Episode 217: Capoferro, AI, and the missing zero, with Dr Marc Heimann

PODBEAN  -  APPLE  -  SPOTIFY

Dr Marc Heimann is a Philosopher of Technology in Hamburg whose work formalizes the intersection of continental logic and the operational mechanics of large language models. (It’ll all become clear in the episode.)

We discuss the ethical implications of AI in enhancing human capabilities versus diminishing them, particularly for students. We also delve into the relationship between Freudian language theory and modern AI.

Despite its current limitations, there’s also a very interesting potential for AI use in historical martial arts research, where we could use it to provide new insights without modern biases.

Don't worry, there'll be plenty of swordy stuff mixed in with the AI! Marc is also a practicing historical martial artist, and we connected over Capoferro’s theory of tempo.

Books and papers mentioned in the episode:

• The History of Zero: The Nothing That Is (Kaplan, 1999).



o   Kaplan, Robert. 1999. The Nothing That Is: A Natural History of Zero. Oxford University Press.

• The Impetus Theory: Maier, Anneliese (1940). Die Impetustheorie Der Scholastik. While this isn't in English, it is widely considered by specialists to be the definitive work on the subject.


o   Maier, Anneliese. 1940. Die Impetustheorie Der Scholastik. A. Schroll & Co.

• Engineering Margins & Philosophy: Research on the "problem of margins" (Eckert et al., 2019) and the broader philosophy of engineering (Aslaksen, 2018; Boon, 2021).


o   Eckert, Claudia, Ola Isaksson, and Chris Earl. 2019. “Design Margins: A Hidden Issue in Industry.” Design Science 5. https://doi.org/10.1017/dsj.2019.7.

o   Aslaksen, Eric W. 2018. “An Engineer’s Approach to the Philosophy of Engineering.” In Philosophy of Engineering, East and West, edited by Carl Mitcham, Bocong LI, Byron Newberry, and Baichun ZHANG. Springer International Publishing.

o   Boon, Mieke. 2021. “Scientific Methodology in the Engineering Sciences.” In The Routledge Handbook of the Philosophy of Engineering, edited by Diane P. Michelfelder and Neelke Doorn. Routledge Handbooks in Philosophy. Taylor & Francis Group.

• Freud & AI: Marc’s paper that goes into much greater detail regarding transformer models and what it means for psychoanalysis. 


o   Heimann, Marc. 2026. “Freudian AI?: Transformer Models as a Proof of Concept for a Central Hypothesis in Freudian Theory.” Lacunae: APPI International Journal for Lacanian Psychoanalysis, no. 29. 

 

·         The Stillness of the Sword: Tempo as a logic of time: Marc Heimann’s paper explores the concepts of tempo and stillness in the works of 16th- and 17th-century fencing masters Fabris, Capo Ferro, and Agrippa. 

 

Transcription

Guy Windsor 

I'm here today with Dr Marc Heimann, who is a Philosopher of Technology whose work formalizes the intersection of continental logic and the operational mechanics of large language models. And yes, we will explain all of that, or rather, he will explain all of that because I didn't understand any of it either. Since completing his doctoral research on Heidegger's Historical Logic in 2020 he has focused on a humanities based framework for AI that remains technically alterable to model internal observables. Again, all of that will be explained. His current habilitation work utilizes an architectural realism account of transformer dynamics to move beyond anthropomorphic interpretations of AI. This is perhaps the least intelligible introduction I've ever done. So I'm very proud of you, Marc. We're going to give you interesting places, but most critically for this show, he is a practicing historical martial artist, and we sort of connected over Capoferro’s theory of tempo. So don't worry, there'll be plenty of swordy stuff mixed in with the AI. So without further ado, Marc, welcome to the show.

 

Marc Heimann 

Thank you for having me. It's actually very nice, nice to meet you.

 

Guy Windsor 

Just to orient everybody, whereabouts in the world are you at the moment?

 

Marc Heimann 

I'm currently in Hamburg. I'm living in Hamburg for about five years before that, I lived in Münster, which is a small university town south of Hamburg, about four hours train ride, and where I finished my PhD, very classical humanities institution, excellent.

 

Guy Windsor 

So perhaps we should maybe start everybody off with some definitions. So what is a Philosopher of Technology?

 

Marc Heimann 

It's a good question, because, as many things in philosophy, definitions are hardly fought for. So Philosopher of Technology can mean nearly everything, okay, but in my case, Philosopher of Technology means for me that I'm not just looking at technology as a sort of tool, but much more as something that interacts with who we are and what we are and how that defines us or redefines us. So this is a very old way of thinking, actually, because we find, already in Aristotle, a philosophy of technology. We find a discussion of what technology is and how it is a form of bringing forth things that is important to understand who we are, and then later in medieval times, for example, it became even more important under the shadow of theology, because once you define the Creator as creating something in a form of technological bringing forth, it influences nearly everything, right?

 

Guy Windsor 

God as technologist suddenly opens up the whole field.

 

Marc Heimann 

Yeah. Okay, this is interesting, because in antiquity, we have philosophies which are more, which are thinking more about thinking things coming to the fore on their own. And in many of your times, of course, because you have the Creator. Everything is created by a Creator. So technology takes on a much more important role then and from there on, we move to modernity. Things get more complicated, but this is to show that the idea that philosophy of technology is not like a tiny sub discipline in the large amounts of discipline that philosophy has, but it's also thinking about the very fundamental aspects of metaphysics, ontology, epistemology, and my specific take on it, is strongly influenced by what is called history of ideas. My PhD tutor supervisor was a philosophical historian, so he analyzed how philosophy over time changed, and I took that perspective on and apply it today onto AI mostly, that's the most the technology I'm mostly analyzing and to locate myself a bit more strongly, I would say there are still in the last century, there have been developing two different strands of philosophy, two different traditions of philosophy. One is the analytic philosophy, which is mostly dominant in Anglophone countries and emerged somewhat in the beginning of the last century. And the other one is usually called continental philosophy. So French philosophy, German philosophy, and nowadays it's a bit broader, and the difference between Anglophone and continental philosophy is no longer that strict, but I would say I'm more aligned with the continental tradition, and I will discuss later why that actually is interesting, once you look at specific types of technology.

 

Guy Windsor 

Okay, there's a lot to unpack there. I'm working on the assumption that the listeners have no experience of studying philosophy themselves, okay, which is not true for every listener, but it's probably true for most of them, and perhaps their most specific exposure to philosophy was when I've interviewed the philosopher, Damon Young on the show a couple of times. And Damon defined philosophy as the critique of abstractions. So basically, you have an abstract idea and you criticize it in the kind of academic sense of criticize. You know, not just throw shade on it, but, like, pick it apart, see how it works, see what could be improved, see what see what it's doing well, and just basically thinking about it out loud. So I'm, I'm thinking that a philosopher of technology may be critiquing some abstractions, but you're also going to be critiquing some concrete things, some things that people have made. Is that true?

 

Marc Heimann 

Yes, absolutely. I mean, I would probably reformulate that definition of a philosopher slightly in a different direction. Searching for problems where people do not usually find them, and that might be one difference between continental and analytic philosophy, because I heard that podcast too, and I think he's more in the analytic tradition. So I would say what I'm looking for is the problematic elements which pop up in, for example, technology use. I've been teaching in a University of Applied Sciences in the last five years, and I'm not located in a classical Institute for philosophy, but I have been working with computer scientists, with engineers and so on, and they what interested me, there were the philosophical problems that pop up in engineering that pop up in computer science.

 

Guy Windsor 

Okay, can you give us an example?

 

Marc Heimann 

Oh, one very interesting thing is margins. So in classical theory of science, you would assume that if you have a model of something and you have the thing, they should ideally be close to each other. In terms of equivalence, there should be some strong relation between them. However, in engineering, you always have sort of margins built in the uncertainties. Little like, yes, it's a bit like a slippage. But for example, if you build a machine and you build it like 100 and perfect to your to your plan, and everything is close to each other. And the thing get gets hot, it breaks down because it changes. So you have to account for these changes beforehand.

 

Guy Windsor 

So like tolerances?

 

Marc Heimann 

Yes, you have tolerances, and these are very often practical knowledge. It's a practical knowledge of the engineers they know. Okay, here, I have to be careful in that direction. Here, electrical voltage might be that high, but I don't know it beforehand, and I it makes no sense to measure it in detail, so I'm using a standard procedure. So there's a lot more indeterminateness there.

 

Guy Windsor 

Can I make an analogy with woodwork? Because woodwork is something I actually know. And you tell me if I'm describing the phenomenon that you describe as margins? 17 years ago, I made this little chest of drawers for my parents’ fancy cutlery, right? And I made it with such fine tolerances that when the drawers were empty and you closed one drawer, the air pressure inside the box would increase, and it would push one of the other drawers out slightly, right? That's called piston fit drawers, and it's like the hallmark of really precise fancy woodwork, right? I was so proud of myself, so you needed to actually put the cutlery in the drawers, so that there'll be enough inertia there that the piston fit wouldn't actually make a draw push out, because when you push a drawer in, you don't want another one to pop out, right? Yeah, okay. I was so proud of myself. Then that was made in my little workshop, sort of in a sort of subset of my salle in Helsinki. Then it gets taken to southwest Scotland, where my parents live in their drafty old, nearly 200 year old house. And of course, the moisture in the air makes the wood swell a little bit. And now my piston fit drawers aren't piston fit. They're just bloody stuck. So I had to go in take them all out, which took some effort, and then ruin all of my beautiful piston fitting and take off quite a lot of wood off the sides of the drawers and a little bit off the top, a little bit off the bottom, so that they would fit more loosely, because basically my tolerances initially were too fine. Yeah. Is that a fair analogy for what we're talking about?

