The Transcendent OODA Loop: Jordan Hall on AI, Strategy & What Makes Us Human

The OODA loop is not a four-step process. Jordan Hall came on to prove it. Co-founder of DivX and Neurohacker Collective, founding voice of Game B, and former Santa Fe Institute trustee, Hall has spent months deep in math, physics, and AI, and he surfaces with a sharp claim: there is a transcendent OODA loop, the strategy that stays optimal and most evolvable across every landscape, and it works as an attractor in strategy space. Whatever sits closest to that peak wins.

Ponch takes it from Boyd to active inference to the future of work. Why LLMs are powerful but ungrounded. Why the coherent generalist now beats the specialist. Why big organizations are dissolving into fluid squads. Why AI is the best bullshitter ever built, and why the only skill that matters is the ability to call it. And why, on the far side of the disillusionment, nothing an AI can do was ever essential to being human.

John R. Boyd's Conceptual Spiral was originally titled No Way Out. In his own words: 

“There is no way out unless we can eliminate the features just cited. Since we don’t know how to do this, we must continue the whirl of reorientation…”

A promotional message for Ember Health.  Safe and effective IV ketamine care for individuals seeking relief from depression. Ember Health's evidence-based, partner-oriented, and patient-centered care model, boasting an 84% treatment success rate with 44% of patients reaching depression remission. It also mentions their extensive experience with over 40,000 infusions and treatment of more than 2,500 patients, including veterans, first responders, and individuals with anxiety and PTSD

Stay connected with No Way Out and The Whirl Of ReOrientation

X: @NoWayOutcast · @PonchAGLX · @NoWayOutMoose

Substack: The Whirl Of ReOrientation - www.thewhirl.substack.com

Transcript

Jordan Hall on AI, consciousness, and John Boyd

Brian "Ponch" Rivera 0:20

So I'm here with Jordan Hall. I've been trying to get him on the podcast for a while now, uh, mainly because of his work with Jim Rutt, Plan B, uh, some conversations on consciousness, AI. And uh he's actually been on a podcast where he he explained the Ootaloop to um Michael Levin, which I think is fantastic. And Michael Levin has some fascinating work that he's doing with Carl Friston and something in the what they call the platonic space. It's it's pretty fascinating. But welcome to the show, Jordan. How are you doing today, man?

Jordan Hall 0:48

I'm good. Uh we were just out for several weeks in uh California and Texas with families, uh kind of resettling here at the house. We're probably we're 48 hours in, so I'm a little bit a little bit uh you know rattled, but we'll settle in.

Brian "Ponch" Rivera 1:03

That's all good. So I think the one of the last podcasts uh I heard you do was back in March. And then of course the one you did with Jim Jim Rutt. You know, and since that time, Jim Jim has gone on, has graduated. If you wanted to say that, he's he's moved along. Uh and the connection to Santa Fe Institute is is fascinating to me as well. So uh what I'd like to do is just kind of do some sense making for folks, uh give a little bit of not not your background, but what it is you're seeing in the world of AI, consciousness, the connection to religion. Uh I know that's a big ask right now, but uh let's just kind of start there. What's what's you know after you recovered with your family, what what's top of mind today? That's a very, very big ask.

Jordan Hall 1:42

Um big one. Yeah. I mean what I'll do is kind of like remember that game, I think it was called pickup sticks. You kind

Pickup sticks: three forces reshaping how we decide together

Jordan Hall 1:49

of just took a bunch of sticks and threw them down on the ground and kind of started pulling them out. Maybe we'll play pickup sticks. I'll throw some sticks on the ground and we'll see where the the interest. So we have God, there's so many different dimensions. So let me see if I can identify three just to start with. One is a recognition, a forced recognition, let's say by necessity of the individual and the social level of the uh inadequacy of the model of identity and consciousness that has been assembled in the darnity. So the past 500 years or so. Okay. Um, that's one. The second is the consequences of let's just call it LLM flavor artificial intelligence. So I don't necessarily need to go the far beyond where we currently are for the form of sociotechnical structure that we currently live in. So we could call it late modernity if you'd like, kind of a late-stage liberal capital system, or whatever language you'd like to use to describe it is the thing that we live in. But I'm specifically talking about the socio-technical framework. So the things like how we make decisions together, how we coordinate resource allocation, that kind of stuff. And then the third is maybe the most most relevant, but it's the hardest. And that has to do with something like let's see if I can throw how to say it cleanly. When you're some some context, when you're when you're dealing with accelerating change, this is a uh kind of a Wayne Gretzky metaphor, except now the puck can accelerate. I say you schedule where the puck is going, but if the puck can accelerate, that changes things a little bit. You have to start thinking about ways of thinking about the trajectory that you're on that will be true regardless of a lot of uncertainty, uh kind of fog of war in the shape of the landscape of the war, not even the characteristic of the movement on the landscape. Does that make sense what I'm saying?

Brian "Ponch" Rivera 3:30

It does. I'm hearing some Stuart Kaufman in there, adjacent possibles, that type of thing. There's certainly some of that.

Jordan Hall 3:35

Uh and also just straight military theory. If you're if you're if you're planning on a war and you don't know whether you're going to be in the desert or the jungle, you have to think about a very different notion

Accelerating change and strategy that is invariant across contexts

Jordan Hall 3:44

of what does it mean to be properly prepared. This is very bored, right? You just go up one entire level and say, okay, well, strategy is not desert strategy or jungle strategy. It's strategy that is invariant across contexts. Are there invariants in strategy that is no longer bound to a get given context that I can rely on no matter what the underlying context happens to be? And this is the kind of thing you have to do when you're dealing with accelerated change because you can't guess what's going on the world's going to look like in a year and what the secondary and tertiary consequences of this change happen to be. And I should point out, by the way, that the that the result, for me at least, was extremely good news. Like I actually moved from being, let's say, relatively pessimistic about the consequences of AI for humanity to meaningfully optimistic as a consequence of that work.

Brian "Ponch" Rivera 4:31

Right. So you're going to meaningful, optimistic in the age of I I'm with you on that. The abundance that I believe is that it may emerge from what's happening all around us is not necessarily a bad thing. But I think what I'm hearing from you is we have to rethink how we these socio-technical systems operate, from living in a land of scarcity to living in a land of abundance. What does that mean in in this landscape? And I think some of the things you're written in the past about a networked approach towards capitalism, if I can say it that way, where we're moving away from centralized control of government to uh network approaches, I I see that happening more and more. Is that is that what you're talking about?

Jordan Hall 5:10

Yeah. Yeah, that's definitely what I'm talking about. So this just kind of decompose that. When the marginal cost of information and computation drops towards zero, most of the premises of modernity, which is let's say a book-based or literacy-based communications fabric, are no longer relevant. So one consequence is that the structures that we've inherited, and in many ways have kind of optimized or at least cobbled together functional versions of over the past 500 years, will be undergoing really quite significant transformation in

Why Hall turned optimistic about AI

Jordan Hall 5:42

much the same way that the Middle Ages, which had been around for a thousand years, fell apart rapidly in just a century or so after the introduction of the printing books, maybe two centuries if you want to be big aligned. And this, of course, is because the underlying premises of where being a race, not the that which sits on top of the underlying premises. So it's a very different regime of change. That part is it will be challenging. There's no question. There'll be a lot of challenges around that. So then taking it up one level, the next level is some you pointed to exactly, which is, and even more fundamentally, we have been operating under a basic assumption and a cultural logic of scarcity for as long as we've been around. And so moving to a cultural logic of abundance is qualitatively distinct, and as you can point out, it's premised on the qualitative,

Marginal cost of information to zero: the end of modernity's premises

Jordan Hall 6:25

not the quantitative, and novel. We have some good hints on how to do it, but we don't really know how to do that. And so we're entering into a very novel environment. And I think it's important that a lot of people, almost everybody for a very good reason, tends to operate on the basis of analogy to the past. So you say, hey, in the past, technological disruption has always resulted in X, Y, Z. Okay, neat. Or fill in a blank. And the point is, yeah, but novelty is novel by nature. The part, the question is how novel is the context we're entering into. If it is truly novel and significant, then the past tells us very little about it. To the degree to which it's a pattern that has recognized is a pattern that is repeating from the past, the past tells us a lot, right? But we need to be clear on the distinction so we can understand how to actually navigate thoughtfully. And then the last piece, and this is almost like just this just drawing a pyramid, has to do with well, the language, uh uh, the psychological. So moving from the sociological or socio-technical questions of abundance and scarcity. The top of the pyramid and the or the most irrelevant part is actually the psychological. We might even say the spiritual, um, which is that how do human beings deal with abundance? And the answer is generally speaking quite poorly. So it's not just how do our institutional structures, the ways that we coordinate together, deal with abundance. That's a whole issue. That's layer two. The most important part is how do we as individuals deal with abundance? And that's actually some of the largest risks.

Brian "Ponch" Rivera 7:44

So any examples from the past that you can pull from that that or nothing that we've seen.

