AI will have its own language

Welcome to another Founder Wisdom podcast. We have Arne Steemers with us today, go-to-market engineer for early-stage teams, uh, previously in sustainable energy adoption and founder of a bunch of companies including Roofshot and Brainfriend. So Arne, welcome to the podcast. Tell us a bit more about yourself, what you've been up to.

Thanks. Um, Yeah, I'm doing good. I'm lately very excited about this topic called self-organization and complexity and sort of how they are building blocks for AI. I see that, you know, like LLMs work really, really well. I think they're doing some things they don't do very well yet. I'm kind of fascinated by, yeah, the idea that things figure things out by themselves, you know. Um, they're, yeah, they're, they're like local learning rules. There's a whole bunch of concepts connected to it. I need to get into it a little bit. But, um, yeah, I'm very excited to see where the future of AI is headed. And with Braveverse, we're kind of, we're a small community, we're trying to, to discover new edges, new, new angles. Um, active, looking into active inference, spiking neural networks. Vocabularies for AI agents to communicate with each other, with all of the idea of creating a more decentralized system where, you know, the thing figures things out by itself and grows.

Right. And to start with the basics, how did you get into all of this, right? You've been a founder for a couple of years, started in energy and sustainable energy, which I guess is quite common in the Netherlands, right? What's your story?

Yeah, I mean, where did I start? I think I started with just the real drive to make an impact, and especially in energy, there's a lot to do. And at some point I just realized like that startups are, for me, a very good vehicle for impact. I mean, there are many vehicles, right? You can become a politician, you can do fundamental research, you can be an activist, but for me, I felt like startups are the right place for, for me to do it. Um, and at some point, the whole startup game kind of took over. Honestly, it's, it's, it's a game I really, really like. It's, it's starting from zero, figuring out what a product could be, who could use it, how to go to market, and, um, yeah, trying to figure those, those very early things out is just super fun, super exciting. And along the way, to be very honest, like, I, I have also been employed. I also just realized that, um, it's simply very hard for me not to work for myself or not to start a startup. I think that's— I think, I think the honest, the honest answer is also I, I just can't work for, for someone else. And, uh, I'm not saying it's a good thing, it's just— I think it's just I need to accept that fact for myself, right?

And You, you're also like a digital nomad type of guy, and, uh, last time you were telling me you don't mind like sleeping on the floor, so the hustle is real.

Yeah, I mean, um, I've been traveling around a lot. I've been camping in close to Stockholm. I've been staying in Stockholm for a bit, was sleeping on the carpet at a friend of mine in Amsterdam. Um, actually surprisingly comfortable. I slept pretty well there. Um, sometimes on sofas, sometimes with my parents. Um, like kind of every— to be very fair, it's kind of tiring right now. Like, I'm kind of done with it and I need, I need some space for myself. I'm going to get that soon. Um, but it, it's both very exciting and also very tiring, and, and I do notice that Uh, yeah, it's great to meet people, but sometimes it's also great to have like a long friend, you know, around.

Yeah, but I've seen that like in, uh, folks with your personality, my personality, Peter Levels' personality. It's, uh, digital nomadism is great to just gather a bunch of data, right, and see and make productive decisions, right? And just sticking around is a bit, uh toxic for kids of our generation, right? Like, uh, that historically we've— not sure about you, but have you grown like in suburbs, for example? Was— how was childhood for you? Is it like, uh, I don't know, boring?

Uh, I did, I did grow up in suburbs.

Yeah, playing video games, right? Sorry again, playing video games, right?

Yeah, too, too much, but also playing soccer.

Yeah, we have like very similar background. I think it's just like, uh, it's a imperative need at some point to just go around and see the world, right? And then we get tired of it. It's like all the same story arc. And then you start building, right? And then you try all these startup ideas and then one sticks out, like, which I guess is the phase you're at right now. You tested lots of startup ideas, right? Oh, and just another parallel, you start your own GTM firm to test out even more startup ideas and see which one works, right? And make some money at the same time. And eventually you find out like one good idea that's, it's like, it's my story arc right there. It's like, it's the same exact thing. What startup ideas have you tested so far and which one are promising?