 

Marc Heimann 

That is a perfect analogy. What is interesting in engineers, in difference to like classical theoretical science, is that engineers pick up these informations quite quickly, so it has a normal part of their work. And that means there that this classical idea of science is okay, we have a model and that is as close as possible to the real thing that doesn't really work for engineering. They have to think differently. And in the last 20 years, a whole philosophy of engineering were came to be because engineers realized, okay, there are philosophical problems in how we conceptualize what we're doing, and so it's less philosophers defining things very closely, and more problems popping up which needed philosopher philosophical language to even describe them properly, right?

 

Guy Windsor 

Okay, so how would an engineer come to you for help?

 

Marc Heimann 

Usually they don't, but they should. I mean, if I take an example from what I've been doing in the last years, for example, I've been doing prompt engineering with computer science students, and their philosophy actually is quite helpful, because we have a lot of theories of language, and these theories of language actually can describe quite well what happens in these models. I mean, we will come to that a bit later, but that is what I've been mostly doing. So for example, and that is a very simple example, we have the Socratic idea that I know that I don't know. So I have an idea as a human that my knowledge is limited, and this is not just something that is abstract, but it's actually very important. Like for example, if we take fencing, if someone starts describing a technique and there's something essential missing, you realize that without needing someone to point it directly out, or a more literary example, there's this famous Sherlock Holmes case where the dog didn't bark, right? So the dog didn't bark, and that's showed Sherlock Holmes that something happened, and the dog was not alarmed by whatever happened, right?

 

Guy Windsor 

So the dog knew the person doing it, yeah. The Curious Incident of the Dog in the Night Time is, I think, the name.

 

Marc Heimann 

And now the interesting thing is, large language models have no concept of absence, yeah, sure. So they can talk about absence if they have a pattern that tells them, here is nothing. But if that pattern does not exist, the model will just hallucinate something, yes, but that's the model itself. The. Sir should have an understanding of absence to manipulate this or not being manipulated to himself, because, like, if you go in to any deep expertise, you start to know at some point, usually not when you come fresh from university, but later on, you realize, okay, there's a point where I don't really know enough about but I can't be careful by avoiding the specific problem here. And these accumulation of knowledge, of absences, I would say, are an important part of expertise. Now, an expert using a large language model can see the hallucination, a student that hasn't learned these specific gaps and holes in the theories and the knowledge he uses doesn't know there is a gap, so the model will just close the gap, some association that will fix, so to say, the whole and that is, I would say, a philosophical problem, because we are not talking about things. We're talking about not being there being important for using a practical technology.

 

Guy Windsor 

Yeah, I'm trying to think of like a fencing analogy. So like if we are fencing with longsword, and the instruction in Fiore is to grab your sword handle with my left hand. It is assumed that I let go of my sword handle first. No human being would need to be told that. Right. And indeed, Fiore doesn't say, let go of your sword with your left hand and grab that. He just says, but when you're supposed to drop your sword, which is an odd thing to do, he will tell you to literally let go of your sword so you can grab both hands, right? Because it's assumed you hold on to your sword unless it's specifically said that you don't. Is that again? Am I on the right track?

 

Marc Heimann 

Yeah, yeah, okay, you're absolutely on the right track. There's a lot of knowledge we actually utilizing every day where absence plays a huge role, but we don't tend to think about it, and more importantly, we don't write about it, and only written information gets its way at some point into a large language model, for example, okay, so lot of implicit knowledge, which we only can conceptualize properly if we once take the time to think about how absences are actually active in our day to day life.

 

Guy Windsor 

I mean, just think, like reading a book without the white space on the page. You can't read it.

 

Marc Heimann 

Yeah, huh, okay, for example, there's a common joke used to explain this. It's mostly Slavoj Žižek, contemporary philosopher, who uses that when you're in a restaurant and you want to order a coffee without cream, and the waiter says, Sorry, we don't have any cream. Can I bring you a coffee without milk? Fine. So the interesting thing is, if we are strict, coffee without cream and coffee without milk are two different things, because, like for example, when the joke was written, cream was more expensive, right? So if you get a coffee without cream, you signal to the waiter that you have more money than someone who wants a coffee without milk. Okay? And these small aspects are, I would say, quite important when we think about language but in language models, they are operating in a way, but on a different level than it works with us. There are a lot of parallels how we conceptualize language in the theories I use, but these are things we actually require a conceptual language to think about them. And I think their philosophy helps, okay, to really get an idea what you're actually doing.

 

Guy Windsor 

Okay, now we are absolutely going to be diving quite deep into the whole LLM and Chat GPT license up in a little bit. But just so the people who are just here for the swords, don't like switch off and go, Oh my God, when are they going to start talking about something sword related? I know you are historical martial artists, so could you just tell us how you got into historical martial arts and what your sort of career path there has been?

 

Marc Heimann 

Oh. Um, yeah, I started three and a half years ago that was relatively close to the Corona pandemic, and I realized working on the desk and primarily in front of a computer, that the desk will ultimately crush myself. And I was in Hamburg searching for something to do against that, and rather by accident, I stumbled on Cornelius’s school.

 

Guy Windsor 

Cornelius Berthold.

 

Marc Heimann 

And he offered one free month as a test month, and I took that, realized I absolutely like the way he teaches it, because it's very focused on technique, on very philosophical. And well, shortly after that, I stayed there and started with sword and buckler, and later, half a year later, I started with rapier, and have been doing, I think, 10 hours a week of training. I had a hernia disc this December, but I'm already again at five hours a week.

 

Guy Windsor 

So okay, so does Cornelius offer that that many hours? Are you doing a lot of training at home?

 

Marc Heimann 

No, he offers a lot. 10. He offers nine hours. Three. Wait, wait, three hours of Yeah, three hours of sword and buckler, then again, three hours of sword and buckler and three hours of rapier every week, every week, yes, and there's one hour I do train at home try to get the stuff that I'm missing.

 

Guy Windsor 

And, yeah, there's two kinds of students, those that train at home and those that don't. I wish we could also say there are two kinds of students, those that progress and those that don't.

 

Marc Heimann 

Yeah, yeah. I mean, progression is very interesting, because I think I mean, prior to fencing, I only did sport in a, I would say, recreational way, so just to avoid breaking down in a way. But I think fencing gave me something I would call a physical literacy, and understanding of my own body that is vastly different of what I had before, and that alone was a very interesting experience, because I'm as a very heady way of operating normally. Getting into that was interesting, and I think it enabled me to get an understanding of learning that is quite different to what I've been doing in the last 20 years in philosophy.

 

Guy Windsor 

It's funny to think, like all people start out learning physical skills, like usually walking comes first, and then speaking is another physical skill. And once the walking and the speaking are, there various intellectual capacities develop, and, you know, and then, and then we have a person, and then at some point, they get into computer games or screens or reading or whatever it is, and then all of their attention and activity kind of migrate upwards until they are basically a brain in a bottle, right with thumbs, a brain in a bottle with thumbs. And an enormous number of my students have come to me from that perspective, and it's really interesting to see how they reinhabit their body again, because very often they move very, very badly because they're just not paying attention to anything their body is doing, unless it's saying, you know, I'm hungry. Feed me. Yeah, right. But also, like, like some people who come from, like a weightlifting background, have a very, very precise in a very specific range of motion and very, very strong but like watching them sort of take that skill of precise movement and that skill of strength, and just suddenly be able to apply it in a much broader range of situations. Again, yes, it's fascinating to watch how people basically rewire themselves when the sword sort of is dangled in front of them and. They get off, get out of whatever it was they were doing before, whether it's like a sport, like weightlifting, or whether it's playing video games or whatever it is. And it just, yeah, it connects them up properly.

 

Marc Heimann 

Yeah, absolutely, I would say that it even gave me more time to earn more active time to actually think about the other stuff I've been doing, because I feel like once my body is in, in, in that kind of movement, I'm a bit freer to do the rest of it, because, I mean, just alone. This, this, this renaissance idea of sprezzatura that you move in a way that is that looks easy and that doesn't get like cramped up and I'm liking the English words here.

 

Guy Windsor 

Yeah, it's effortless grace. I mean, is usually translated as effortless grace, and it is what every Italian martial artist or Marc, sorry, every practitioner of Italian martial arts, let me be precise, because I'm talking to a philosopher. That is, that is what we aim for, particularly in like, 16th century and earlier martial arts because the deal is public, everyone is watching. It's not enough just to, you know, stop your opponent from hitting you and whack them and they fall down. You've got to make it look good. And ideally, you've got to make it look easy.

 

Marc Heimann 

And, I mean, this is so different to our modern way of approaching work. I mean, yeah, if you go into a gym and everyone is grunting yeah and showing how hard the work is that they are doing. And I think this idea of effortless grace is something that I would say I strive to integrate it in more parts of my life, because it's it also makes you a bit more relaxed, because the body is it's difficult to be stressed when the body is relaxed.

 

Guy Windsor 

Yeah, that's a good point. Yeah. And it can help in all sorts of situations, like I am a very easily annoyed driver. And you know, I can, I can absolutely lose my cool within three seconds of getting behind the wheel, because people are just so fucking incapable of driving. Like sensible people, this is ridiculous. Like the level of driving out there is so bad. But of course, I occasionally, I'm a very bad driver myself, right? So, so on those occasions when I sort of think, well, okay, let's, let's, let's do this with some sprezzatura, it's like, yeah, you know, it's easy, it's like, and it actually makes no difference to how quickly you get there, but it makes all the difference the condition you get there in.

 

Marc Heimann 

Yeah, totally. I try to integrate that approach to my when I get reviews, because academic publishing is sometimes difficult. And yes, like with driving, you sometimes get people who simply don't understand whatever you're doing and trying to get this idea of at least making the answer sound effortless actually helps. I think. Yeah, because it's all also sort of cultural communication problem, often because I write in English, but I haven't integrated the usual ways of let's say, let's say there are differences in how I communicate normally in my northern German approach to words, because we're pretty direct, and there's no sugar coating or anything. At the moment I publish in, for example, in American journal. I need to be some what sugar coating, my approach to review.