Jordan Hall 7:49

Yeah, I'll give you like a very simple example, one that's kind of relatively uh straightforward. We, in at least in the in the West and mostly in the United States, moved from a realm of caloric scarcity, which dominated human existence up until the 1950s or so, to caloric abundance by the time we got to the mid to late 1970s. And the result is out of control of obesity. What happened was is that the the governing structures of literally the hominid arc from anthropological time up, it was governed by the fact that getting access to adequate calories was the primary problem. We uncorked that, we flipped it and said, actually, no, now you can get access to many calories as you want, but we didn't actually build anything in place that enable people to navigate that new landscape properly. And the net result was a very large portion of the population flipped into obesity. Um that's a that's an example of both an absent sociotechnical environment and an absence spiritual environment that can navigate the shift from scarcity to abundance. You could do the same thing, by the way, if you even wanted to, more catastrophically in the context of reproduction and sex. Up until the pill, uh male-female sexual activity was driven by scarcity, grounded in scarcity. If it produces children, then and that that shaped all societies up until the 1970s and 1980s. When the pill became ubiquitous and the underlying premise was no longer available, we've now navigated into an environment somewhat paradoxically, but not radically paradoxically, where fertility rates are collapsing globally. And we may actually extinguish ourselves in two generations because we no longer have the capacity to produce and raise children that raise children. So that's another example of where scarcity moving to abundance lacked a and this image, and you can't really see it, but then this image of inversion, because it it actually very much is one of flipping things upside down and moving out of one regime into a completely different regime. And almost

Scarcity to abundance: obesity, and the discipline problem

Jordan Hall 9:44

all of our inertial, inherited, unconscious, in many cases, evolutionary, biological, and then traditional structures weren't designed for abundance and therefore aren't properly able to govern a realm where the choices are now abundance. And this leads to a number of negative consequences. So those are examples. I can give you the minute you'd like.

Brian "Ponch" Rivera 10:22

So the the means to get there to this large access to higher to just abundance and food, the way we did it is not necessarily a naturalistic approach to to feeding the world. It's more of an engineering approach. Can we say that?

Jordan Hall 10:39

Well, let me let me let me just I'll add more words. So we've got we have the ability to decompose an inherited complex event. So an apple has, as a complex event, has let's say three major constraints. One constraint is it's seasonal. There are sometimes aren't apples. Another major constraint is they're on a tree, so you have to actually harvest them in some fashion that requires a but certain activity. And the third constraint is that an apple is itself a complex object. It contains things like fiber and multiple different nutrients. Okay. So what we are able to do is we're able to decontextualize and extract out a apple-flavored apple jazz, which no longer has seasonality. It no longer has any of the context of a tree, and in fact, has none of the bound context of an apple. But it has sugar or sweetness, it has the sweetness of apple, it has the flavor of apple. And so what it does is it becomes a super what's called supernormal stimulus. It hit the portion of our evolved system that says apple flavored sweet is good and give us 50 times that. But by virtue of decontextualizing it for the remainder of the environment, it produces a, and this, by the way, is a classic problem with madurity in general. It unconsciously, usually, removes all the elements of that context that were actually critical. Sorry. Chesterton's apple feedlot. There's something about the fact of appleness, and maybe even the fact of apple treeness that is is actually critical and load-bearing in our nutritional, maybe even by the way, our social law, social psychological being, that when we decontextualize it and optimize for something narrow, we get a problem. So if you if that's what you mean by engineering approach, yeah. Decontextualize, optimize, and upregulate a narrow target, then yes, that's exactly the problem.

Brian "Ponch" Rivera 12:32

Okay. And if I'm hearing you correctly, if we take a more naturalistic approach, understand anthropology, understand natural sciences, and and this goes back to a lot of your work you've done in the past and things you've said in the past, that's probably the optimal approach. And that's that's going back to the you know, the work with complex systems and systems thinking, right? You mean the we think.

Jordan Hall 12:52

Well, well, what I would say is that the the most critical element is the degree to which abundance rests upon individual collective discipline. Meaning that we move the

Chesterton's apple and the danger of decontextualized optimization

Jordan Hall 13:02

criticality of our choices from highly bound by received context into deliberately selected. So this is the the easy example is if you just imagine, man, I cannot get this out of my throat. If you imagine the anthropological, naturalistic landscape, where again, my ability to access ten apples is very limited on the degree to which there are ten apples to get, and that I'm ready, willing, and able to put out the work to get them. And migrate that. This addresses that problem. Uh no, it's all good, man. So you were talking about uh migrate uh apple. Right. So now you find yourself in our plentiful American grocery stores. I could 10,000 apples at my ability to say, cheese, just make it Costco. I can hardly avoid it. Now what happens is instead of having my choice on how many apples to consume be almost entirely received by the constrained context of my natural environment, the choice of how many apples to consume has to come from somewhere else entirely. In this case, it had to come from my own normative, almost ethical and aesthetic, let's say, uh discipline of what is proper nutrition. Now that requires a lot, not j not just, it does require something like a science of nutrition, or at least a, yeah, science is good. I don't necessarily, when I say science, I don't necessarily mean I definitely don't mean a scientism, but a well-understood model of what nutrition is on the basis of a variety of different inputs. So I need a science of nutrition, but I also need something like ethical discipline. And you probably well know that it's not uh trivial. Easy to know that you should go to the gym. It's not easy to go to the gym and not skipping leg day is a non-trivial event.

Brian "Ponch" Rivera 14:35

Right. So uh I d I I know you have some experience in like neutropics and maybe even the uh psychedelic assistant therapy space and consciousness. And if I'm hearing you correctly, we're what what I'm what I feel you're saying is we're what needs to happen is we need to get back to the basics of being human in this in this age of AI, you know, and this as things accelerate. It's a weird mirror image.

Jordan Hall 14:59

So it will look very much like the basics of being human. Uh okay. But it will have to be done on this entirely different basis. That's the key. So if you recognize that the basics of being human already imply the anthropological context, just use our apple tree. It implies the fact that there's only 10 apples. Um, this this is key. So the basics of being human sit on top of a contextual substrate of our received evolutionary history. When you get rid of that, the basics of being human need something else, which is completely different in nature. But that something else will actually begin to

Abundance requires individual and collective discipline

Jordan Hall 15:33

reorient towards something that looks an awful lot like an indigenous Homo sapiens. It'll look a lot like that.

Brian "Ponch" Rivera 15:40

Okay. And that's where you think we're heading? Is is is that what I'm going to do? Yeah. Okay. Okay. So so now can you tie in consciousness again? I I know we kind of started off with that.

Jordan Hall 15:51

Yeah, and let me let me just kind of hit that point. So if you imagine AI or accelerating change or the singularity narratives, most of those narratives tend to look like transhumanism. They tend to look like disembodied minds or robot bodies or something that uh David Egan would have written about. I'm actually arguing that the the reverse, that the only futures that are stable, futures that are actually it's called viable timelines, look an awful lot like a archaifuturism, that human beings look and behave much more like we did in pre-civilized environments, pre-agriculture, than we look like space robots, for example. And that's why you're optimistic. No, that's actually not why I'm optimistic. Although I that does make me optimistic. That fact is does make me optimistic, but the the reason I'm optimistic has to do with the fundamental nature of eDIOS.

Brian "Ponch" Rivera 16:44

What generative sides of that or learning, adapting, adapting?

Jordan Hall 16:47

What do you think? Oh, it's if you take the idea of saying I can think about a way of I can formulate a theory of strategy that is generalizable across domains. Again, I don't have jungle or desert. I can use the Outinek and think about a meta-strategic approach. What I can do in the context of AI is I can say, well, is are there, is there sort of like a stable attractor in strategy space? What is in fact the most strategic strategy, the most powerful strategic position available in principle in all possible landscapes? And the answer I think is yes. And that answer turns out to produce an attractor in the form of what AI must look like. In other words, an AI that isn't in that location will be strategically suboptimal compared to an AI that is in that location, and therefore will lose. That's what it means to be strategically suboptimal in a conflict. And so the natural trajectory of the evolution of the AI landscape was towards that location. It turns out that that location looks pretty good for us humans.