One thing I started or tried to do was creating a one-stop shop for home renovation. So solar panels, heat pumps, battery, insulation, all those things in one, you know, one digital sort of almost e-commerce level shop that didn't work out. Many people tried. It was like people just didn't— they were like, it was bad at the end. I pivoted with a co-founder, with a new co-founder actually, into small construction management projects. So more on the like real construction side of things. That worked for some time. We didn't break through with that, unfortunately. And then later I tried to do, start something in food. I had, I was literally cooking for 15 people every week just to try it with, as a pilot. And I would, I cook pretty well actually, and people like my cooking generally. So I would cook for them one meal and everybody could just take that only meal. There's no option. And it was a kind of a community dinner. It was very fun, it was very fun, but like commercially not viable at all. Um, yeah, and then I've been working for different— yeah, in and out of startups, to be honest. Um, I like— I'm kind of a bit— I, I'm getting a bit like I have specific knowledge and expertise and, and interest, but to be very fair, I'm a bit lost. I do have a sort of like a bigger long-term vision, but I Currently I'm a bit lost. Where, where is the next thing? I don't know. I just, uh, I think that's okay too.

Well, AI is big, right? AI is, uh, gigantic. And my opinion's gonna be like a one-sided door. Uh, I do think most humans are not gonna make it to the other side of that door. Um, and that's, that's like your thesis, I guess. You want to work on something fundamentally big. Are you trying to be an AI research lab right now? Like, what's your main thing?

Yeah, I mean, I'm on the lookout for the right idea with the right people at the right time, right? So I personally, like, again, like this personal fascination, it's a bit abstract and philosophical, but with self-organization and complexity that has been here for years, and I have, I think I have a few real ideas there. Um, how to turn that in actual AI, I think there are a couple of directions there, but this has— like, I, I definitely haven't solved it, you know. Uh, we're experimenting, we're trying, we're having discussions. There's a bunch of researchers involved with us, we'll build prototypes. Um, but there's no breakthrough idea yet as far as I can see. But in the meantime, it's also just fun, right? It's, uh, I, I love connecting about Very technical, technical.

What's your thesis? You talked about vocabulary and agent-to-agent interaction, which I feel in 5, 10 years from now will, will be like 95% of all interactions on Earth. Um, what's, what's this about? What's your thesis about?

Yeah, so one of the researchers involved with us is, uh, created a protocol for a decentralized protocol for agents to create new words, um, to create a new vocabulary that they can all interact with. And, and it's kind of like a second brain for agents, to be honest, that facilitates agent-to-agent, the multi-agent collaboration. So, you know, like, in my head, it totally makes sense that at some point, you know, my AI will be talking to your AI. AIs will start having contracts with each other to deliver actual work because either within departments of companies, but also between companies. And within that view, I think it's kind of, regardless, I think it's kind of inevitable that AI will start creating words at some point that we don't understand anymore, start creating a language that we don't understand. So in my head, like something like this, and maybe we're onto it, is kind of going to be fundamental infrastructure for the future of AI agents that I'm pretty convinced of. And I really wonder, like, what's going to happen if you create a specific structure for those AIs creating words, because we see, you know, we see that with human language, we see how successful LLMs are. It's because they create— they use the large, very, very large body of human words and language, right, in a specific structure. So what if you start kind of collecting metadata on what words agents create and then use in specific context. Like, if you start collecting the data, then honestly, I believe you have a gold mine. So maybe I should not be sharing this with you guys, but, um, I can totally see how this is going to, you know, that could be— that could become a wave in itself, I think.

And what's the commercial model behind creating that language? Is it open source or is like—

Right now we're open source. I can share the link with you.

And how will you make money with this?

Well, we'll see. I think there's a— I mean, you can go freemium, you know, like that's an obvious example and there's different ideas for that. I think what we could do is we can optimize AI agents that already exist if we start interacting with this vocabulary. I'm pretty sure about that. We haven't tested it. Then another thing is that if this really works, then, um, I could see how at some point this will allow you to bootstrap AI agents. So, you know, like Lovable is allowing anyone to build websites with AI right now. So what if, you know, anyone can start building AI agents for their task and just make it happen? So bootstrap the harness specific for a particular task. I can see how that will work out.

And what would that language look like? And why in the first place would it be like fewer words? Would humans be able to understand? Would there be words that are not existent yet that mean a precise something to AI? Tell me a bit more.