 

Guy Windsor 

I've lived in enough countries to have learned that politeness is entirely cultural, right and when a Finn says that’s shit, do it again, what they mean is, I respect you enough to believe that this is worth further effort, and I have confidence in your ability to produce that effort and get a good result out of it. The current standard is perhaps below where it what it could attain to. But if you would like to do the further work, I'll be happy to review it again. Right, right? That's what they meant. What they said was something else. And there are, you know, I've gotten in trouble before by being a bit Finnish when I ought to have been a bit more Italian and or sometimes being a bit too English. I mean, like the English and the Americans, they we have very, very different ways of approaching everything. Like, you know, if an Englishman says, that's really not bad, right? Like, that's basically, you're probably in for a Nobel Prize, right? If an American says, that's really not bad, it's shit, he wouldn't wipe his ass with it, right? It's the same thing. But like, you know, an American instructor coming to Finland, I would always prepare them for how the Finns give feedback, right? And, you know, unfortunately, it's a visual joke. It doesn't work very well on a podcast. But imagine a student standing there completely stony faced. So I say, right, okay, if they're not enjoying the seminar, this is what they look like. And I stand there with my sword stony face. And then if they are enjoying the sword, if they are enjoying the seminar, this is what they look like. And I stand there in exactly the same position with a slight relaxation of one facial muscle, and I say, and when you've been here 10 years, you can tell the difference, right? And, yes, that sort of prepares them for the complete lack of the demonstrative feedback that, like an American class moves around a lot, right? They're constantly moving their faces and then, like, moving their arms, and they're sort of like, you know, like showing that they are engaged by moving around a lot, whereas a Finn is showing you they're engaged by standing completely still. So, yeah, it's very much a cultural thing, but your rapier stuff, I mean, we've been talking a little bit about academic journals and a little bit about rapier, and regular listeners to the show will be familiar with Cornelius, because he came on the show to discuss Capoferro’s tempo. And you have actually written an article which is a post modern reading of Capoferro and Fabris’s tempo, okay? So firstly, I mean, I did an English degree in the early 1990s I was familiar with post modernism, but for the non technical listener, if you could define post modern and then tell us something about Capoferro’s theory of stillness in motion.

 

Marc Heimann 

Absolutely, I'd start with the post modern thing, because post modernism in literature is, I think, a bit broader than post modernism and philosophy, because in philosophy, it's not that strictly defined, but I would say it's the rejection of a grand structure, like on the very fundamental way to conceptualize this is I've started my academic career, my doctorate with a work on Martin Heidegger. And Martin Heidegger is probably one of the most prominent atheist philosophers, and he's not an atheist because he rejects the idea that there's an important part of philosophy that discusses God, but he's an atheist, in a way, because he fundamentally assumes that there is no core unity of the world. There's no core structure that keeps everything together. But at the core of it, there is an upgrade, as he calls it, and if you translate it in English, which is a bit difficult, because it is usually translated as abyss, but abyss is sort of wrong, because in German, it's much more like an unprincipled or unreason, but it sounds exactly the same as abyss. So the word has a double meaning, right? And what that means is essentially that he assumes that whatever ontology, metaphysics we're creating, this is, in the end, not built on something fundamentally reliable. And I would say the last 100 years in most sciences, we have discussions of these problems, like in physics. The classical idea of 19th century physics, where everything is calculable, broke down, and for example, the early philosophical papers written by Heisenberg, for example, discuss that in detail. And so when I talk about a postmodern reading, I would say, I reject the idea that we have something like his approach to historic thinking, that is, that can be true to history, and which means, in practice, in my paper, that I explicitly say I'm interpreting them from a modern continental philosophy standpoint, and I'm trying to make that explicit. And the reason for that, and I have one example I usually give to my students, because it's so easy to grasp how this change of perspective, because we have no ultimate grounding, is actually affecting us, is mathematics. Because usually we would say mathematics is something we can rely on. It's stable. And if anything is stable, then it must be mathematics, because everything else is in some way. Let's say the art sciences are all somewhat dependent on that. However, our modern mathematics is totally dependent on zero as a number which was found or invented, depending on how you see it, in the 14th and 13th century in Europe or in the 11th century in India. But antique mathematics had no knowledge of that. So whenever we try to reconstruct Greek mathematics, we have to sort of remove our understanding of zero to even grasp closely at what they have been thinking.

 

Guy Windsor 

Sorry, just give me a second to process that, because that is what we do with interpreting historical martial arts all the time. And the problem is, zero is a big fat obvious thing that makes maths work, whereas the things that we are inserting, I'm not, I'm not talking about, you know, in a particular martial arts source text there is, you know, maybe they never actually tell you how to deal with a backhand cut, right? That can happen. And so we have to figure out how to deal with a backhand cut from other things. And so there's this thing that's missing in the text. It's not that it's that our relationship to the world is mediated through language. The way that has occurred has changed radically over hundreds of years, and when you change from one language to another, it changes automatically, or so, not automatically, but it's sort of implicit. But if we accept that what we're trying to do with, for example, Capoferro. So when I'm studying Capoferro, what I am trying to do, I will be explicit about this. I am trying to figure out what Capoferro himself meant in terms of how this thing he is describing in the text should be embodied as a physical practice. So this person is doing this physical action, that person is doing this other physical action. They're supposed to interact in this particular way with this sort of timing and this sort of measure and this sort of level of force or that sort of thing, right? So that is my goal when reading the text. But obviously it's impossible, because I hate Roland Barthes with a fiery passion, but he wasn't actually wrong in his death of the author, the death of the author essay, but so leaving Barthes out of it for a second, the question really ought to be, what is the zero that was either there and is now missing or is with us but shouldn't be? Yeah, fuck okay.

 

Marc Heimann 

You know, the moment you start to conceptualize that, and that was essentially what Heidegger did. He analyzed several ways of fundamental linguistic structuring of what he called the world, and he says it's operating on. A central metaphors, essentially, like antiquity had physics, which is the emergence from some something, and there's very little explicit conceptualization how that works. But his example is, for example, you sit in a on a around a fire in the dark, and someone emerges from the dark. So that is how things come to be in antiquity. That's the dominant metaphor, yeah.

 

Guy Windsor 

And spontaneous generation was thought to be how flies are created, right? I mean, they, they were literal about it,

 

Marc Heimann 

Yes, yes. And in in medieval times, you have then the Creator. And everything is determined by being created, except the Creator, which is the uncreated, because he didn't create himself. But so it's a linguistic idea. It's a practical idea that organizes the metaphysics. And when I read Capoferro, I try to approach it in terms of like, like, with the death of the author in mind. So I take the text and try to read the text from the standpoint I have today with explicit making that explicit, because if we don't make it explicit, we have that problem with a zero. We might push something into the text that we can't really assume being there by making that explicit, my idea is to look at a aspect of the text that is, from our perspective, coherent, even if it wasn't intended to be coherent in the way it is today.

 

Guy Windsor 

Okay, here's my question, and as someone with an English degree, I probably shouldn't ask it, but I'm going to ask it anyway, right? What is the point of a post modern reading of Capoferro, if it doesn't make you better at fencing?

 

Marc Heimann 

Oh, I think it offers a understanding of this idea of tempo that is actually applicable.

 

Guy Windsor 

Let's propose a thought experiment, right? If you took your understanding of tempo from your post modern reading of Capoferro, and you went back, and you met Capoferro, and you taught one of his students in front under the Master's very gaze. And you explained tempo in this way, and got made the actions work according to that theory of tempo, because tempo is highly applicable right to the to the hitting people side of things, right? Would Capoferro say that's exactly what I meant?

 

Marc Heimann 

He would probably call me a heretic.

 

Guy Windsor 

But, yeah, well, that's a given.

 

Marc Heimann 

I mean, I doubt it. I doubt it, because we today can conceptualize what I've called in my paper a vanishing mediator. A vanishing mediator is a Hegelian term, in a way, and it says you have something that seems to appear at some point and mediates between two states, but afterwards that mediation is gone forever. You don't realize it ever was there. And I think our language would give Capoferro a more specific vocabulary to express the problem he has, because he talks about motion and stillness. Yes, and when we see a competent fencer, we of course, don't see stillness.

 

Guy Windsor 

You know, this has been bothering me for a long time, because, I mean, in theory, yes, and when you have two people with sharp swords, there is an awful lot of stillness if they're doing rapier in that sort of context, right? Like in 2005 I think I went to the Wallace Collection with a friend of mine called Alex, who's very unreliable and a bit mad, and my friend David Edge, who was the head of conservation there, let me play with some of the rapiers in the Wallace Collection. So I put one in Alex's hand, and I knew there was a non zero chance he might actually try and stab me in the face with it, just to see what would happen. Not a high chance, maybe, like a 10th of a percent, but a non zero chance, right? Because he's unreliable, and so when I faced him holding this antique sword, and he's holding this antique sword, my body just went into a perfect Capoferro guard position like it has had never done before, right? And I didn't even know I was doing it. It was just every cell in my body was trying to get my head away from that hideous, sharp point that I couldn't really see, and to get my sword closer to this lunatic you might try to stab me. And suddenly all sorts of things sort of made sense, right? And I stepped forward, and there was a very, very clear tempo of stillness at the end of that step. And then I stepped forward again, and there was another very clear tempo of stillness. And then I backed the hell out, because, of course, we won't actually get a fence with these things, because they're, like, 400 years old or something, and not ours. And David was there being quite very sort of relaxed, actually, given what was happening right in front of his face. But yeah, so that when fencing with your friends or fencing in a tournament, there is practically none of that, right? You don't stop and wait to see what happens. You do everything in motion, but with sharps, you do actually get these Tempi of stillness if you're doing rapier, not so a side sword or long sword, but with rapier, certainly these, it's really clear.

 

Marc Heimann 

I mean, I haven't fenced with sharp rapiers before. I've fenced with I fenced 1.33 with sharps.