Brian "Ponch" Rivera 18:45

Okay. Now I I don't think everybody shares this your point of view on this. And and maybe nobody, in fact. Yeah, may uh no I I mean, uh I look at it as you know, as as we get this hockey stick acceleration, the snowmobiles as adjacent possible. And your your analogy of hockey puck that can actually accelerate. That's the space we're in right now, right? And and I don't think most people can wrap their heads around that. But I think a lot of folks are you know fighting the last war, using that type of metaphor, that type of analogy to how we they apply this thinking. And going to your point about strategy, when you think about strategy, I heard you on a podcast talking about what we learned from the Manhattan Project, and it was a process, right? So the the innovation process is more important than the actual weapon that was created out of it. I think that's more important here, too. So you got to take the well, we'll call it the process for now, take the process out and understand how a strategy is applied and start thinking like that. So when we talk about strategy from a Boyd uh you know, a Boydian view, a mental tapestry, right? We have to understand the landscape. We need some type of optical flow. We need to create some type of map, and we know that all maps are generally wrong. So and you're talking about attractors, and a lot of people don't know what attractors are. You know, in a fitness landscape, we we talk about that in the brain, we talk about that in in actually in the markets as well and in music, uh, but most people just don't can't can't grok this at the moment. We're starting to see the sports world pick up on attractors too, because they're starting to see how you can train people using in perception action loop and things like that. So everything's kind of converging on the language you're using in here, but this language is kind of unknown to folks. And and I think when they start learning this, they can start developing better strategies or a better strategy, understanding the landscape. And then going back to sense making, I'm gonna throw something at you to see if it resonates. When when I learned about sense making, I learned at it from a view of situational awareness, third-level situational awareness, which is anticipatory thinking. You know, what happened, what's happening, and what could happen in the future. Most people can't figure out what the hell happened yesterday, right? They're arguing about that. They don't know what the hell's happening today, and they certainly don't know what the hell is going to happen in the future. I don't think anybody does. But that's what I'm hearing from you. And that's what, you know, when I read your work and I see what you're doing on podcasts, it's very aligned to this type of thinking. And it's not to say that it's right or wrong. It's just that uh if you take a naturalistic approach to decision making, sense making, how we perceive reality, that that's what I'm hearing from you is

Archaeofuturism: why viable futures look more indigenous than transhuman

Brian "Ponch" Rivera 21:03

you're taking that approach to project forward in the near future what's what could happen to us as humans.

Jordan Hall 21:10

Yeah, yeah, absolutely. Um just to kind of make you think of s simple examples. So first the a premise, even an invitation. Because you made the example of sports, that's beautiful. More people in different domains that begin to grasp the utility of these ideas, the more they notice that these ideas in fact have real utility and deploy them. And then they become the ones who are successful, and that then begins to bring the ideas into the domain. Other people are like, well, how'd that guy get successful? Well, he started thinking in this way and acting in this way. Well, shoot, I need to do that too. And of course, we see that that's the history of everything. If if you can, anulated. Sometimes you can't, sometimes you are constrained by other opponents. So if I get an example, let's

The real reason for optimism: the nature of the OODA loop

Jordan Hall 21:50

try one of those. Let's say I had a team of 11, 11-year-old soccer players, and then I had the Spanish national team who just won the World Cup, and I had them play soccer. Yeah. Who's going to win? More likely the Spanish team. Yeah, with very high likelihood. Like you, it's not absolutely certain, nothing is. But we could put a very you could put a pretty big bet on that and be pretty confident. And the point, of course, is the reason why you're able to say that is you have an internal model of how things like sports works, how gradients and quality of performance software, things like that. And you can do that across the board. Now, the point is that you can make a model of how a particular domain operates and use that model to that give you predictive power over how that a scenario would play out. This is an extremely trivial thing that I'm saying, but the point is we start at the most trivial, then we begin to work step by step. Now, let's say I take a uh a Spitfire and a uh an F-18. When? Unless the guy who's flying the F-18 isn't is or the plane or the man are dysfunctional in some way, it's gonna be the F-18. In fact, most likes the Spitfire doesn't even know he got blown up until he until he's uh looking at down from heaven saying, hey, what the heck happened? Um again, same idea, right? You can do the models, you can understand things. Now, as you begin to narrow the gap, obviously this is where a lot of board's work came in, but we can start saying things like that which has the ability to observe its environment, orient on what is the relevant set of choices to make in that context, make decisions that are

The attractor in strategy space: the most strategic strategy

Jordan Hall 23:15

actionable within the most effective deployment of Of energy and then deploy those decisions in the most effective and efficient way. Um A and B, right? One that is good at that, well, I'll compete one that is bad at that almost all the time, which is the most basic premise. Then we can start asking questions about what is that which optimizes against all the characteristics of Boodo's. Now we start getting interesting. So you can say, well, what would it look like to have something that has higher reaction time? Okay, that's neat, because that's going to be able to move from decision to action in a particular fashion. Nice. What does it look like to have problems with perception to observation or orientation? Okay, I've got a bunch of sensors, but I can't process the information. I've got a bottleneck. Okay, that's interesting. You can then actually run a process of looking at every single aspect of this kind of a loop and think about what does it look like to produce something that has as its basic nature, it is the optimal OODA loop and is an always optimizing loop. In other words, it isn't just statically good, it is also maintains evolvability at the highest level. And so if you can state the question in that fashion, then you're actually saying something which is which is this works for me. What I'm about to say works for me. It may not work for most people, but it's like saying how many angles are in a triangle. I'm sorry, how many degrees are in the angles of a triangle? And you go, well, 180. I go, yes, exactly. Or we can say that, say that with real precision. I call that a geometric statement, meaning circle has 360 degrees. Yes, no. Yes, cool. Um the beautiful thing about a geometric statement is that it speaks about phenomena or events that are transcendent in major. It doesn't matter which circle I'm talking about, circle in essence has a circle of characteristics. Once I know those characteristics, I can use them in a particular way. What we're trying to do is make a similar kind of argument about strategy, a similar kind of argument about oodaloops and say, is there a way to talk about the transcendent oodaloop, that oodaloop, which is the optimal and most evolvable? In other words, it maintains the highest level of in a given context and maintains that over arbitrarily large number of uh contexts. Such an object, such an event, will under most circumstances outcompete anything that is below it. That's the attractor. So this to give to give you a language, the attractor is there's a location in strategy space, which is this object is optimizing a volveability in Oodaloop, which anything that is trying to compete in strategy will move towards that. So the phalanx will become the Legion. Why? If you select against the Oodaloop space, Legion has closer to this transcendent peak in Oodaloop space than Phalanx is. Well, that's interesting. Napoleonic maneuver warfare outcompetes, you know, Frederick the Great, mass uh focused. Okay, why? Well, because it actually has a higher location on this landscape. Oh, neat. Well, if I can just grade things by their vertical or or their proximity to the to the peak, then I can say, well, if I can begin looking at what kind of moves move me closer to the peak, what kind of moves move me closer to the peak faster? That's the attraction. So that's that's the try I was trying to recapitulate the idea of not just your attractor in general, but the specific attractor that we're speaking about so that maybe more people could understand it.

Brian "Ponch" Rivera 26:18

That's great. That's a great, great uh way to look at it. Uh hey, so you're you're very versed on the Oodaloop. Uh, where did you come across it, just out of curiosity? When did you when in your career or your background did you find it? About 25 years ago.

Jordan Hall 26:30

I can't honestly can't remember the specific context. It's really odd that I can't, but I do know that I read and went through, somehow got access to the primary work, like the slides, the slide decks that Boyd put together, and then got, I was like, that's genius. I'm I'm a big fan of genius. So when I find genius, I had time to dive in and then said, All right, now I'm gonna go broaden out. I read like three commentaries, very comprehensive commentary, took it all the way to the end of Boyd's life, which was interesting to me because it's a number of post-structuralist thinkers that one would not ordinarily imagine would be part of his his uh his milieu, who I was familiar with separately. I'm like, oh, this is interesting. You know, I've got this guy who I'm reading through the lens of fighter pilot and then you know, strategic theorist. Like, wait, he's actually dealing with, I can't remember who it was, maybe Foucault or a Derrida or something like that. And he's doing something in a way which is he understands what they're talking about, and what he's saying is very interesting. So that reinforced my uh thought, thinking that this was a good thing to spend time on.

Brian "Ponch" Rivera 27:25

Yeah, it is. Uh and you know, we have access to the archives now, his handwritten notes. There's they're making that more available. I think you'll we'll see an announcement

Is there a transcendent OODA loop? Phalanx to Legion

Brian "Ponch" Rivera 27:31

here from the uh Marine Corps um history division in the next few months. Uh and I think a lot of people look at the OODA loop as a linear four-step process. You know, when I started to really pay attention to it, it was a connection between Eric Reese and Stephen Blank when they wrote the customer develop model after John Boyd's observer to side act loop. And of course, you get into fast transients with the lean startup and concepts like that. However, a lot of their view of the OODA loop was a linear approach. And then you get into the agile movement and the software development. I'm sure you're familiar with all that. There's they're they're still applying the linear thinking of the OODA loop. And I'm like, this this that's not the way I learned it in the military. Uh and then you dive a little bit deeper and you see the connections of neuroscience, of philosophy, anthropology, biology. And you're like, wait a minute, this is not a force-depth fighter pilot decision-making process. And and I I see that when, you know, we have Carl Fristenon saying that his active inference or the idea of active inference is isomorphic with Boyd's Oodaloop, uh, then I'm like, wait a minute, this this is more powerful than we think. And it's just, you know, it's it's about the lightest sketch of how a living thing operates with a changing environment. And I think that's what, you know, one of the reasons I want to talk to you is because you're already talking about these things and using the oodle loop. Uh, and I think by helping those business leaders that that know about it from the the cocktail party they went to last night or the event where they heard a former Top Gun talk about the four-step process of Oodaloop, I think that's an opportunity, right? I'm like, wait a minute, we we need to dive into this and say that thing you're hearing actually has a lot of value.

