Rich Robinson:
And we are back. Welcome everybody to the Intercognitive Podcast. My name is Rich Robinson. I'm the founding chair of the Intercognitive Foundation, with our mission and vision to make the physical world accessible to artificial intelligence. I am joined today by two luminaries in the field from this amazing company with an amazing story called "NuNet". And NuNet is going to tell us a little bit about their personal origin stories and also the origin story of the company. Please put your hands together warm welcome for, Pravar and Kabir everybody. Gentlemen, thank you very much. Welcome. I know that you're joining in. Kabir from Belgium and Pravar you're in India, I am here in the warm embrace of Mother Ubud on the island of the gods, and on my way to China soon to frolic with robots, but it's a pleasure to see you guys here. I'd love to hear, we start a little bit with you Kabir and your origin story and like your idea for NuNet and how you were bitten by a radioactive software bug that leads you into this coding superhero please

Kabir Veitas:
Hello, everybody, good to be here. Thanks for inviting me Rich. So, origin story starts from the origin, right?

Rich Robinson:
All the way from the beginning. I want to hear that beautiful accent. Where that beautiful accent is from?

Kabir Veitas:
So the beautiful accent comes from Vilnius, Lithuania. And even though I live for 15 years now in Brussels, and before that I was studying actually, I got my... 75% of my degrees I got in English, it didn't kind of eradicate my beautiful accent. So, as the origin story started in Vilnius. And maybe I will jump to my education because the first thing that I decided to do is to study business management and to be a management consultant, which at the time seemed to be a good thing to do in order to, let's say, build things rather than, well, maybe change things rather than build. And then probably so around 8 years into that or 10 years into that I understood that I actually like to build things rather than... to actually build things. And then I decided to turn my gears towards artificial intelligence. I pretty much remember that I was choosing between three fields. AI, biotechnology, well, actually genetics, biotechnology and nanotechnology as a fields that at

Rich Robinson:
What year was that?

Kabir Veitas:
Let me think. I think it was somewhere 2006-2007. So, 20 years ago.

Rich Robinson:
Yeah, so 20 years ago. So if you're choosing AI two decades ago, you're definitely seeing something we didn't see.

Kabir Veitas:
Well, so there are two things in this. One thing is seeing something. Another thing, I realized that since I was interested in computers since my pretty much childhood and I was doing all kinds of programming, hacking, Linux, things as well. Although I was working full-time job in business consulting and management consulting, that gave me just a better background to start in AI because that's related to computer science, also AI back then, 20 years ago, it was first of all, it was AI winter. There were people who were talking about the AGI but not at the level that it is now. Second, even deep learning revolution didn't have happened back then. So it was really winter and people were struggling running around and figuring out how can we kind of move along. And neural networks were kind of exotic. Exotic chapter in AI, you know, textbooks. So, yeah, so it was two things. One thing is that I was interested in the computer science and computers and generally intelligence. And that led me to actually to Complexity Science research group in Brussels. Well, actually before that I got into AI education and computer science education, but finally I got into the Complexity Science Research Group in Brussels.

Rich Robinson:
Complexity science. you can you can you double click on that?

Kabir Veitas:
Yes, well, Complexity Science studies complex systems. Complex systems are basically systems which are alive. We can very, and I think Pravar maybe can jump in because he has a vocation on another side. So basically we can split systems into control systems and complex systems. Control systems is the ones that usually we engineer and we have a clear way how to control all the parameters of the systems. Which, most of the programs, most of our physical devices are control systems because by doing them, let's say cars, by doing them in a controlled way, we can predict how they behave. So when we turn wheel right, that was right, not right, then the car goes right, not left. Now complex systems by definition, they have nonlinear behavior, which means you can do very small interruption to the system and the system starts to behave unpredictably. And usually those systems are biosphere, human body, all the living systems, which have their own identity and you cannot define a system just by putting, what is input, what is output. And so it's basically systemic thinking about these systems at large. So you try to figure out what are the general laws that govern this nonlinear behavior complex systems. How can you actually build or interact with those complex systems and not to drown in the complexity and so forth. And interestingly enough, the current large language models and AI, current revolution of AI, let's say, is built on large language models, which are neural networks, which are complex systems. And one of the problems that all the frontier labs and basically everybody who tried to do neural networks back then, mean, neural networks were invented a long time ago, what is the recent advances were based on certain ways of how to deal with neural networks. Let's say the famous Google's paper, the deep learning and so on and so forth. And then lots of compute which basically technical capabilities allowed this to happen. So people were struggling with non-linear behavior of the neural networks all the time. And the main, I believe the main kind of questions that everybody's asking is still that: "How can we make them?" And now they became super big and very complex, how can we sort of deal with those complex systems in a way that they're actually predictable? And I have to say that it's amazing to see that it can happen. So for me, when neural networks or, you know, lot of times they hallucinate or people say, why was they talking nonsense? We cannot know exactly what they're doing. It's amazing that we can say what was it. the level that we can say how they're behaving is amazing from that perspective. On the other hand, human brain is also this kind of system. Therefore, there's a lot of unpredictability there. And it's amazing

Rich Robinson:
Yes.

Kabir Veitas:
that we somehow figured out how we can, you know, live in the same space. I think I expressed my enthusiasm

Rich Robinson:
That's beautiful. So management consulting to AI to complex systems. And now you've been digging into that and all of the studies and groundwork that you laid in your career. It's all coming to bear, right? And it's a perfect timing.