Yeah, I mean, you know, the thing is like, we're— humans invent words all the time, right? And why do we do that? It's because there's a new, or, you know, there's a consistent pattern of behavior or a set of attitudes, ideas, uh, attributes. And at some point we're like, okay, this is, this, you know, we need to call this a new thing. Um, so I don't know, dating apps or cryptocurrency, they're relatively recent words that we invented because there was a new thing and we, we kind of wanted to, instead of saying, describing the whole thing every time, which is like 5 sentences, we want to just have a word and be like, okay, yeah, of course, all of us kind of know what that, what that means. And, you know, AI agents, they, they do different things than humans, and they operate in a different way. So, you know, they execute code, they make tool calls, they, they execute API calls, um, And so I think, you know, that's— those things are really not covered by human language. So I think it's— I think we're all like, this is going to happen very soon, that AI agents will be much more efficient and effective if those little modules, those little nuggets, will get their own words. And if you define it properly, then actually it's more, more effective and, and more certain than human language because we often disagree about definitions. But if you have a verifiable hash, um, for the— for that word with an agreed-upon, like, with a fixed definition, then, you know, then you always agree that you always communicate with definitions that, that you share. And I think that's actually a lot more We humans don't do that, right? So that's going to be interesting.

Why would OpenAI or Anthropic use you and not create their own?

I think, I mean, I think this is more of a thing for those people who create AI agents more than it is for the model developers, honestly. This is really about your AI talking to my AI.

It's like people open sourcing their model, like me running my own OpenCLAR, what?

No, you open source the vocabulary, but I think actually one of the business models we could be doing is actually allowing people to have private vocabularies that no one shares. So all the words you create within your system remain private. So because also this is, this is like knowledge, right? This is persisting, creating words and defining them properly. That is a way to persist knowledge that you acquire.

Yeah. And how are you making money on this side to finance all of this?

Well, right now we're not making money like this. This is a side project.

But you personally, how do you stay alive? Me?

I, I'm an independent go-to-market engineer. I love talking to startups and getting them, uh, getting their engine of growth running, right?

Um, how will you use that own GTM engine that you're good at building to fuel the growth of that company?

Well, we're starting, uh, we're starting a small experiment there. Um, So it's a, it's a simple, it's a pretty simple hypothesis. We have, we have, we have the, we're creating the ability for people to create their own vocabularies. Um, they're published by default. If you want to have a private one, then you, you gotta pay. That's, that's, it's a very simple hypothesis. We were in touch with a bunch of, uh, AI and agentic engineers. Um, In the meantime, trying to validate whether this is actually more efficient. So like, honestly, I think, I think, you know, we, we're not really focused on, on turning this into a money machine. We're focused on like, where is the real value? What is the real thing? How is this going to evolve? What, um, and, and, you know, if there's real value, then I'm sure there's a way to make money, but that, that's for later. That's for later, right?

And How do you picture the AI future to, to be like? Uh, do you believe in concepts like superintelligence, AGI? Have you studied, uh, this part of the game? What do you feel will happen to world economy once this hits?

Um, yeah, so I mean, I think, I think like the AI superintelligence will get there, that is more intelligent than humans and more capable than humans in, in, uh with a big difference. Um, and but I think to some, I think economic principles will still apply. It— I don't think it's necessarily true, especially if a lot of people is going to be unemployed, that human labor is going to be more expensive than robots and, and, uh, AI. Like, that, that will need to, need to be there. Um, the other thing is that I think what's going to happen is a lot of people were not going to be happy about it which is very understandable, and they will either boycott AI solutions or create, like, you know, fall back into smaller communities. Some of, like, I, that would be very scary, but, you know, those people can also be mobilized for politics and us-and-them dynamics. That's gonna, that could be very bad. I also see, you know, this is more idealistic, but I also see a world in which if a large chunk of the human population is unemployed, like, what are we going to do, right? So I think at that point in time, it's really going to be our job to be stewards of the planet and restore nature to what I call pre-David Attenborough level. So David Attenborough, the great documentary maker, he started filming and traveling the world when um, you know, large parts of the world were still untouched, and, uh, this has drastically changed during his lifetime. So once we get there, like, I think, I think the real vision is like to restore nature to pre-Davin Attenborough levels. I think that's a— apart from, apart from achieving ASI as humanity, I think that's going to be a massive, massive achievement.

And do you think humanity, uh, is gonna be left with jobs, or AIs are gonna take the piece of the pie?

Um, no, I think AI is going to take a bigger piece of the pie. Um, but you know, like, who is AI, right? So I think it's going to be— it's in the end going to be companies and people who build the AIs that kind of take the job. And, uh, that, that in itself is going to be a political question to a large extent. Um, and again, um, people— like, many people will not be happy. So we'll see where that goes, right? So I, I don't think— I, I don't think it's as simple as, okay, we have AI, it's smarter than humans. Okay, now we have robotics, it's capable of doing and moving through the world in a way that humans can do too, but in fact better. So everything is going— everything in the economy is just going to be automated by AI. I think it's— I think it's going to run into a lot of different issues.