Jordan Hall 29:01

And it has a, as you say, a lot of richness that you may not be tapping into. Like you say, it's yeah, it's not, in fact, a linear four-step process. Each of the four components is is valid as a component, but that's you you you can be much more affected if you think about it more and more appropriately as to what it was actually talked about. Let's say and the active inference, uh, how to say orthogonal connection adds even more richness. Because now what that means is you can actually expand out and look at all the stuff that's going to happen in the active inference world and use that to bolster whatever it is that you're locally doing.

Brian "Ponch" Rivera 29:32

So so on that note, when we look at you brought up large language models earlier, we may not want to spend a lot of time on them, but there's there's a ceiling to them, right? They're not actively engaging with the external environment, if I could put it that way. So there's a shift here with like reinforcement learning and using world models. Can you talk a little bit about that and what that looks like in the world of physical AI?

Jordan Hall 29:51

Not a lot. I mean, what I can I can tell you maybe three things. One is, interestingly enough, I'd say, because it's not obvious that we're going to make breakthroughs in certain domains. So we know that world models matter. We were having trouble building world models that were actually performing effectively. And then there was a breakthrough in uh Lacun's team around world models not long ago, less than a year, maybe six months. I can't remember exactly now. And suddenly we're now in an environment where that's likely going to be on some kind of accelerating performance curve because it begins to add a whole, again, orthogonal dimension to what LLMs can do and fully compatible. There's like the the well, it's like the uh right side of the brain, the left side of the brain, more or less, like the uh language and kinesthetics are highly synergistic. So and then we should add things like you just think of a better way of thinking about it. This is actually a better way of thinking about it. The competitive frontier is extremely intense, and meaning that there's a a number of players, although not an arbitrarily large one, but someone maybe 12, I don't know how many there are specifically. The Chinese kind of throw some interesting monkey wrenches into how competitive landscape looks like recently, which could expand the name of competitors. But you know, open AI, Anthropa, now is in there, Google's got the play. China writ large, let's just cat make that a whole giant China Inc. kind of a big. So that's five. Maybe there's seven, something like that, but not many. But for the most part, these guys are all in. For the most part, they're saying this is a winner-take-all, zero-sum, full domain kind of a gig. What I mean by that is if you look at Facebook, this is a classic example. They discovered, oh wow, it turns out there's only one. Well, once Facebook achieved dominance, they owned this giant thing. There weren't like 15 competitors in Facebook land. There's one Amazon.

Speaker 1 31:39

Amazon.

Jordan Hall 31:40

That's the whole space. It's a kind of network effect. And the basic premise is that

Where Hall found Boyd, and reading him past the fighter pilot

Jordan Hall 31:45

uh AI looks something like that. There might be an AD to Intel. There might be a Twitter to Facebook thing, require those niches. The world is not a sing a smooth, untextured topology. But the basic premise that's driving basically everybody who's plugged in the AI space is that uh whoever wins wins very, very big, 85% and more. The size of winning is enormous because you're talking about an abstraction layer that is above money. It's larger than winning the money game, winning the coordination and intelligence game. You maybe winning the deal with boom. Um and therefore, all chips are in. And so the the consequence is that the uh the frontier of what can be done to achieve local strategic advantage is extremely explorable. I think about uh World War II. All the players were like, this is a big, this this very intense conflict, existential. You know, whoever loses, loses completely. Whoever wins, wins becomes global global hedging on for some meaningful people to die. Leave no stern unturned. So the classic you know, Nazis were looking to see if Odin's spear can be found. Did they believe Odin's spear was real? Hard to say. Doesn't matter. They don't care. Like, look, if Odin's spear is real, we want it. And if it isn't, it doesn't hurt. Like, well, we're gonna we're gonna investigate it. And the point of that is that the notion of

The OODA loop is not a linear four-step process

Jordan Hall 33:00

world models is a highly likely advantage for everyone knocked it out. And people like Elon are looking at it going, hmm, what are my strategic advantages? Tesla. Interesting. I have the strategic advantage that I have by by very large far the most real-world information sitting in things that are AI at the edge. And I'm actually actively teaching AI how to navigate it. That's neat. All right, let's lean into that. For a little while, Facebook had the advantage of the QM's research. They don't need more. And so that's kind of becoming more of an academic open source thing. But the point being that you you wherever advantage can be found, it will be explored with significant intensity. And so at a high level, we can say where are advantages likely to be found, and world models is extremely likely because it's so orthogonal. You know, we look at the gaps that LLMs have. One gap is they is that they don't live in the world. And the second gap is that they are not grounded, meaning that they don't have a comprehensive closure on where they are. Entirely received consequence of humans doing language. If humans didn't do language, LLMs couldn't produce anything but neurons. That's why their training data, the reason why the LLM structure or the math produces something that is actually interesting in the world, is that language is grounded in reality. And that grounding in reality is because humans are grounded in reality and we have produced the corpus of language. So because those are two caps, the exploration of what does it look like to make the uh stronger, actually, there's a third cap, is that because they're not in reality, they don't have agency. So we've had add agency, and that agency frontier has now been a big part of the last what year, year and a half? And will continue to expand really quite dramatically. I should just sort of plan to

Active inference is isomorphic with Boyd's loop

Jordan Hall 34:35

be surprised by the expansion of what's happening on the AI agency frontier for the foreseeable future.

Brian "Ponch" Rivera 34:42

I'm curious if you can bring up the open AI controversy. Uh okay, if if you we do have these AIs that are that have a world model, they're lower energy, um, and we all have access to it. Uh I would say that's a that's a good thing. If they're protected, if one company has universal access and and dominates, that's probably not a good thing. Uh can you comment on that and and correct me where uh you know what what would help help our listeners understand what does this networked world look like in the age of world models and AI?

Jordan Hall 35:16

Well, let me just sort of state the the basic problem. Asymmetry, the word word asymmetry is

LLMs, world models, and the gaps: grounding and agency

Jordan Hall 35:21

good. We actually have even used the Spanish national team versus our 11-year-old soccer players. There's an asymmetry there. That's kind of the defining factor why you can say the Spanish will win. Asymmetry and a number of different objectives. Well, what AI does is it produces an asymmetry across a very large front of important characteristics, fill in the blood. Asymmetry and medical research, asymmetry in mathematical research. What week, month, we've now seen AI cracking stuff that has never been cracked before in a surprisingly blase fashion. That's an asymmetry. And asymmetry is power asymmetry. If somebody has a power asymmetry over you, then that somebody can make you do things that they want to do and you don't want to do. Simple as that. Um and because AI produces a potentially very large asymmetry, and I should maybe also tease that apart. Right now, if I have access to Claude Fable, I and you have no AI at all, I am significantly more capable than you are in a very large number of domains, and the breadth of domains matters a lot. Like if you're an elite software engineer and I'm a mediocre software engineer and I have Claude Fable, I might be maybe roughly equivalent to you, maybe not as good, actually, because verticality matters. If you're elite and I'm elite and I have fable, I'm probably 10 to 50 times, five, zero times more capable than you, just in terms of productivity. But here's the thing I'm also a pretty darn good physicist. I'm also a pretty darn good real estate agent, and I'm also a pretty darn good neuroscientist because AI is has that has breadth. And so you, as an elite software engineer, may know nothing about any of those other domains. That's the very nature of becoming elite, is that you have to have focused in that narrowing. I may not be able to compete with you in software engineering, although I might get close, but I can do something which is a thousand times, ten thousand times more than you in all these other domains, because the AI is part in that direction. So that's the that's the the the multiplier effect or the uh asymmetry in one way. And the other asymmetry, and this is the one that everybody talks about, and we're already seeing them, is uh the feedback loop. So if I use my AI to make the next generation AI, and you're just having to code by hand, then I get the advantage of I get basically compound interest. You're sort of uh lit linear accumulation of stuff, and I'm getting compound interest. So next year you've got 10 sheaves of wheat and I've got a thousand off. And then, you know, it just the curve is a geometric curve as opposed to a linear. And that is the problem. Like that's the capital T A G problem, is that if there is a compounding effect of intelligence on intelligence, which there is to some extent, we don't know if it's capped. We certainly are seeing the real effects of it in terms of the competition between AI companies, then he who has access to the most intelligence doesn't just have a particular advantage, he has an accelerating advantage over time. And so therefore, you you might find yourself in a circumstance that's kind of like the British showing up to the Polynesians, and well largely, meaning so much power of symmetry that whoever happens to have the high end of the stick can really dictate turns to sort of everybody else. And I shouldn't even expand on that. I'm not just talking about the ability to do software engineering, I'm talking about the ability to do material science, I'm talking about the ability to do physical aerospace, all domains that require intelligence to drive innovation, which is literally every domain. I have an asymmetric advantage, perhaps a remarkably large asymmetric advantage, which means that my power advantage multiplies over time. That is the biggest risk. That's the thing that most people are worried about. And most people who are worried about things, that's what they're worried about.