Kabir Veitas:
I didn't put the last moment, but the last piece, but that last piece was also five years now, which is when I pretty much finished the studies, which was doctorate in Brussels, that's why I'm in Brussels.

Rich Robinson:
And I'm sorry, the title of your PhD thesis was.

Kabir Veitas:
The title of PhD was "Synthetic Cognitive Development," the open-ended.

Rich Robinson:
Wow. That's,

Kabir Veitas:
Yes.

Rich Robinson:
first of all, I can understand the title. You didn't make my brain explode. And secondly, that sounds like that's probably gotten more and more important over time. So it's actually, what a great choice.

Kabir Veitas:
Well, I didn't tell the subtitle. So, I hope not. So,

Rich Robinson:
Okay, now I better hold on to my skull. All right, yeah.

Kabir Veitas:
Synthetic Cognitive Development, In open-ended decentralized computing systems. So, then, pretty much immediately after we started, we co-founded NuNet. And to a large extent, NuNet is, I couldn't say now attempt. I think back then, five years ago, I would say it's attempt. I even was not connecting the things that we were doing at NuNet with the actual research. And then it started to fall into pieces probably a couple of years after that. So now I can say NuNet is basically implementing certain principles that I was researching for 8 years before. I couldn't say post-doc. Post-doc is another thing. It's more actually

Rich Robinson:
It's your postdoc. Yes. Okay.

Kabir Veitas:
going and building it, building it now and seeing how it works and making an impact. And I think this is...

Rich Robinson:
And what was that first step in the initial vision versus like what it's become now? And then let's bring Pravar in after that to see how he ties into the whole thing.

Kabir Veitas:
I think, so since I come from, I mean, at least this past, I cannot say that I come from the original story from academia because actually I come from the business world into academia, which was just the first step into academia. And then back to let's say, engineering and business. However, the premise why we started, why we founded NuNet and why we started to do, what we started to do is a premise that the compute industry is built on top of certain premises that basically primitives, computing primitives and all the technologies based on the premises that all these systems that we build are actually control system and we can control them from single point. It could be, and we can think about, let's say, Amazon. They control all the data centers, they have one login. It's a very complex system also, but it's a control system. It's a kind of way of thinking about how do we build systems. So the thing that I wanted to bring is actually the reverse perspective. Let's think about those systems as complex systems and complex system by definition. That comes from all this field of research. By definition, combination of different, more or less independent entities interacting together. So you can kind of see complex system is always a network. Always a network of something interacting, exchanging information and behaving independently or semi-independently based on this information. And I believe you can already see the network of robots interacting and robotic world, which comes from the systemic perspective. But, so that's the main thing, is I wanted to see how can we build a system or AI in terms of looking at that as a network of interacting components rather than a single monolithic something that you can control. And then since we started as a hard, let's say, hardware orchestration layer for the AI agents. That was the first impulse to start NuNet, which also from the systemic perspective is basically the same. You can, kind of, abstract all these things into the network of interacting components.

Rich Robinson:
Hmm.

Kabir Veitas:
Then we looked at what do we need to do in order to enable that. And it appeared that the technologies that are most prevalent in the compute industry now. As I said, based on the centralized kind of way of thinking about how those systems are built. And my understanding back then and now is that that was the reason why edge computing, IoT, which also are quite old terms, it didn't take off. Because the technology that we built, mean, the computer science industry built until then. was not designed for that, which made us to go, let's say, into pretty much fundamentals of deep technology development, in terms of what kind of protocols do we need to establish, what kind of security primitives do we need to build on top of those protocols, how to deal with the decentralized systems, whereas a network is insecure by definition and we cannot trust them, but every entity of the system has to establish trust by themselves. As always, or maybe not as always, is that we basically kind of stand on the shoulders of giants. Most of those concepts were invented in the 60s, 70s, 80s. actually, most of were being developed over the years, but they were not in the mainstream, because the main money were going into another direction where industry was going. However, with connectivity advances, with computing advances, with basically robotic advances, and now with AI and AI agents, these things, also computing advances, meaning the mobile phones became so, or mobile phones and computing, our desktop computers and laptops became so powerful that actually we can do reasonable stuff now in our homes, which back then, 20 years ago, so it's just impossible. You had to have our data centers, which were huge built. And that's why industry was moving there. So right now we are at a place where we can do these things, which were basically, you know, brewing for years and years and years in the computer science field. And yeah, it's amazing, amazing kind of interaction.

Rich Robinson:
Amazing. Wow. What a great journey. Yeah. Decades, decades in the industry and decades of your life. And you're kind of a overnight success story after two decades of digging into this. It's It's a great journey. And I love it. I'm glad that I double clicked on "Complexity Science", and on your doctorate you're sort of a kind of genetically engineered in a lab to be able to do this. I think. And you talk about your accent. When I speak Chinese, I have an accent. But your accent makes you even more suitable for what you just outlined. You sound like the perfect mad scientist to pull all of this together. And you're doing it. So I think I wouldn't have it any other way. It's fantastic. I love it. It's so great. Excellent. And tell us, Pravar about you hitching your chariot to this rocket ship as it's going along? What was that connection? Tell us about your background as well, please.