 

Guy Windsor 

Yeah. I mean, you're from Hamburg, you're in the sharps club.

 

Marc Heimann 

Yeah, we do that once a month. So, I mean, I would say what I found interesting, what Fabris is proceeding with resolution, because, in his proceeding with resolution essentially turns the stillness into just a moment of reduced speed, and then It moves on so and the moment you read stillness within the framework, the philosophical framework of his time, and the way he uses the terms in Italian are, I would say, linked to the academic discourse of his time. So he uses, I think, quieter. And if you read stillness in the Aristotelian way, it's a form of energeia, so it's a potential from where something can happen, which applies easily to fencing, because you have this stillness because you want to be able to act. But energeia never appears as such. So we only see that something could happen from that point, because something has happened, okay? Because if I just stand still, that is a tempo. Yeah, that's a movement. Essentially. It's a movement that is not very active, but it's still a movement. And what I found interesting in these texts was the following, like Capoferro says tempo is essentially the time it takes for me to realize an action I have planned. Yeah, and that sounds at first, like a subjective form of measurement, like we could measure it with a single fighter. And then he says, No, this is not how we use it. It only can be, probably be applied once we measure not only one fencer, but also his adversary. Yeah, and only of both movements, we have something that we can call tempo, and that makes a lot of sense, because the example I give in the paper is it's also used in the texts. If I lift my arm and I strike, this is either one tempo or two, because if my plan comes to fruition, why should I call it two?

 

Guy Windsor 

Tempi, yeah, that's fair. Yeah. And. It's also sort of worth, obviously, you must have read the Gianni los Guerra.

 

Marc Heimann 

I've read into him, but I have not used him for the paper.

 

Guy Windsor 

No, I know, I know he's not in the paper, but you must, you must have, you must have read it because he defines, he literally, explicitly refers to Aristotle physics, seven and eight, if I recall correctly, and motionless terms, whatever. And he says, like, between two guards lies a blow, and between two blows, lies a guard. And that doesn't necessarily mean that you strike, stop, strike, stop. It just means that if you're if your sword is moving in one direction, and then it changes direction to come back along the same path. It must, it must momentarily have a speed of zero, because it is going, I don't know, 10 meters per second. I'm making numbers up 10 meters per second in this direction, and then it goes 10 meters per second in the opposite direction at the point where it must slow down and then come back right now, of course, fences have ways of making that so it doesn't slow down. You do the whole thing without the sword actually coming to rest. But in theory, at least you're going to get that moment where the velocity must be zero. And so there's that sort of theoretical stillness, which I think is also something that Capoferro sort of includes in his idea.

 

Marc Heimann 

It's also quite important if we conceptualize the ideas about physics of that time, because we're not in the in the modern conceptualization of physics, but we are in the phase where the classical approaches all sort of break down. I mean, the Renaissance is one great breaking down of classical assumptions about reality. In a way, it's a highly volatile intellectual landscape in which these texts are written. And in that time, we still have the impetus theory, which, and there are absurd depictions of that. The impetus theory assumes that movement requires an impetus and a first push for in a way, in one direction. And the idea was, once this impetus is gone, the object will not have any mass pulling it in one direction, but it will go into that direction it naturally tends to. So the cannon ball, if it loses its impetus at the top of the arc it flies, it will fall down directly to the earth. It will not continue on his path. That's an idea that comes into physics only later.

 

Guy Windsor 

But they can see that cannonballs didn't do that.

 

Marc Heimann 

Yeah, doesn't matter. It's okay. We have physical books, books by written by engineers who have built cannons who use this theory. There are very interesting books about how this idea, despite seemingly being disproven by visual experiences still are active at that time. I mean, I'm not a specialist on that. I quoted the specialists in the paper so people who are interested in that can follow up on that. But the interesting thing is, it's pretty close to experiences of fencing, because that is actually close to the this assumption of tempo Capoferro uses, I have an intention. I move the sword in one direction, and it will only stop at that point when my intention, the movement I have started, comes to an end. Now the interesting part of that is now we have this thing of movements which are defined by two people, and once we define it by with two people, it makes sense to say, Okay, I lifted this the arm, and the other one poked me at that point. So maybe it wasn't one tempo, because my plan never came to fruition, sure. So we have an inter subject metric, which is what we then call tempo. But for me, the interesting thing is that the individual stillness defines the inter subjective length. So the length of a tempo is defined by the movements of two bodies. But if I am able to have a moment of stillness that allows me to act when the other one is still moving, then I retroactively, I make his one tempo two.

 

Guy Windsor 

Okay, I'm not home, so sure Capoferro would agree.

 

Marc Heimann 

Yeah, I'm probably not. Why I said reading, yeah, based in continental philosophy, yeah, but this, this retroactive redefinition of the situation, is actually something that happens in language all the time, because when you read a sentence, you only know what the sentence says at the point of end.

 

Guy Windsor 

It’s especially true in German. Yes, less true in English, but it's especially true in German.

 

Marc Heimann 

I mean, there are German jokes. Unfortunately, I don't know any good English jokes about that. But in German I have a joke.

 

Guy Windsor 

I have an English language joke about German language, if you'd like to hear it? I'll tell it in the first person, because it's more fun. So I went to this lecture, and there was this German philosopher, and he was giving the lecture in German, and honestly, my German isn't that good. So my friend who I went with, his name was Marc, lovely guy. He was going to translate for me. Okay. And so the lecturer starts talking, and he's talking, and like, five minutes go by, and I sort of nudge my friend, and he sort of, he's looking at me like, I haven't forgotten. 10 minutes go by, and I nudge my friend, and he looks at me and like he clearly hasn't forgotten. I'm like, when the hell is he actually going to tell me what this man is saying? And I nudge him a third time after about 15 minutes in, and he tells me, and he whispers, I'm waiting for the verb.

 

Marc Heimann 

I mean, if you ever read Kant in the German original.

 

Guy Windsor 

I don't read German at all.

 

Marc Heimann 

Sadly, this is actually there are sentences over page long, but there's a very simple joke and it's easy to translate. So in German, it is why Jake invited by the toad. So I can translate it in two ways into English. Two hunters hit each other in the woods, both are dead, or I can translate it, two hunters meet each other in the wood, both are dead, because in German, treffen also means hitting with the bullet or hitting or meeting someone, but that verb is only defined once I say that both are dead, right?

 

Guy Windsor 

Okay, so,

 

Marc Heimann 

and that means the end of the sentence, and even if we have a even if different languages operate differently, but even then you can, like, add for even in English, I would say you can add a sarcastic ending to it, so the whole meaning of the sentence shifts from the last you said, Yeah.

 

Guy Windsor 

I mean, friends made that really popular in the 90s when, when they started adding, not at the end of a sentence to make it kind of humorously negative, like, I really like that bloke, not. And it's like, it's a completely stupid thing, but it became like normal, and now everyone uses it in all sorts of ways. And, yeah, it's, it's, it's just inherent in the way people use language, okay? And actually, that's how we do fencing. Like if, if you and I were fencing, and let's say you hit me in the face, and we were going to figure out how you just hit me, so that we could reconstruct it, so that I could figure out what I should have done, which is, like standard fencing practice, right? We would start with the hit, yeah, yeah, and work backwards from there, because it's the end thing that defines everything that came before it. If it is an attack or a feint, we don't know until we know whether it hits or not.

 

Marc Heimann 

Yeah, exactly. And I think this is at least implied in the text, okay, if we read it with this idea of, for example, a vanishing mediator in mind. So stillness intersects, but it intersects essentially as one movement ends and the next directly starts, because if it doesn't directly start, I can't interrupt the other one. If I stand still, I'm not interrupting anyone, right? The strike will just go through. And that's sort of the what I've tried to. To make visible in this in this paper, I've tried to make visible that the way we talk about tempo in this text, or this text implies that is a metric. It's a way of measuring things. It's not a it's it allows us to measure our movements in combat, but it has this idea of retroactively redefining what and how long a tempo actually was, and this retroactive redefining, and that is why I started with the zero is, of course, something that happens constantly, sure and in a way a bout, is, if we take it from this perspective of tempo, also not just a conflict between two bodies, but also a conflict between two measurements or two matrix mattresses, because once my mattress is the definitive one in which we measure the bout I also have one.

 

Guy Windsor 

Yes, that is true. So what is the vanishing mediator in this context?

 

Marc Heimann 

Stillness? Because it's, I would say we can talk about stillness only as a energeia, so where something appears from only when something actually has appeared from it. So if I freeze in combat, that's not stillness, that's just another tempo of not doing anything. Okay, but the moment I come to the end of one motion, use this moment of not moving, which is vanishingly small, to turn to go into another movement. I have actually used this stillness there to as an moment of energy, as a moment of bringing forth another movement, which then could actually interrupt the tempo.

 

Guy Windsor 

So it’s also sort of like the square root of minus one, where you use it to solve an equation, but it doesn't actually exist. There are equations you can't solve without it, but there is no square root of minus one because there is no number that you can square to make a negative number.

 

Marc Heimann 

Yeah, or zero itself. I mean, the empty set is in every set, right? So we need it to define set theory. We need it for set theory, and it is like, like some ghostly eminence in every existing set. But on the other hand, it's the empty set.

 

Guy Windsor 

So Capoferro’s stillness as the empty set.

 

Marc Heimann 

Okay, if we assume stillness, there is the Aristotelian concept of stillness as the possibility of something to emerge.

 

Guy Windsor 

I think that's fair. I mean, really, I've read Viggiani, I've read Capoferro. I've even, obviously, in translation, read the relevant books of Aristotle's physics when I was studying Viggiani. And yeah, I think his stillness and motion thing is very much Aristotelian in that sense. Yeah, okay, oh, I mean, some more thinking. You're making me think a lot Marc. I'm not sure I like it.