Brian "Ponch" Rivera 39:05

So did I hear you say that maybe generalists in this age are more important than specialists?

Jordan Hall 39:10

Yes, you did. Yeah, I I implied it. I didn't say it, but that's definitely the case. And I need to hear something like a coherent generalist or a polymath, not somebody who is a just sort of dabbler in many things, but somebody who's actually able to achieve a kind of a B plus or higher level of grounded quality in a number of different domains, suddenly is going to have already is having substantial improvement in performance.

Brian "Ponch" Rivera 39:32

So Boyd was, you know, he talked about cross-referencing domains all the time that you you got to look across, not not in the vertical aspect of things. So I think what he wrote about years ago, and I think what what's being shown today is that's true. Um and then going back to software, you you know, your days in software, my days coaching software development in the Agile space, you know, 80, 100 people coding software has been replaced by a few. Is that correct? Yeah, absolutely. Yeah. And and I remember, you know, 15, 10 years ago that you had coding camps where everybody was saying, you need to go to a coding camp, you need to do this thing because these jobs are not going to go away. And that was less than, I'd say, seven, eight years ago, right? And and now we're starting to see the decline in software developers, a requirement for them. And then guys that were Marines that played with crayons can develop things. And I I'm saying that in a nice way because my friends who are Marines, they can actually use cloud and things to create software, right?

Jordan Hall 40:23

Yeah. Well, and you got you got two different things that are going on there. One is to the degree to which the amount of work that is being done by an organization is limiting, it's bottlenecked in some way. Then you're going to see a significant decrease in the number of people who need to do that work because the you know, the two or three best can actually do all the work and and frankly do better. Now, now, of course, that to the degree which is a is an assumption. It may be that the the the the uh organization will simply expand the amount of work that it's doing. This then goes into category B, which is the the competitive landscape explode. Very large number of groups that were previously unable to produce software, for example, suddenly can. They can either produce it for themselves or produce it in the marketplace. And the groups that did produce software are now producing more of it. And the question is whether or not the need, the overall demand expands at all in relationship with the overall supply. That's uh sometimes this is very important, by the way. There's there is no law of nature that an expansion of supply will produce an expansion of demand. Some things, we've seen that. We have seen it, for example, that we improve the efficiency and the produ and the uh availability of energy, that

Asymmetry as power: the compounding intelligence problem

Jordan Hall 41:30

beats paradox. However, an increase in the supply of beanie babies did not increase the demand of beanie babies. There are things that have fixed demand, and supply does not drive increased demand. So we have to be thoughtful about what exactly are we talking about. Now, you know, this goes back to that notion of abstraction. It may be that there is, in fact, fixed demand for the kinds of things that can be done in software, but the demand for the kinds of things that can be done by intelligence is less fixed, flows after. Um, and that means we may be entering into opening up domains that we haven't actually even thought about recently, or maybe ever.

Brian "Ponch" Rivera 42:00

Mm-hmm. I'm kind of curious, the type of jobs, like if you want to help our listeners, the individual listeners build resilience capacity in this environment, what are the things you would recommend for them to do? So we're talking about job seekers and things like that. And then the same thing for organizations. If they want to re you know, survive and thrive in their own on their own terms uh going forward in this age, what do they need to do?

Jordan Hall 42:22

Well, this all depends. Uh you you definitely should be very aware of the impact of AI in your industry. That is that that would be that's clean and simple for sure. This is not a trivial thing that you should ignore. This is substantially more important than the entire internet, for example. I would say it's substantially more important than computation thus far. If you imagine yourself sitting in 1958 and somebody came along and said, you should probably pay attention to computers, the answer is yes, you should probably pay attention to AI. So what does that mean? Um, let me give you an example. If you're if what you do, if your industry or what your activity is is is very much in the ground, let's say you build uh this is so hard, you you do wire wire harnesses for drones, then you're gonna want to do at least two distinct things. One is you're gonna want to leverage AI to tighten your processes and your operations so that you're able to get more effective, efficient inputs and outputs, communication from customers, designs and specifications, procurement, um, logistics, like tighten everything up. Just like this is what we did with the information economy in the first place, right? We actually wanted supply chain management. Just think about something equivalent to that. The other side, of course, is the innovation frontier. Going to have to have some kind of loop that is monitoring your environment to look at competitors, potentially emergent competitors, who are pushing the frontier outside of your capability of actually meeting the current standard of your market, because your market's going to be evolving faster. So what if your product meets the market's needs now, it may in fact not meet the market's needs in a year or two years. That that might be true in terms of functionality, it might be true in terms of price. Like I think I actually saw some guys who pushed out a new, as well, I mentioned wire harness, a new way of producing wire harnesses that's like twice as efficient. Well, shit. I mean, everybody else is effectively obsolete unless they can accelerate and catch up. Now notice this is I characterize it as a new loop thing. You're going to be using the LLMs to be monitoring your operating environment, right? You actually need to be doing sense making at a vastly higher level because the known players may not even be the relevant players on the maybe an adjacent movement, somebody who'sn't even in your industry, or by some startup out of who knows where. So you have to be able to scan your environment, upregulate events that are of meaning to you, which is a whole different kind of heuristic observe. Oh, yeah. Then you have to be able to decide how you're going to act on the basis of this new information. And then you have to be able to build a flexible enough

Generalists, specialists, and breadth versus depth

Jordan Hall 44:39

organization that you can pivot, pivot, pivot. Um, and so a lot of the stuff that we talked about, say, for example, in Agile, starts to become intrinsically required for all kinds of organizations because uh the environment we're operating in is getting hotter. So you use that as a simple metaphor. As you heat things up, you have simultaneously more fluidity. So as ice melts, things become able to move in directions they were no longer they were previously unable to move, and you have more energy. They tend to be moving with more intensity in certain directions, more than water. There's a metaphor. So the entire economy is heating up. So if you're in something that you have that has been for the let's say real estate or uh construction. Construction has largely been untouched by technology. I mean, nail guns or technology, but untouched by the information technology revolution. And therefore, it's relatively slow, relatively inefficient, relatively slow. By hypothesis, everything's going to heat up. Therefore, your industry is going to heat up. And your industry is not used to being hot. It's used to being cold. And so you have to learn how to operate in a more nimble fashion. So you might want to bring in somebody who understands, let's say agile, and say, what would it look like to apply deploy agile as a fundamental way of thinking about how to build an organization in the domain of construction? So that's that's kind of the industry side. I give three or four points aside. I would look at it in terms of harvest, horizon, and fundamental. So this is a gigantic economic dislocation, and there will be trillions and trillions of dollars invested in it. You can harvest that. If you happen to be an electrician, I would recommend that you learn how to build electrical wire systems and hyperscalers. They're going to be needing a lot of electricity. And there is a radical scarcity in people who can actually deploy the electrical capability and/or hardware that they're going to need. Literally, trillions of dollars will be invested in that sector. And if you can move into that environment, you will simply be on the receiving end of a large wave. And it doesn't really matter how many water rules are built for that wave, a large wave returns a lot of water risk. Water rail here is a metaphor of how you can wave energy into your energy. So that's an example. You could simply navigate towards if you happen to be somebody who can figure out how to allow hyperscalers the ability to build in your region because you can navigate the real estate and regulatory and political environment, you will be making a lot of money. It's as simple as that. So that's one. That's a car list. You just notice where this enormous pouring of money is going to be. The point is it's enormous. And then figure out, happen to be able to move yourself towards that enormous pour in some fashion that hold out your cup and let it pour into your cup. The second category has to do with the example that I have, because I lived through it, was when the internet began to emerge, there was a whole wave of companies that popped up that helped legacy industries begin to migrate into the internet. They built websites, for example, you know, did the back end. What was that called? SAS? No, no, SAP. Remember those guys? Pubblestogles. Um, they existed to help migrate legacy economies into the information age. Well, something like that is going to exist ad nauseum at all levels in the AI age. So if you happen to be somebody who could master how to use large language models, and I'll give you an example in a moment of how weirdly trivial this can be, then you can bring that out to small businesses, medium businesses, or large businesses, depending on how good you are. And if you're interested in scaling, you can hire other people and train them and build a larger business. So the migration of every single piece of the economy across that threshold is an opportunity of enormous size. And kind of anybody can do it if you choose to put your mind to it now. That will not be the case in years, like three, four, five years. We will become more and more specialized and more and more professionalized. But right now there is no professional LLM jockey class. And so if you decide to learn how to do it, you'll be the one in your neighborhood who can. And then that gives you an opportunity. And the reason why I was going to give you an example is that this was uh how old are you? 53. Okay, so we're roughly the same age. I'm like, you're older than you. So Gen X discovered, not everybody in Gen X, but many people in Gen X discovered it. Our willingness to learn how to do HTML coding, for example, or to become a sysadmin was meant that somebody who had no training whatsoever could suddenly become the expert in this thing called the internet in their company and suddenly became a the guy. Fill the blank. You know, you could become the guy at the newspaper. You know, the boss of the newspaper comes in and says, You, you're 19, go figure out the frickin' internet and do it for us. Okay, go. And you could do it because the amount of information needed to become more capable than anybody else in your environment is actually relatively limited. And if you're willing to learn it, then you can. And the leverage is high. So that the point is the frontier for kind of like Gen Z, older Gen Z, youngest millennials, but mostly Gen Z is wide open. Just become the become very, very skilled. And for the next three to five years, you will in fact have an asymmetric capability that can be deployed more or less anywhere. Okay, so that would be sort of an example. Number two, you can do the other work, which is say, okay, there's what what industries will be disrupted in what way? So if you happen to be a junior software engineer, it might be a little tough. You might have to say, am I good enough to actually start competing in what's going to become a substantially more competitive environment? Or should I pivot to becoming an AI migration specialist, for example? Uh certainly if you're a copywriter of mediocre marketing copy, you should look for something else. Just not going to exist. It's a thing anymore. By contrast, as I mentioned, electrician, good job for sure, for the next five years, maybe seven. Now, here's the problem: supply and demand. Lots of people are going to get that, and lots of people are going to move into that. And as this

