, I recommend everybody to read that. So we're probably on the verge of a whole new kind of chapter in science that will emerge on on the back of this new renaissance we're in. So fasten your seatbelt. It's gonna be an exciting ride. I mean, I I think that's the thing we can. I think the main message I would have uh really, I think, out of the these kinds of conversations is uh um we're really heading into a positive future. It's got doesn't mean there won't be bumps as we transition from the industrial age to the age of intelligence and superabundance. But um, over the course of the next 20 years or so, we'll kind of figure things out and and figure out how we take care of everybody and how we deal with universal incomes and things like that. But but we're headed toward a uh an extraordinary uh future. And uh that's uh that's something I think we all can sign up for. It's something we've been dreaming about. And technology is going to help us as it has historically. I mean, uh, you know, the plow allowed us to uh grogue more food, I think. These little breakthroughs really open up uh huge things, and uh AI is uh a huge breakthrough. So LLMs will have their role and uh active inference will have its role, and uh humans interacting with them and will take the uh co-evolve all of us uh over the next uh 10, 20, 30 years.

Brian "Ponch' Rivera: 1:02:47

Well, hey, I think that's a great place to stop. But before we do that, you brought up books. I just want to ask you, you brought up the case against reality. You got another couple books you want to share with folks that you you think are important to read right now, or what are you what's at the top of your reading list?

Dan Mapes: 1:03:01

Well, gee, yeah, yeah, that's always a good question. I mean, uh you we kind of started this conversation talking about the power of meditation and uh how it can free your mind from just the being too narrow thinking, so broadens it out. So uh I think any any of the great teachers uh in the meditation traditions, I think are really valuable. I think looking back to some of the uh kind of ancient wisdom that has survived more than a thousand years, whether it's Rumi's uh poetry and uh whether it's the Tao Ta Qing, things like that, even the Bhagavad Gita, all the great uh teachings that we've come from Judaism, Christianity, and Islam, if we pull out the essence of uh the values and the teachings and that kind of thing. I mean, uh you can see what people were struggling with in those days, the Ten Commandments and these kinds of things. How do you build a society that really works well and functions? And so as we even the AI will take all this stuff in and probably help us uh really build uh new legal structures and things like that that help us just live together on this planet more effectively and with less violence and and uh more s more security for everybody. So that's uh that's generally a direction I would go. But yeah, books-wise, I mean, you know, gosh, I mean, there's just so much great stuff coming out all the time. I mean, gosh, I just can't even put my finger on all of it. But I I do like Donald Hoffman's uh latest book. I thought I think everybody should read that. It's definitely food for thought. And uh certainly uh some of the work that's being done around uh active inference uh as well. That's uh gonna be a crucial understanding uh in terms of careers and inventing things.

Brian "Ponch' Rivera: 1:04:32

Uh I couldn't agree more. Uh Dan Mapes, thank you so much for being here on No Way Out. Uh we'll we'll end it there, but I'll keep you on the call for a few more minutes. So thanks again for your time. Appreciate it. Thanks a lot, Panj.