 

Marc Heimann 

I mean, I think there's an influence. There's a non Aristotelian influence in there too. I also try to mark that in the paper, because at that time, we have a strong Neo Platonism coming to the fore in Italy and with Camillo Agrippa, we even have this explicitly marked in the in the pictures. He’s using him to describe his own position, where he refers to mathematics instead of the books of tradition.

 

Guy Windsor 

Yes, he's absolutely explicit in it. Just for listeners who may not be entirely familiar, he published in 1553, so that book had been around and quite influential for a long time before Capoferro probably. And the second edition of Agrippa, I think it's the Science of Arms, or the Knowledge of Arms. The second edition came out in about 1600 so I think it, I think the second edition of Agrippa predates Capoferro. So, yeah, it's very likely he would have been familiar with it, or have heard of it, at least.

 

Marc Heimann 

Yeah. And I think the strengthening of formal and mathematical thinking at that time, because that was what the plutonium was pushing in a way that might be another influence, because the way stillness is used here is, I would say, in a way, somewhat un Aristotelian, because Aristotle didn't think of two of two actors in a conflict, and for us and for offenses of that time, It allows them a metric, a form of measurement that allows us to think the conflict between two adversaries in a way that we mark something that never appears physically, but We can mark it via the formal structures of tempo, this stillness, because if it appears in reality, it's essentially a form of freezing. If it doesn't appear and the fencer acted then and only then, we actually have stillness. This stillness as inner gear, but we never see it, yeah?

 

Guy Windsor 

But it has to be there for the motion to be able to occur, yeah, it's sort of, it's almost like the chain ring of an action, yeah? Hmm, that's a lot to think about. So where can people find this paper of yours. I know where I can find it in my email inbox, but

 

Marc Heimann 

It's slated to appear in the Acta Periodica Duellatorum. Is it Acta Periodica Duellatorum?

 

Guy Windsor 

Yes, that's the name of the historical martial arts.

 

Marc Heimann 

It should be published with the Rapier issue that is slated to come out this year.

 

Guy Windsor 

This is academic publishing we're talking about. This year is actually pretty precise by academic standards when it comes to publishing. Okay, so when it does appear, just make sure you send me a link and I'll put it into the show notes at some point. I doubt it will happen before this goes out, but hopefully when it does, and then I'll also stick it, stick it out in the newsletter as well.

 

Marc Heimann 

If someone made aware of it, extremely interested, you can also always write me, and I sent the early version manuscript out. I sent it to some people already.

 

Guy Windsor 

Okay, that's very kind of you. Thank you. Okay, now, LLMS are extremely topical at the moment, and like generally speaking, this podcast is never current, right? It is. It is never, ever, ever talking about something that just happened last week. It's always sort of more evergreen in that sense. But given how much work you do with like the theory of large language models and whatnot, I thought we should probably start with the ethics of large language models, which is, there's several bits of that. There's the ethics of the use of copyrighted or other material in the training of them. There's the ethics in how people use the output of them. And of course, there's the ethics of the physical processes that that are required. So energy usage, water uses that kind of stuff. And then there's, of course, the ethics of the companies that use LLMS to bomb orphanages. So where would you like to start?

 

Marc Heimann 

I mean, I have to, in general, say I'm not an ethicist.

 

Guy Windsor 

So, and ethics is a branch of philosophy, so it's like we're talking about swords and stuff, and then we start talking about, I don't know, katanas or something. I have an opinion, but I don't do Japanese martial arts.

 

Marc Heimann 

I'm just giving out opinions because I'm planning to do a paper on the ethics, but it's not coming along soon. And then, given academic publishing that we are talking about something that might come out in two years or so, so I would say one of the things that I find too on not discussed enough. In regard to ethics is the tool character, because I would say we have, we have these big firms, but we have also, like millions of models which are free to use online, which people can essentially download and run on every gaming computer like they some of them are good, some of them are bad. Some of them are as good as Chat GPT 4, and they are free to use.

 

Guy Windsor 

So that's right, and you can, you can download them onto your local machine and they will run on your regular computer. So you don't need a great big data center. You don't need all of that.

 

Marc Heimann 

No, you only need the data center if you want to use, like, the latest state of the art thing, okay, you want this integration within, like, for example, if you have Google Gemini want to do integration in documents, and you want to use the ability to scour.

 

Guy Windsor 

I hate that stuff with a fiery passion, like I'm on a Mac and it keeps wanting to, like, improve my emails with AI and improve my I'm sorry I write fucking books. I can write a fucking email without some fucking computer telling me how to do it. Fuck off. And so I sort of I found out online how to switch it all off. And it's all switched off because when I want the opinion of an AI, I will bloody well ask for it. And I do actually use Chat GPT for things that it's good for, like helping me create templates for my syllabus wiki, something I would never I'm not a programmer, but I can tell it what it needs to do, and it creates the code, and I can put that in and have a go at it and actually it needs to be tweaked like this. And I tell it what tweaks I want, and then it creates the code, and I put it in, and it's really, really good for that, in that it doesn't replace having a computer science degree. But for my use case, it works really well. It saves me a lot of time. More importantly, actually, it saves my friends time, right? Because otherwise I would get friends who have these skills, and I would bug them for help, and it would be very tedious for everyone concerned. And this way I can just, I can, you know, Chat GPT can't get bored, so I don't have to worry about the 15th stupid little tweak that I should have thought about an hour ago. So it's fine. So it's not like I'm personally against them. It's just that I don't want them. I don't want them by default.

 

Marc Heimann 

I mean, for me, it's like one aspect of writing in another language than you were usually thinking in. Means I would have been using an editor for most of my publications if AI wasn't there. So it makes it easier to ensure that my English is on par with what I require. I generally tend to write everything in English, and I think my English is relatively good, but they're still.

 

Guy Windsor 

Your English is superb, I have to say.

 

Marc Heimann 

Thanks. But there are aspects of the language that, because I'm not a native speaker, I miss, and for that, I think it's, it's, it's a absolutely great tool, because they the things were originally built as translation machines, and that's what they are really good at. But back to the point I was trying, was trying to make. I think there's a possibility of a democratization of these language models. We don't have only the big firms. We also the big companies. We also have, like several companies in Europe, which are much more careful in the way they use your data, and which are from the ground up, setting up systems which are not like sending all their data to America, which is problematic out of several. reasons right now. So, for example, Mistral in France, they have legit which is coming close to the functionality of the big American systems. So we are no longer bound to open AI and their model, which is also already failing, in comparison to some of the other bigger firms. But I think one of the bigger problems in AI use is, or the big ethical problem I see is making that is, I would say, coming from an institution that grades students,

 

Guy Windsor 

right, yes,

 

Marc Heimann 

what I saw and what I heard from colleagues and so on, is students use AI more and more to get through their courses. And that in itself, wouldn't be a problem if the use of language and the expertise of your own disciplines language wouldn't, on the other hand, be the very tool you need to utilize your AI, right? So a student that uses AI to write his texts also de learns the ability to use AI on a higher level, and that is, for me, one of the bigger problems with the current AI use, because we are an expert. We already had the problem of gaps. An expert knows the gaps, but an expert also knows the language of his field. And with AI, it's not like it's not good to speak in common sense with an AI, it's good to speak in the thickest jargon you can actually muster in your field, because that will allow the AI to start in its internal geometry of associations at a point where it can actually utilize your can actually conceptualize your problem. So if I, as a philosopher, use the thick language of philosophy for an AI, the results are better. If I use the common sense language I use every day, the results will stay in the statistic normalcy that language also carries. So what we're currently having, and there I see really an ethical problem, is we're not training current students to actually be able to use AI, right?

 

Guy Windsor 

I mean, one, one thing that really struck me when I'm working on Domenico Angelo at the moment, and I thought it'd be convenient to have a transcription of the School of Fencing from 1787, and so I just to see what would happen. I gave it to Chat GPT to create a transcription, which it is incredibly bad at. The first five pages or so were almost flawless. Little bit of editing here or there. Other than that, it was, it was amazingly good. The next five pages were really getting rougher and rougher, and by the time we were about 1215, 20 pages in, it was flat out making shit up like it had no bearing to the page at all. But the thing is, if I wasn't referring back to the original text, I would have no way of knowing that, because what it confidently stated was, this is an accurate, like, like, publication level, precise transcription. And eventually it admitted that they can't actually do that right, but only when I kept saying, but actually, hang on, this is wrong and this is wrong and this is wrong. And again, just to see what would happen, I took a photograph of this early, mid 16th century English handwritten thing about sword stuff that's pasted into a book owned by my friend Malcolm Fare. And it's kind of gnarly English handwriting from 500 years ago, it is hard to read, and yeah, it confidently produced this absolute shite. And the one bit of that of that text that is easy to read is a list of numbers, 10, 20, 30, 40, 50, written out. Anybody can read it, it was entirely missing from the confident transcription. So I went online and I found a transcription engine that specializes in old documents with handwriting that should do a really, really good job, and I sent the same photograph in, and it confidently gave me absolute horse shit. It got some of the numbers, but not all of them, and it was just absolutely terrible. And at the same time, somebody who I listened to on a podcast, one of their listeners, wrote in with a thing about how they're writing this stuff about 17th Century Holland, and they had all these Dutch manuscripts, and they don't read Dutch, and they can't read the handwriting, so they just uploaded it to Chat GPT, and Chat GPT transcribed it all and then translated it all, and now they're basing their work on that. And I'm like, you fucking moron, I wouldn't say such a thing for the person in person, but like, like, No, that is not what these things do. And that's, I think, a large part of this problem is people don't seem to know that that's what these things do. They make stuff up. That's their job.

 

Marc Heimann 

Yeah, yeah. They do it confidently, nearly always. And I mean, if you limit through the input the LLM to a field where its training data is actually good, and then you will most likely have zero hallucinations, right?