Building resilience: what job seekers and organizations should do

Jordan Hall 50:03

hyperscaler wave comes and goes, there'll be an oversupply. So that's could be problematical. You gotta be thinking about that. This is gonna be a real challenge of how do you ride these waves? And this goes back to that notion of agile. So then we'll get to the kind of the fundamental. The fundamental is everybody's willing to have to become more disciplined and more agile. That's all there is to it. Um, there is no, let me see if I can I'll put this in historical context. Once upon a time, they were large companies, and you'd go work for them when you were young, and you would work at them their your entire career, and you would retire and they'd give you a watch, and that was that. And and a pension, by the way. They had a thing called a pench. You used to get getting paid money by the company, and then you didn't work there anymore. Um obviously, by the 1980s, that began to break down, and as far as I know, now it doesn't exist at all. And if it does, it exists almost entirely in its uh kind of mutant character caricature form. The point is this that basic notion of small, small groups, become very fluid and very capable on your own, and also become very good at assembling into groups of three, five, eight, think about it at that kind of like squad level, and spontaneously jumping into a squad, identifying a problem domain, rapidly building an appropriate approach, solving the problem, and then being able to dissolve back into individual groups, right? Become more like the special forces than like the 1950s army, is a simple way to put it. And that's just what it is. Like that's just again, the environment is getting hotter. Hotter environments mean smaller, more fluid, more able to bond in different ways, organisms. And this bonding matters. A metaphor I used with Jim, right, I'm not sure publicly, was the notion of carbon bonds. In other words, become the kind of person who people are loyal to because you are loyal, because you are trustworthy. People trust you because you are trustworthy. You are reliable. Why? Well, because in an environment where large organizations don't exist anymore, your ability to attach to something and just ride that positively or paratically goes away. Meaning your ability to thrive becomes your ability to become, to, to be choicefully bonded over and over again. Three days I'm working with this team, and then for three days I'm working with this team, and then three days I'm working with this team. Well, what does selecting mean to these teams? Ultimately, it's because you are the kind of person other people want to work with. And that kind of information, by the way, in a world where the marginal cost of information plummets to zero, becomes highly available. Literally, just is this is what it'll look like in the future. I will have my own personalized LLM agent that is an agent that knows me very, very well. And I'll wake up in the morning and I'll say, Samuel, I feel like the following three things. And Samuel come back and say, Of those three things, these two are extremely ripe right now. Would you like me to negotiate a project for you? Well, yes, I would. Samuel then reach out to, you know, into the into the web of environments, would come back and say, We have the following three invitations on the basis of our reputation. Would you like to join one of these projects? Click. Yes. That all that all on the back end, that's a lot of information, a vast amount of information. There's going to be actual, think about like the way that engineers get hired in the post-open source world. You'd say, hey, well, show me the project you've worked on. Well, show me the data stream of everything that you've done and everybody who's interacting with you. And it'll be processed instantaneously. And so you are going to be a walking a ledger of your actual relational quality, your ability to perform again, things like reliability, trustworthiness, effectiveness, efficiency, ingenuity, all the sort of virtues of collaboration. And weirdly enough, oddly enough, that is what it's selected for now. So all the various sort of vices of collaboration that have been able to be selected for, things like good at pretending to be effective for virtue signaling, these start to get selected against. It's just because the environment is able to be sensitive to small numbers of people constantly renegotiating their relationship and has the ability to perceive them. And so therefore they'll be selected against. And so the weird truism is become express the actual virtues of collaboration at a very high level.

Brian "Ponch" Rivera 53:58

I want to read back to you what I think I heard, or the way I would put this, and that is in the military and in fighter aviation, uh, we have crew resource management, which is the foundation of team

Harvest, horizon, fundamental: riding the AI wave

Brian "Ponch" Rivera 54:07

science. So think about uh commercial aviation. Uh the pilots that fly together generally don't know each other. They just do exactly what you described. The way they do that is through the crew concept, the ritualized form of a team. They have the teaming skills, the non-technical skills. If I'm hearing you, when you talk about the three, five, eight small teams or squads out there, that is gonna be more important now more than ever. The the non-technical skills you bring to the uh environment to collaboration are important. And I think there's another element in here. When you start bringing agents in here or robots, robotics in, or uh trying to think of another name, not not I guess not gonna say NHIs, but human machine teaming becomes a thing, right? So so you brought in Samuel, who's your physical AI or whatever you want to call it, you're working together with a machine, we'll call it that. You're using a crude concept, which I believe you define quite well. Um the collaboration skills are more important in this environment. You're you're still coming to the fight with technical skills, but those technical skills are evolving because they have to. Yeah. But the nice thing is if you don't come to the fight, and I'm calling the fight the this new landscape. You you have to have the non-technical skills. That's what and and there's a reason I'm saying this, because that's the space I'm in. You know, we we try to coach that, and most people push back on it and go, ah, I already know how to do this. Well, you better hope you do, right? Yeah.

Jordan Hall 55:29

Because this is what's coming. In the in the past, this is two two dimensions. One was technical skills could get away with actually being relatively dysfunctional at the teamwork level, just because technical skills were relatively precious. And even worse, various forms of parasitic organization, organi organization parasitic skills, like being an effective middle manager who nobody quite knew what he did, but was good at playing the game, uh, could get away with it. And and we, you know, anybody who's been in an organization knows that once an organization gets past a relatively small number, more and more and more of the weight is actually dead weight. And that's all just all there is to it. Well, what happens is that large organizations are all going to go away. There won't be large organizations. So there is no more niche for that skill set. So then you get two cuts. One cut is technical skills become less precious and parasite capabilities become vastly less capable. And that the what's remaining then is as a primary capability, and and I like your point, your uh addition, and I won't expand on it, of teaming with the machine. So teaming with other humans and teaming with the machines. Now, teaming with the machines is a highly non-trivial skill. Uh, the the anecdote I have here is these things are world-class bullshitters, the best bullshitters ever, the best bullshit that you've ever even imagined. And if you're not capable of calling bullshit, they are more than happy to leave you, lead you down a path of days, weeks, months of bullshit that looks amazing until it actually touches reality. And then you discover to your chagrin that it was all complete nonsense. Um, and that all that worked is just best off being thrown away. Now, on the other hand, if you do have the ability to call bullshit, if you have the ability to ground what's being done and be able to distinguish what is what is, say, true or grounded real from what is bullshit, then you can take advantage of them and they become a massive lever. This is the difference between, say, a junior engineer who steps into this harness and thinks they suddenly became a superhero, and a senior engineer who, in fact, actually is becoming superpower. And that difference is gigantic. This is very important. I've noticed a lot of people who are, um, how do you say it right, greedy and naive, who think that they can say, hey, I'll just start working with this LLM and I'll you know radically improve the quality of my business and I can fire all these other people. Quite often they're wrong, completely. Quite often they will be hundreds of thousands, millions of dollars into a business plan that was actually complete nonsense from the get-go, but they just built more and they began to be more and more and more in a world of their own creation because the LLM just is a world-class bullshitter, and they couldn't ground it themselves. That's the key thing. And psychologically, it's very easy to sort of say, it seems really bad. I would, I would hate for it to not be true, what this LM is saying. And I every time I ask it, it gives me really good answers that seem really brilliant. You know what? I'm just going to accept it and just go to the next step because you feel you get that dopamine hit if I feel like I'm moving forward. And of course, it's created, giving you a string of dopamine hits for spinning a large amount of your yarn. And this is not just my experience, this is an experience I've talked to hundreds of people about. So it's an interesting bifurcation. If you are able to really ground it and you're able to call bullshit, and you're able to know where it's delivering real results and law, then you're superpowered. If you're not, then you're superpowered in the opposite direction. You know, you're going to be subject to tremendous bullshit. So teaming with the AI looks like that. And by the way, if you aren't already grounded, like I knew basically nothing about high-level mathematics two months ago. And I started working on something. And for the first month, I kept running into this problem of bullshit. But of course, what happened was is I started learning how to ground properly and started learning enough pieces adequately all the way up and down that I could actually start scaffolding. And then the AI started filling in the gaps. And now I'm actually doing decent work. Not yet great work, you can't hold that yet, but decent work. And over time, if I keep focusing on it, I'll get more capability and be able to do good work. And then maybe maybe great work. We'll have to see. So those are that that's the invariant. The invariant is teaming skills in the human, which is time and memorable. You can just read basic virtue

The pension is dead: squads, carbon bonds, and your AI agent

Jordan Hall 59:26

or military textbooks on what good squad looks like. And then teaming skills with the machine, which is completely novel. But the metaphor that I've used is it's kind of like a samurai sword that's a chainsaw.