Um, precisely, um, let's say if 10% or less than 1% of humans have jobs, you think the concept of UBI makes sense, or is, is that like the best we can come up with?

I mean, uh, it might be inevitable. Um, I think— I don't think it's the best concept we can come up with, but it's probably inevitable. It's good.

It's—

but in practice, that just means that, that most of the population becomes a lot poorer than they are right now.

Um, I think not necessarily if, like, um, because UBI would be like cash, right, for stuff that is not already provided and free. And superintelligence would provide a bunch of things that would be free and would sort of give us limited options, right? I even think that cash would become quite archaic at some point. So I don't think quality of life and I think people would get way more than what they would get nowadays, like 2, 3, 10x more by definition.

Yeah, I mean, I think, I think one, I think AI is to some extent like it's going to be a bit of a serious sum game. And, uh, I think the success of the big labs and the people who know how to build stuff with AI is it's going to create a lot of inequality, um, and not on— like, yes, it's going to create a lot of economic growth, um, but I think, I think it's also going to create a lot of inequality.

So, but what is inequality if not like, uh, people being jealous, right? I think like that's a mental thing that superintelligence could quickly solve. This is like, stop comparing yourself, right? It's like easy. If your baseline is 5, 10x what it was 5 years ago and it's, it's 5, 10, it's 100x better than what it could have ever been without like superintelligence, you could have never brought yourself to so much abundance, but then you're still comparing yourself, right? Like, is that a good, uh, or fair shot at inequality?

Um, I mean, I guess, I guess it's part of human nature to compare themselves, uh, either to who they were yesterday or to other people. Um, I'm not saying— I don't think it's, uh, I don't think it's necessarily a good thing, by the way.

But I mean, it's never good. Um, and even people nowadays focusing on inequality, right? Like, um, they're complaining about being in an age of super abundance, right? Like, 200 years ago, even 100 years ago, women could not even vote. Um, 200 years ago, couldn't have a bank account, stuff like that. And by the way, I'm not saying— I'm not focusing on women. I'm just saying, like, anyone that's complaining about their current state of things. The struggle of nowadays is not the struggle of 200 years ago, which was literally life or death, right? And I think, I think it's a mental trap, right? I think you can be poor and you can eat ramen on a daily basis but still be grateful. Of course, there's a physiological aspect to it, right? Like there's brain chemistry, but A lot of people, they keep bringing them themselves back in the hole with a constant mental loop, uh, of cortisol, right? And I think, yes, uh, you need— we need to give them access to, um, mental health, uh, care, um, and psychotherapy and these types of things. But I think, uh, humans, you know, we need to be careful. We've got like a very strong weapon up there. And even if like physiologically we're way better than we are— we ever were, and we're swimming in abundance, it's so easy to, to be like, oh, I don't have anything, right? And I'm guilty of that as well. Even like in the high spheres of things, I'm, I'm like, oh shit, this is too slow, but I have like everything, right? And what do you think about that?

Yeah, that's a good question.

Um, I, I mean, you slept on those carpet floor and you said it was comfortable. Isn't that a testimony to that?

Yeah, I mean, I mean, I mean, uh, we're grateful for that carpet. Yeah, I just, I, I do really love like moving around with my two backpacks and, um, living simply because, you know, a simple life is kind of good life. But I think, I think the sort of, um, I'm not materialistic, like I don't compare myself to others materialistically, but I do compare myself to others in, in terms of success, you know, whether I achieve— I think for other people it's experiences, like, did you go to this place? Did you experience that? So it, it's still the same human nature, but it kind of moved, you know.

Um, so you have stuff that billionaires want, right? Like billionaires, some of them are like so not flexible, uh, they're stuck, right? They're stuck in their habits and the, the mental patterns that they built for themselves, and they, they they wish sometimes that would— they would just have a backpack. And, and by the way, this is mental, right? They could literally like take a backpack. Of course, they would probably suffer backlash from their investors and all of that. But, you know, a lot of people, they create their own mental prison, their, their golden handcuffs. Arne, where can people find out more about you?

I think the best place is probably LinkedIn.

LinkedIn, Arne Steemers, right? Yeah, with two Es. Um, take care, Arne. That was Arne and Charles, uh,