 

Guy Windsor 

Because my wiki template thing, brilliant, amazing. And, you know, I needed to get, like this great big folder full of folders of raw image files. I wanted them all converted to JPEGs, so they'd be much smaller, and so I'd have, like, a working set that I can work with them when I decide I need to take this picture and make it, you know, do something technical with that picture. I know which like raw file to take and do stuff with, right? So I have a reference set, and the only way I need to do that was to open up a raw file and manually export it, right? And so I asked Chat what I should do, and it gave me a script to stick in terminal. I did that, and it literally it just created a new folder with all these sub folders and processed, took, like, two hours for the computer to crunch it, but created all of these JPEGs out of all of these raw files with no effort on my part at all, other than, like, writing a prompt into Chat and then pasting it into like, in its domain, it's superb,

 

Marc Heimann 

But you need to be able to identify the domain. And I think that's the problem we need at the current state. And I don't think this would change very quickly. We need an expert to use AI, not a beginner, right? And the way, currently, students are essentially losing their transition to expert, to AI. We actually have a problem, because the moment you have an expert, and I would say my own work would have been nearly impossible without AI. I mean, the time it took for me to work myself into the architecture of large language models as someone coming from a humanities right standpoint, is was certainly shortened a lot by being able to utilize AI to explain stuff, to research stuff, to make sure, with several approaches that going from several directions, I got the thing right. I mean, I'm coming from logic, so it wasn't that hard to understand the formal stuff. It wasn't that hard to understand the mathematics. And I can follow that, but I'm not able to do that fully on my own. And once you have a grasp of the foundational structures of the principles of the field you're asking for, I would say it becomes extremely strong, but you need, essentially this classical hermeneutics of the field you are analyzing to get really get an AI to do what you want. So expertise usually give that to people because, like, for example, in fencing. You know where the bullshit is, and you know where the stuff is. You actually, you're actually looking for. You don't need an AI to tailor that, but if you don't have that, then nothing of worth comes out of it. However, for example, if I give an AI and now and there, it depends on the model. If I give an AI several of my own papers, and I say, Okay, I need a simplified approach for that, for, let's say, a talk I give in front of non experts, this is usually really good work, because it operates not on the basis of pure speculation, broadly connecting philosophy to large language model, but it has very strict constraints because it has read the text. And this is, in a way, the most important thing. If you use AI, you need to be aware that your input creates the horizon in which the continuations the model produces are actually constructed. So it's not like you have a huge database, and this database is just called and everything in the database the model knows, actually it's like, I've been using the mental picture of walking in a fog, like you walk in a fog you see only very short passages in front of you. And for language model, it turns a bit like a horror story, because every step you take the fog changes the terrain right? So every token, the model works itself through changes what can come next. And this what can come next has a certain horizon. It's not fixed too closely. But this also means that, like jailbreaks, I can give a certain input and the model will suddenly ignore that it shouldn't tell me how to make motor cocktails, which means that my input is so strong that it can de stabilize the inner alignment structure of the model. So which means that the language I use to actually get the AI to do something is much more powerful than we usually attribute it to language like normally we would say, Okay, it's a form of information here. It's a form of structuring the field of possible answers which boil down so to the very simplest principle of AI is shit in, shit out, right?

 

Marc Heimann 

And a student is usually not capable of doing more than that. And that's a problem, and that is, for me, one of the ethical problems of using AI in university, school settings whatsoever.

 

Guy Windsor 

Do you have a solution for it?

 

Marc Heimann 

Well, I would say, and I actually did close reading classes with computer science students, I would say, Read, read, read, before you use AI.

 

Guy Windsor 

Okay, so your solution is basically to have the sort of students who aren't going to cheat.

 

Marc Heimann 

Essentially, but it's like working.

 

Guy Windsor 

My daughter is at university at the moment, and her view is really simple. She said that if you're the sort of student who just wants to try to pass the test and get through without actually learning anything, then AI will really, really help, right? And if you're the sort of student who really wants to study and really wants to learn everything and really understand everything, AI will really, really help, right? So you know, you can get AI to explain trigonometry, just to pick some random thing that can be quite difficult, and it will just patiently explain trigonometry over and over and over and over again until you get it right. Or, you know, my daughter's doing economics, so she might use it to, like, discuss, because she's one of the people who actually really wants to learn the stuff. And so, you know, asking it questions that can that, you know the professors would happily answer it, but you know, she's asking a question, not necessarily trust the answer, but then use that to go and look something up, or whatever, and then come back with so and what about this thing? So the mental model that Cory Doctorow uses is of a centaur, where a person who is using AI correctly is able to do a whole bunch of stuff they couldn't do on their own. Yeah, right. And so I think, like the ethics of AI, really, if it's if you're using it to create centaurs, it has it is largely positive. But there's also the reverse centaur, where the corporations want to use AI to basically make a person into an AI checking device, and so basically, that person's percentage is massively diminished. I mean, the classic standard example is AI can be trained, not LLMS, obviously, but artificial intelligence can be trained to read medical scans the radiologists would normally read. And in a perfect world, radiologists would have access to that kind of AI and use the AI to scan anomalies and whatnot. And you know, if they're not sure about something, give it to the AI to see what the AI thinks, but basically use it to make them faster and more accurate at looking at these scans, whereas what actually is going to happen in a properly capitalistic use is, okay, well, we've got we can fire 90% of our radiologists, because the AI could do most of it, and we'll just keep one to check all the output. And so you have somebody who isn't really a radiologist anymore, because all they're doing is clicking, yes, yes, yes, yes, yes, yes, yes, yes. And they're just not in a in a position to actually look at these scans, because they're just too many of them. I mean, sorry, go ahead. No. So obviously that that is wildly to my mind, that is wildly unethical. So if you're creating centaurs, basically ethical, if you're creating reverse centaurs, basically unethical,

 

Marc Heimann 

I would absolutely agree there. I think one of the problems is that we have this idea of automation, like the current, often repeated idea is that we can automate much of like bureaucracies and management whatever. Now the problem is the one person, or let's say, the team that still exists at the end that needs to be there, because AI can't operate with absences. That is, it will just hallucinate stuff. The worst point this team needs to be understanding of their field on a level that a student that now uses AI will never reach right? So if you're now a student and you're out there, think about if you really want to be employed in 10 years, because in 10 years, I think we won't need mediocre understanding of fields. Yeah, we will need expert understanding that is willing to delve into the theoretical foundations of their own field, and not the student that says, Okay, I'm here. I'm sitting here. I want to get a good grade for doing essentially nothing, because that student will not be able to use AI on a level that will be required at some point. And I think this is a problem that universities have to solve at some point, because we're just creating problems. By not training them with AI, and we're creating problems when by train by training them without AI.

 

Guy Windsor 

So I mean one, one obvious, not solution, but one obvious safeguard is exams are sat in you on university controlled computers which have all AI functions disabled, so you have to pass your exams without the use of AI, yeah. So you can use them for revision, you can use them for training. You can use them for research. You can use them for developing your understanding, or have you. But when it comes down to the actual exam, no AI allowed. So you have to actually know it and understand it and be able to explain it.

 

Marc Heimann 

Yeah. One other thing is, you can do it in, in person. I mean, oh yeah, especially in in classical humanities, I would say, if you are able to talk about, let's say Ulysses, then this will show in a direct way.

 

Guy Windsor 

I’m just thinking that PhD students have to do a viva, so a live exam, but like, if you've got 1000s of undergraduates doing engineering exams, you can't have every single one of them individually examined by I mean, it's just a it's a logistical nightmare that you'd have, like, a battery of professors whose only job was to mediate the exams, if you have to do them one at a time.

 

Marc Heimann 

Yeah, sure, that's the problem.

 

Guy Windsor 

Or you train an AI to be a really, really good examiner, and they have to do it live to the AI.

 

Marc Heimann 

That might actually work.

 

Guy Windsor 

It might actually work because the AI could be trained on the stuff the student is supposed to know and pose live questions. You just have to be a bit careful that the student can't hack the AI to answer the questions that the AI is asking.

 

Marc Heimann 

Yeah. I mean, you probably have to set a guard in the room.

 

Guy Windsor 

So, yeah, probably okay. So leaving aside the ethics for now. I mean, I have friends who will explode at us, not basically condemning every instance of AI as inherently evil. But let's just do that aside Association versus thought and the Freudian machine, to quote you, how does Freud explain Chat GPT?

 

Marc Heimann 

Actually, surprisingly good. So Freud is, today, I would say, part of a tradition of a theory of language. And here again, we have this analytic, continental philosophy divide. We have, for example, Noam Chomsky, whose assumption about language, who is on the analytic side, whose assumption of about language is language is fundamentally organized by grammar, by logic, by structural rules that organize how language functions. And this distinction is very much fundamental for this divide between continental and analytic philosophy. And on the other hand, you have authors like Freud, and Freud notably assumes in the interpretation of dreams that we have a primary process and a secondary process, and the primary process is associative. It's simply a pattern of association, a pattern a memory pattern of association, and these associations are then the foundation for the secondary process, that is the logical, grammatical speaking about things. Now this, this is a very much this fundamental assumption. And there have been several authors in the last 100 years who have taken that up, refined it, improved it and so on, like Jacques Lacan, Derrida, Martin Heidegger.

 

Guy Windsor 

I fucking hate Derrida with a fiery passion. The reason I didn't do a PhD in English is entirely because Derrida was popular back in the early 90s, and I couldn't stand him. But just to make sure that everyone is sort of on the same page, we are talking about Sigmund Freud, the psychoanalyst, yes, who he basically, sort of, he formalized the notion of talk therapy as a thing. Yeah, right, so his therapy is entirely language based, yes, and so linguists and theory theorists of language have looked at the way he is describing that language works from a psychoanalytic, analytic perspective, and applying that to linguistics more broadly.

 

Marc Heimann 

Yeah, that's fair. I would say that's what that we what we find in in Deleuze and Derrida in parts.

 

Guy Windsor 

Quick question is the is the abyss of no meaning that Derrida wanks on about at such tedious length. God, I hate that man. Is that the same abyss that Heidegger is talking about?