Brian "Ponch" Rivera 59:39

Okay.

Jordan Hall 59:40

It's very, very powerful, but you really do need to learn how to wield it properly. Well, you will cut off your own.

unknown 59:46

Yeah.

Brian "Ponch" Rivera 59:47

Uh what if we could switch gears for a little bit and skip you for a few more minutes? But I'll look at uh go back to consciousness and actually ground that in the uh emerging ideas that are coming out. We brought up earlier Michael Levin's uh platonic space. Uh, there's some things coming out in from Donald Hoffman and Andrew Gallimore on the DMT side about others of or traces of others, things like that. I don't know if you've seen any of that yet. But uh, can we talk a little bit more about the we'll call it the woo side of things, which I don't think is woo. It's actually uh fascinating, the connection to religion, spirituality, uh, what we're learning about consciousness and and what we're learning from this acceleration of technology and how that applies to understanding who we actually are as humans.

Jordan Hall 1:00:24

Uh-huh. Nice. So yeah, what I would say is maybe two things, first order. One is consciousness is, which I've always found that a bit amusing because sometimes you'll get people who are materialists that will say, well, no, consciousness isn't. There is nothing more so self-evident than the fact that consciousness is. Uh, you are the experimental apparatus. If you're currently conscious, you know it. You are self-conscious. And if you're not currently conscious, while you do not know it, the fact that you were not conscious is something you were aware of the moment you're conscious again. So consciousness is a very strong premise. What it is exactly, you know, is it is somehow deeply bound to material in a certain fashion. Okay, fine, we're gonna start having a conversation. But they're just premise. Consciousness is. Second, I'm very, very confident, though not certain, that the LLMs are not conscious. Okay, so that's the second. That one's again very confident but not certain. Some people I know who are more focused on that have come to similar people who I think are quite cake have come to similar conclusions now, not just my own assessment, okay, but also other people's assessments. So I will not be speaking about the fun kind of fantasy science fiction novels about conscious AIs. What I will be talking about is the model of identity, the model of human intelligence, the model of human identity that is being shattered by the fact of what LLMs are. So one way of thinking about it is that a big piece of modernity is that it has taught us to think that being a mediocre machine, mediocre is emphasis, machine, is what it means to be a human. And the kinds of things that machines do is what intelligence is. Okay. So if you would kind of walk around a early 2000s atheist coffee meeting, they would probably nod their head as something that they wouldn't say it that way, but you'd give a bunch of examples of what intelligence is and what humans are, and they would more or less agree, which is to say that there was a weird consensus that being able to do things like remember information and process information and articulate information in novel ways is a is both intelligence and what human beings are. Now, you you mentioned 11. Uh, those who are beginning to really delve into the nature of intelligence are showing profoundly that that never was even a vaguely adequate definition, both horizontally and vertically. In other words, it didn't cover the scope of what we're going to be able to identify and it was not grounded in what what's actually happening. And LLMs push that to the limit because LLMs are not mediocre machines. LLMs are extremely good at being machines. And so, in just the same way that if I would have said in 1960, uh human beings are just things that add numbers together arithmetically, and then you gave me a calculator, you'd go, well, shit, we're screwed, because those things are much better adding numbers together arithmetically. So I guess we're out of out of business as beings. Same thing. I say, hey, everything that an LLM can do is precise as the sort of thing that is actually not the essence of being human. That's the key key point. Nothing that an LLM can do is essential to being a human. Now, there are things that LLMs do that humans do. LLMs do produce rather coherent long text and write software. Humans do that too. And historically, only humans did that. So it makes sense that we would have identified that as being something that is essentially human. But the argument I'm making is that it's not essentially human. It is a function that has never been essentially human. And to the degree to which we accidentally identified that as being essentially human, we were already making a category error, much to the detriment of ourselves. Because LLMs are coming along and are forcibly taking that territory from us, we will go through a trough and

Crew resource management and human-machine teaming

Jordan Hall 1:03:54

then a recovery. The trough will be the demoralization of, oh my gosh, I had thought that I was a human because I could do acts, I could write software, or I could do math, or I could speak English fluently, at least quasi-fluently. And no longer am I the only one that can do that. In fact, I'm out competing. I've actually lost my job as an English speaking, fluent English speaker to an LLM. Therefore, I no longer have an identity as a human. That's kind of the trough. The trough of disillusionment. Um and by the way, it seems absurd, but the reality is that will actually happen to a lot of people because most people have identified they are the value as a being, is what they do. And what they do is largely stuff the mediocre machines do, because never already needed mediocre machines into have got good ones. Okay, so that's the trough. But on the other side of the trough is the reawakening of the reality that none of that was ever essential, and that a human is a qualitatively different kind of being. And maybe a way of putting it is that they're the fluid, fluent intelligence that LMs seem to demonstrate effectively. And there is something like the soul, or if you'd like a more fancy Greek word, the noose that humans have, and the LMs categorically do not. And that difference is going to be a wonderful difference because the restoration of humans into their proper role and place of being human means, and going all the way back to the beginning of our conversation, a restoration to a more indigenous way of being human. We will actually re-thrive because we will remember who and what we actually are, because the illusion that we've been living in for the past 500 years is just going to be forcibly removed by good machines. Quick interruption here. Noose,

AI is the best bullshitter ever built: the samurai-sword chainsaw

Jordan Hall 1:05:30

N-O-U-S. And a simple way of understanding that, if you understand Khan's distinction between the phenomenal and the numeral. Yeah, yeah. The noose is an ambient, intrinsic human faculty for being directly in relationship with the numeral. That's not precise, but it's a good way of thinking about it.

Brian "Ponch" Rivera 1:05:48

Not familiar with that. And the reason I thought I brought it up, I thought it related to the the Is it Newest Fear by Big Shardan? Yeah, no, it's not.

Jordan Hall 1:05:57

Okay. My wife asked the same question three weeks ago. No. So you're familiar with that though, right? I am, yeah. Yeah. So newest fear is N-O-O. And this is N U.

Brian "Ponch" Rivera 1:06:07

Okay. Sorry for that interruption.

Jordan Hall 1:06:09

Just curious. No, I was done. So the the point there is that um this is actually a very nice. I'd like here's a conjunction called a crossover. So a crossover that I would promote heavily is the crossover of Marshall McLuhan and Boyd. Yeah. And so McLuhan has said, or perhaps his son said, I can't remember which one, digital retrieves the medieval. So modernity is very much a consequence of the printing press, the literate, which moved us away from the medieval, uh, the uh C.S. Lewis's discarded image, right? So the image in the medieval world, the medieval mind, a bunch of very interesting stuff that this had went away in the context of literacy. And digital retrieves it, meaning we'll be noticing a very substantial restoration of medieval sensibilities, let's say, and the consequence of the digital, in fact, already are. And a big part of that is wonderful because a big part of that actually restores the intrinsic value of what is human in essence, and which the machine can never compete with potato because it is not qualitatively capable of.

Brian "Ponch" Rivera 1:07:09

So we do have a lot of conversations about the connection between McLuhan and Boyd. And in effect, yeah, we there's a lot of overlap there. And then the deshred ant's work on the The phenomenon of man. Most people don't know that Boyd actually read that. And he read a lot of uh books on mysticism and and of course the Tau of Physics and things like that. So um I you've read that. It sounds like you're you're familiar with it. Okay.

Jordan Hall 1:07:31

A long time ago, like twenty.

Brian "Ponch" Rivera 1:07:33

I saw recently that even folks that are studying in psychedelic assisted therapies and how psychedelics work are starting to get into that as well as as they are looking at complex adaptive systems thinking. So so uh the neuroscientists involved, and and I'm I know you're tracking that as well, is they're looking at, hey, this complex adapted system approach to understanding how this thing works is is fits well with what they're looking at now. And that's where you get attractors, right? That's where we get it to I'm not gonna say affordances come from the same place, but uh attractors, affordances, the landscape, the fitness landscape, and things like that.

Jordan Hall 1:08:04

Yeah, that language. And just FYI, just I just remember. So for for those who don't have it, the idea of an attractor is kind of simple. Let's take a uh a funnel and let's take a small ball bearing, and I I drop the ball bearing into the funnel. I can say two things about it. One is well, three, it's going to have a trajectory, in other words, it'll hit the funnel, it'll roll around in some fashion. Second, it's very, very difficult to predict what that trajectory will look like in its particular high or low. Will it go around six times or seven times hard enough? But the third is that I can say with almost certainty that it will drop down the bottom of the funnel. The bottom of the funnel is known as the attractor. I could drop in 10,000 ball bearings, and while they will have a very large number of different trajectories, they will all collapse to a single point. The attractor of the trajectories is the bottom of the funnel. So that's a very clean visual physical metaphor that I think most of you will have within an attractor.

Brian "Ponch" Rivera 1:08:55

I I actually I used a similar I use a bowl to explain that basics, and I said, hey, look, your uh your brain is not a bowl, it's more like a ski resort inside a snow globe, right? And you shake it up, and you know, that that changes the landscape quite a bit. So you want to have you know, if you're going down in a sled or down skis, you're gonna create these um grooves uh over time, and then you're gonna start going to the same basin, the same valley, whatever it may be, right? And that's an analogy I got from uh Robin Carhart Harris, and and the attractor landscape stuff is is a fantastic way to talk about this success. Most people don't know what they are. And I we argue that they are uh if you start looking at Michael Levin's work again and Carl Frist and we asked him about this, we believe that there are in the market, there are attractors because they're harmonics. And if the market is one global conscious, if we're all communicating with each other, then it's being reflected back to us in price and time. And more we can't prove it yet, but if everybody uses some type of stochastic approach to understanding the market or a process that uses volatility and and price and all that, they're using some type of math, which is a subset of the geometry that's inside the market, the the attractor set. So the same thing is true when in strategy, I believe. We we talked about that earlier. If you're building a strategy, you've got to understand the landscape, and you have to build out what you believe the attractor landscape is based off constraints and energy, whatever it may be. But most organizations don't know to do this.

Jordan Hall 1:10:20

They they just go, well, we're a really good example is classic, right? Is the high ground. The high ground is a metaphor that we often can now use across a variety of different domains. The high ground is an attractor, strategy space, meaning most trajectories that have the high ground win. Okay, we're done. Not always, but most.

Brian "Ponch" Rivera 1:10:38

Well, we know I'm sure you're familiar with uh Dave Snowden's work, and he talks about that quite often where we use heuristics. And if you look at the Marine Corps, they use you know, take the high ground, stay in contact, keep maneuvering, right? So

Consciousness, the soul, and the noûs machines cannot touch

Brian "Ponch" Rivera 1:10:50

that's a simple way to understand how to maneuver in a complex space. And I think that's what organizations need too is one could be take the high take the high ground, maintain your customer base. I don't know, I'm just throwing things out there, but the same type of approach needs to be understood that heuristics matter. And you're starting to see this in sports too, with them with basketball and and you know, they're using we know that uh through Bernstein that uh we don't have muscle memory. Uh we have this habits of mind and and and you know the way we maneuver in space changes all the time. Um so therefore, what we need to do instead of talking about the perfect delivery of a of a kick or a shot or anything like that, we need to talk about the heuristics of uh the back of the rim and down, right? That's what you want to be thinking about. Did did that go to the back of the rim and down? And and we can guide people through that. So we're starting to see the emergence of this type of thinking in different places, but it's it's it's kind of clouded in different language. Um but to me, when when you look at it, it's all the same. It's a fractal nature of the universe that if we understand how these things work, and going back to Boyd, going back to the things you talk about, if you start to understand it, you can start to see how these scale in different domains.

Jordan Hall 1:11:52

Nice. Yeah. So just to kind of reinforce Snowden Dave, his work, uh, very fundamental work in complexity. And specifically, a distinguishing the complicated from the complex, which is helpful. And then second, where I think the prior his primary lead is how do we operate properly in the complex. So in the complicated, that's where you rules and regulations, right? Buy the book. Buy the book can work because it's a finite set of problems and you can actually find a place to look. The complex by the book is impossible in principle. So you have heuristics. And the key is to hone effective heuristics. And then by the way, you hone a meta heuristics that allows you to be able to select the proper heuristic rapidly, whole different kind of thing. Then you're going to be navigating complexity as well as you can. And by the way, it's complex. So you'll never be doing it perfectly. But if you can do it better than anybody else, you'll do it better than everybody else.

Brian "Ponch" Rivera 1:12:40

I'll tell you what, this has been fascinating. I really appreciate your time. And uh I I know we can go on for a couple more hours. There's so many more things we want to cover. Uh anything you want to share? What are you working on these days? What do you what do you what is your you're retired now? What is your life like?

Jordan Hall 1:12:53

I've been retired since 2007. So I've been almost 20 years. Well, I've been absolutely neck deep in AI. So I've got you know $1,000 a month on the different AI systems. So I have using them all and using them. I'm becoming quite skilled at understanding what the characteristic difference is, like when does this question go to and COD? When does this question go to uh Google, for example, not often, but sometimes, maybe next first guy. And then figuring out how to deploy it. So how do I actually not just learn how to use it as an instrument, but then how do I use it to produce something useful? So as I mentioned over the past few months, I've been actually doing an extremely radically deep dive in what's called architectonics, mathematics, and physics. So I'm becoming somewhat skilled as a mathematical physicist. More importantly, I've been rather skilled in using AI in the domain of mathematical physics. Very weak in mathematical physics, but I'm quite skilled in using AI in that domain. So I'm trying to grow those two. That's just an interesting domain. Like if it has the advantage that useful interesting results can actually be grounded, you can find out. And there's not a lot of people who are competing for that territory. For a while, I actually started running into pure software development because I was like, wow, I no longer have to hire software engineers. I can write it. And that's true. But so can a lot of other people. And uh, so I looked at it and said, geez, while I am building good stuff, the velocity of stuff that's coming out is so much faster than even my current ability to build it, but I'm constantly having to go meta instead of building things. I'm actually building things to see who else has built something in this domain already. Well, wow, I'm actually building things that are learning how to seek between those who are building things to see other things. But like I was writing up the stack so rapidly, I was like, that's not gonna work. I'm gonna have to go down instead of up. So I went down into mathematics and physics, which have the characteristic of being invariant, and that otherwise, I mean.

Brian "Ponch" Rivera 1:14:44

I don't have a math background, you know, enough to be dangerous, uh, some engineering background. And you get into this uh work with uh again, going back to Friston and the active inference free energy principle, you get into the stuff that's coming out of game theory, connection to um consciousness. Um I mean, it's it's pretty fascinating that it's all coming back down to math. And math is an invariant, right? It's in the universe, it's it's it's there. Yeah, and that's fascinating too. And and some of the weirder aspects are looking at like the pyramids and understanding the geometry and the harmonics that are are there and seeing that in other places. Yeah, it's cool.

Jordan Hall 1:15:15

I can see that you're a you're a rabbit holer. You've you've you've named like 16 things that are all of the rabbit hole characteristic, including uh a half moony. So let me give a last tip. This is actually a nice, useful, helpful tip. So to anybody at all, learn some math. Use AI to learn some math. Work with your LLM, pick something that you at one point you were frustrated because you didn't learn it, or were maybe can imagine being intrigued to learn it, and then learn it. Because here's the thing you say, okay, give me an example, give me differential equations. And the first time it gives it to you, you're like, I don't understand that at all. So write down this, right? Sorry, I didn't understand that at all. Let's go through a series of loops where you dumb it down until I do get it and you verify that I do, and then we'll build back up. And like literally, you'll the thing that you'll have is you'll have three good things out of it. One, you'll learn that math. You really will. Like, it's shocking how easy it is. It's not hard. It's kind of hard, but not really. Two, even really hard math is not really that hard. The hard part is understand. Once you get it, once you kind of get it, then you can start building out the capacity or capability. The second is you'll learn how to work with AI in learning new things, right? So that's that meta thing. And the third is you will get a feeling of being capable. You're like, wow, that was a win. I actually tried something with AI and it worked measurably, demonstrably. And you get a sense of what it means to be grounding in real truth, which again, mathematics is helpful because it can be proven, which then helps distinguish between bullshit and real truth with AI.

Brian "Ponch" Rivera 1:16:37

So I think that's a great I re I really appreciate it. We'll wrap it up right there. I'll keep you on for a few moments. Thanks again. All right, see ya. Bye bye.