 

Marc Heimann 

yes and no. I mean, yes, always,

 

Guy Windsor 

When you ask a philosopher, that's the answer you get. Well, yes.

 

Marc Heimann 

People are willing to make that jump, this interpretive jump, and say that the continental tradition that we have that starts essentially with Heidegger and Freud as the influential pair that then moves to France, then I would say, yes, it's a more modern interpretation of the same abyss, okay, but this interpretation, of course, changes things. So yes, it's the same, but it's a interpretation, and it relies fundamentally on the ideas that Heidegger, the ideas that Freud brought in, but it's obviously not identical, because Derrida added stuff of his own. He focused on other things and so, so we have a tradition that starts with very reduced. I mean, I'm simplifying here, but it starts essentially with Freud and Heidegger coming to France. And Freud and Heidegger both have this assumption about the fundamental structures of language being words in association to other words. So we have not just Freud, we have also Heidegger, and both assume that logical language is a secondary phenomena of that. Now, if you if we just had Association, we wouldn't be able to do anything because, like, for example, if I talk about the bank and I do that in the context of a sure it means something completely different than when I do that in front of the financial institution. Yeah. So words need not only Association, they need also context, and they need to change according to the context.

 

Guy Windsor 

Okay, so just to be clear, what is the difference between Association and context?

 

Marc Heimann 

Well, bank is associated with a river and bank associated with money. You can use the same word. You can use the word associate to mean both context and association if you could just sort of clarify

that a little bit, I would say association is sort of the individual link. Context is the chain in which this individual link is operating. Now, one of the interesting things that Freud already marked is we don't have just these associations. We also can fundamentally change them, like when I say, hopefully not in rating anyone, lawyers are like sharks, right? I'm connecting lawyers and sharks in a way that exists already, but potentially I could also say lawyers are like flowers and you have a combination, sure, a new pattern, essentially that just makes sense in the text we are now. We are now creating in this discourse. However, these two fundamental operations like Association and metaphoric rewriting of association, we find those two structures essentially as the foundational building blocks of large language models, right? So a large language model has no understanding of a word outside of its context. It has a huge geometric representation, essentially of its learning, and in this geometric representation, it can it this it can use to represent words like, for example, if I use the word lawyer, it has a multi dimensional vector that represents this word. And this word is defined by its adjacency, but simplified to other word representations. So we have a concept of language, an approach to language where everything is only defined by its relation to other words. Every single word is defined by these relations. And these relations not only include like the common stuff, like I have lawyer and jail or whatever. Lawyer is also, for example, identified with the style of talking in a court. It's part of the definition of lawyer, and that is how, essentially, information is saved in the large language model. Very simplified, of course, now what the language model we have now can do, and that is something that is a relatively new development, is they can dynamically re contextualize. Is language like when I give a language model this sentence, lawyers are like sharks, it will actually take the word shark and reposition it in its active representation of this text closer to lawyer, because it recreates essentially, it creates a new word this exists only within the what computer scientists call residual stream of the current chat, so the input that is given to the model, but it allows the model to be extremely dynamic in the way it operates with language. And this is why stuff like jailbreaks work. I can redefine adjacency in the geometric representation of language, and thus move my input in a way that the common pattern that would allow the model to not answer my question are just avoid it. You have like, these are not three dimensional. But if I say there's one edge of three dimensional space where the information lives, and normally I start in the middle of that. I can, with the proper language, move that starting point closer, and the point where the avoidance happens is closer to the middle. So that is the technical reality. And now these models operate then, essentially with the two fundamental principles of language that we already find in Freud. They have Association, the relation of one word to another, and the this relation is important for choosing the next word, and these relations can be rewritten dynamically. They are not rewritten by an intent of the model. They are rewritten by the structure of the text that it reads. So it's not like the model actively rewrites, but I write in a way that these are adjacent or different to each other, and then this will cause the model to do that, like when I say lawyers are like sharks. Then this like sharks. This like in the middle, this star changes things. This changes things, yeah.

 

Guy Windsor 

Whereas if you said lawyers are sharks, then that would be slightly different.

 

Marc Heimann 

Probably comparable, but slightly different, yeah. And then the next important thing, and then that is also existing in this theory of language. But I just, we just had this joke about things being defined from the end. Now, a language model operates like that. It doesn't know what it's writing before it has written what it's writing.

 

Guy Windsor 

Neither do I, honestly, I'm trying to find out what I'm thinking.

 

Marc Heimann 

Yeah, me, me too. I've never written with a plan before.

 

Guy Windsor 

So, yeah, I said when I'm writing a general idea, yeah, I have a general idea of the book, and I start writing it, and at some point I get stuck, and when I get stuck, I create a plan, and that gets me unstuck. And then I write for a while, and I realize that I've abandoned the plan, and so I then I get stuck again, and then I reconstruct the plan to be closer to reality, and that gets me going again. And so there's this constant sort of referring to the plan, but the plan changes when I'm actually writing. And so at the end of the book, I'll write the introduction to the book, because now I know what's actually in the book. Yeah, yeah. I think

 

Marc Heimann 

That's normal. Yeah. I think so too. I've never written anything in differently.

 

Guy Windsor 

I do. I do know people who will literally plan the whole thing out, okay, and then write it. I mean, in non fiction and in fiction, Saul Bellow, not Saul Bellow, not so the other one, Lord of the Rings, Chappie, not Lord of the Rings. Ah, my God, Lord of the Flies. What's gone with my brain? Oh, we philosophy for nearly two hours. So my brain is fried. William Golding, William Golding, that's the chap. Anyway. When he got a Nobel Prize for Literature, people asked him a bunch of stuff about literature and whatnot, and he famously described he would plan his novels down to the last flicker of an eyebrow. And, you know, I know of people who will write 100 page outlines for their 300 page novel, right? And they work it all out in advance. I mean, people do that with woodwork. They will, they will create this absolutely, like, accurate plan of the thing they're going to make. And then they will make everything to the point that they can literally just take all the dimensions off the plan and make the piece of wood. And that's not me at all. I am, you know, I start, I start with the general, like, sizes that I need, and then I sort of make the first piece of wood, and then everything else is sort of related to that first piece of wood. And it's all, you know, I'll actually measure something, you know, like maybe five times in the whole process of putting something together.

 

Marc Heimann 

Yeah, I usually start with a problem I have try to get it into words, and from there, the rest emerges, right.

 

Guy Windsor 

Okay, so Freud and Chat GPT, so what you're basically saying is the way large language models work is closely related to how Freud described language. Do you think that that in has any bearing on whether Freud was actually accurate when it comes to describing human language?

 

Marc Heimann 

I mean, famously, Freud was criticized for conceptualizing, for example, repression, and he was called completely unscientific in conceptualizing this idea, and funnily, large language models, repress, exactly like Freud said, like there's an association that blocks the access to what you want to say, and this association sits in front of it like that is ethic refusals. It's point by point, like Freud describes in the psychopathology of everyday life, so no trauma. This is exactly what he describes as repression. So I wouldn't go so far as to say they prove Freud, but I would go so far as to say they demonstrate under engineering constraints that what he describes is not just scientific, but very much possible.

 

Guy Windsor 

Okay, that's a very interesting and clear way to put it.

 

Marc Heimann 

Yeah, okay, so there might still be aspects of Freud which are wrong. I'm not disagreeing there.

 

Guy Windsor 

Oh, I think, I think a lot of his psychotherapy stuff is nonsense, but that's just maybe maybe.

 

Marc Heimann 

But the point is, his description of language is actionable, right?

 

Guy Windsor 

Actually, as a historical martial artist, that is more useful to me than being actually sort of externally empirically Correct.

 

Marc Heimann 

Yeah. I mean, it is empirically correct because it is actionable, right?

 

Guy Windsor 

Yeah, okay, but so, but like, Okay, well, in fencing, you can do things that are theoretically wrong but they work, and you can do things that are theoretically correct and they fail. But the like, a necessary, but not sufficient condition of an interpretation being accepted is, does it work at speed under pressure? Because the assumption is, if it doesn't work at speed under pressure, it can't be what the author intended. Yeah, so how, how far do you think these large language models can actually help in the interpretation of historical manuals?

 

Marc Heimann 

Well, historical manuals have one major difficulties here, and only one, yeah, I mean for the statistical analysis that is basically of the basis of that is that they are usually singular, and they are probably not very High in importance to the big corporations that currently train AI models. So if we had a cheaper way of training a model, and we can feed it the whole victim hour and retrain it on that, we might be able to get a model that is more. Competent at helping us think about martial arts, because it might be able to show us links that we might not see individually. However, this that still requires deep expertise, because we're talking about a text that describes bodily movement, and a large language model has no understanding of bodily movement outside of language. There are studies in robotics which show that even the purely linguistic understanding of language models is still extremely powerful in controlling robots in 3d space. Okay, so language holds a lot of information about movement, perhaps information about movement that without language we couldn't conceptualize, like stillness. Okay, but I would say the same problem that we applied essentially in when we discuss ethics, is we still need experts. So I would be very careful, and would say I would not advise using an LLM to teach you something that you never tried before.

 

Guy Windsor 

I that's one thing I was I was thinking of, is if I created transcriptions so machine readable versions of, for example, all the smallsword texts we have, I could get an LLM to give me a list of all the techniques that are used in smallsword organized by how commonly they're referred to in the text I could get it to so that would give me an idea of how historical fencing masters in the period how much they expected these things to occur, perhaps. And I mean, it would be maybe possible to figure out, like, how things changed over 100 years, if we're sticking within the same system. At the moment, my head is entirely full of 18th century small sword stuff, and that has the pro and the con of most small sword treatises are very, very similar to each other, from de la Touche in 1670 to the School of Fencing in 1787, for over 100 years. If you looked at the content of these manuals, there are some outliers. Sir William Hope's New Short and Easy Method of Fencing would be one, which is basically a different system altogether, but generally speaking, they're all doing French smallsword, and they all are all using the same sort of language. They're describing the same sort of actions. They are functionally describing the same system, which means that we have multiple windows onto a probably relatively common body of fencing theory and fencing movement. So, like, you know, uploading all the smallsword treatises to an LLM and getting it to do some textual analysis might be useful.

 

Marc Heimann 

I mean, what could be interesting is taking all that smallswprd data and all textual data we have up to that point, and nothing that is what was written later, and see if a LLM that exists essentially as a representation of the intellectual discourse of that time and not of our time, could tell us something different about fencing, yeah, solving the zero problem, not completely, right?

 

Guy Windsor 

Yes, it says so, basically making sure that it doesn't know about anything after about 1790 and actually, like fencing masters in 1750 didn't know anything about Capoferro or Marozzo, still less, Fiore, right? So, actually, that, my God, that's a really interesting thing. So we could basically create an artificial researcher that doesn't have our modern biases, yeah, wow. Oh, that is interesting. Maybe I need to talk to those Mistral people, because, I mean, this stuff is French. I mean, smallsword is a French system, and I'm sure their large language model speaks excellent French. So maybe I wouldn't even have to translate it all. I could just stick it in in the near in in the original French.

 

Marc Heimann 

Yeah. I mean, the moment training these models gets cheaper and cheaper and it's already getting cheaper, we might think about essentially creating models that are pictures of a certain era, right? So there is some research potential there, and I think even for historical martial arts, because it's conceivable at least, that a model, let's say that is only trained to 1600 and has only the ideas and concepts of that time might allow us to read Capoferro differently. Yeah, so, and that is something that is incredibly hard to replicate as a human, yeah, exactly.

 

Guy Windsor 

So that's actually maybe what they could be most useful for, for a historical martial arts researcher, is, is artificially simulating ignorance of everything that happened after the period in question. So we don't see it through a modern lens. We see it through a contemporary one. Yeah. And you know, my younger daughter still lives at home, thankfully, and she likes watching certain TV programs. We watched this new Young Sherlock thing on Netflix, or whatever it is, Prime maybe, and it's set in like when Sherlock Holmes was young, so maybe 1860 or something like that. And the physical culture stuff, you know, the clothing and the technology and everything was pretty good, but the characters keep using modern terms of phrase and modern ideas in a way that somebody in 1860 I mean, that's not even my period, but I've read enough 19th century literature to know that people like that could not have said that, yeah, because it wouldn't have meant anything to them, right? It would have been like meaningless babble. Yeah, there they are just saying it, you know, and then, and then acting on these things which they couldn't possibly conceptualize. And it drives me quietly nuts, quietly because, because I am not going to, yeah, my daughter's young and, you know. And I looked at the positive, and I was like, Oh, I really like the clothes and, oh, I really like, you know how they're doing, this, that, and the other, and, you know, I was as positive about it as I possibly could be. And kept, kept all of that kind of, what's the word? How we know what we know, ontological stuff to myself, right? I just left that out of it because it was my own private little torment. But I love the idea of getting the AI to not know anything after a certain date and that and make its analysis based on from that perspective,

 

Marc Heimann 

yeah, I think this is a potential field of social analysis that we haven't even breached yet. Yeah, and it could. I mean, for example, think of an AI that only knows what someone in Japan in 1800 knows.

 

Guy Windsor 

Wow, yeah

 

Marc Heimann 

The potential of talking to a simulated subject of that time is suddenly there. Yeah, it wouldn't be perfect, because it would act essentially as a very, very literate.

 

Guy Windsor 

Yes, and it's also, it's, it's unverifiable, because you can't dig somebody up and have a similar conversation with them and go, Sure. And of course, people vary hugely. So maybe the person you dug up is the wrong person to dig up. You should have dug up the person to the left of them, they would have actually given you. Actually given you the exact same conversation.

 

Marc Heimann 

No way to know, yeah, I mean, it's an abstraction. It's perhaps a form of computational necromancy.

 

Guy Windsor 

I love it.

 

Marc Heimann 

But I think it's, it's at least a worthwhile idea to for historical martial arts, because, oh yes, having even a simulate, I mean, imagine a someone. We're really really deeply into Italian intellectual life of that time to answer questions about the concepts in Capoferro and Salvatore Fabrice, yeah.

 

Guy Windsor 

I mean, there's plenty of medieval philosophy and medieval sources, so we could train somebody up as a medieval scholar, you know, with only medieval knowledge, and ask them, What does Fiore mean when he says, volta stabile, yeah, right. I would happily pay money for that. Oh, my, yeah. Okay, all right. I have a couple of questions that I ask most of my guests. The first is, what is the best idea you haven't acted on yet?

 

Marc Heimann 

Well, I think the best idea we probably just discussed. I've been thinking about models, which are replicants of a specific era, and I haven't have neither the money nor the team to realize that okay and well, the second best is probably the I want to write a book about violence.

 

Guy Windsor 

Okay, from what?

 

Marc Heimann 

Well, somewhat influenced by my fencing, because it gave me a different approach to it. But there's also a long tradition of thinking about violence in philosophy, like starting with Plato, the cave isn't something you leave freely, but you're dragged out of it, right? And we have the idea in German idealism that systems, of these complex systems, have a sort of foundational violence that starts them. The master slave relation in Hegel and Fichte has something that he calls a declaration of reason, which is in German, even the Mach Sprech. So the Power Word, or something like that. And I do wonder if the concept that I am discussing in the text about tempo has some further openings for philosophy, but this is an unrealized book, and as I just told you, I don't plan my books before. It might never appear beyond an initial paper or something like that.

 

Guy Windsor 

You might wake up at five o'clock one morning with the book knocking in the back of your head and not letting you sleep until you get the first idea of the first chapter written down. That's what happens to me.

 

Marc Heimann 

Usually, yeah, so usually, right now, it's these ideas. I create a paper out of them and send that out, and then return to the same thought, because I, I usually don't have the time to write a book at the moment, but papers are also, well, also,

 

Guy Windsor 

Yeah, and papers are academic currency, in a way that books are not.

 

Marc Heimann 

I mean, humanities, the book is still worth something, but,

 

Guy Windsor 

But, yes, a paper in a prestigious journal is worth more than a book published by a regular publisher, yeah, for sure. Okay. And again, my last normal question is, somebody gives you a million dollars to spend improving historical martial arts or philosophy worldwide, given the context of our conversation, I think you might spend that money on creating this large language model that is limited to a Am I right in thinking that's where you'd go with it, or would you? Would you do something else?

 

Marc Heimann 

I think it would be not enough.

 

Guy Windsor 

So it's imaginary money you can have as much as you want.

 

Marc Heimann 

Now I’m thinking, I've been thinking about that question, because I think one of the aspects of martial arts that I really enjoy is the details. For example, manipulating a sword with your fingers instead of your wrist, was for me, a considerable step forward in my fencing. And I think these small details are something that we in the way our society is organized, usually don't have the time to think about so creating a stipend where people can just take one month to improve a tiny bit of their stance, or a their grip, or the way they strike, and afterwards write a paper about it. That would be quite reasonable with a million dollars, yeah, would actually think, I think help historical fencing.

 

Guy Windsor 

I mean, I do it all the time, and I write blog posts and I write books and whatnot, and I will literally spend, you know, a month, two months, six months, a year, however long it takes on something like grip or the position of the back foot, or, you know how exactly you extend your arm, and you know which finger you're intending through when you go into postal longer and, yeah, all that sort of stuff. I mean, that's that is exactly my jam.

 

Marc Heimann 

Yeah, I think these details that that would be something I would want to support.

 

Guy Windsor 

Okay, because so, so if I gave you a million dollars for that, you give me some of it back to do some research. Yeah? Okay.

 

Marc Heimann 

I think this, this, this physical, physiological literacy of understanding these small details is something that I find incredibly fascinating in fencing. And I think that that, that is what I would support.

 

Guy Windsor 

Excellent, okay, yeah. And again, it part of the difficulty is we don't really have a good language for it, yeah, because the language of medical anatomy is inadequate the man the language of sort of scientific movement, so kinetics and whatnot is too abstract. And then you get into things like the language of how Tai Chi people talk about intention. And then you have how weightlifters talk about focus. And you have, yeah, it's, it's, there are lots of different languages from lots of different physical activity subcultures, and just creating a common language so we know what we're talking about when we talk about tailbone position would be helpful, absolutely.

 

Marc Heimann 

I mean, we need essentially, our own jargon.

 

Guy Windsor 

Maybe we could get an LLM to produce it for us.

 

Marc Heimann 

I'm not sure.

 

Guy Windsor 

It's going to, because as soon as you start asking it to invent stuff, it invents, slop, shit, rubbish, and everyone would instinctively hate it, because it would just be completely wrong.

 

Marc Heimann 

Yeah. And if something good comes out of it, something good have must have gone into it. So you're right, two pages of an input, and then it produces, on the basis of that input, something reliable. So I think there can be, it could be a help, but you would have to control your language in detail. Like, that's what I mean. That's what I told earlier. The control of language is report important for the output. So, minimal inputs leads to slop. And if you really think something through it, I think it can help. It could help. Yeah, this is also a job that is without existing, reliable patterns, because you want to recreate, you want to recreate, essentially a new way of linking existing words in a way that they mean something different, like when I like an engineer talking about margins means something else, like a historian talking about the margins of books, right? So we need to redefine language to get a language into historical fencing that is able to describe that. I mean perhaps calling the tale in something else is.

 

Guy Windsor 

That's, that's the easy bit, though, is then getting everyone else to adopt it. Yeah, that's the tricky bit. Brilliant. Well, thanks so much for joining me today, Marc. It has been educational to me.

 

Marc Heimann 

Well, thank you for having me here. It was really a joy to talk about that, and I think we got pretty close to the bar type talk.

 

Guy Windsor 

Yeah, I think we did do.