AIAW Podcast

E152 - Empowering Human Autonomy Through Personal Digital Twins - Ingo Paas

• Hyperight • Season 10 • Episode 9

Share episode

In Episode 152 of the AIAW Podcast, we delve into the dynamic intersection of digital transformation, innovation, and AI with award-winning leader, author, and CIO at Green Cargo, Ingo Paas. Drawing on a career that spans retail, sports, and logistics, Ingo shares strategies for building Digital Composable Enterprises, bridging legacy systems with cutting-edge AI, and harnessing horizontal and vertical innovation for holistic growth. He also offers insights into scalable investment approaches, the disruptive technologies set to reshape industries in the coming years, and the ethical implications of an AI-driven world. We then explore assistive technology and the transformative concept of personal digital twins—AI extensions of individuals that embody their values and preferences, functioning as intermediaries between humans and machine AI. Originating from Ingo’s efforts to support those with severe disabilities, this approach offers a holistic alternative to fragmented assistive solutions and highlights the importance of principle-driven AI systems that keep humans in control. By emphasizing interoperability, robustness, and modularity, composable architectures enable resilient, evolutionary investments in AI. Ultimately, Ingo envisions a future where AI empowers autonomy rather than restricts it, and shares a glimpse of his upcoming book, Empowering Human Autonomy, revealing how semantic, generative, and agentic AI capabilities can unlock new levels of accessibility for individuals and businesses alike. 

Follow us on youtube: https://www.youtube.com/@aiawpodcast

Ingo Paas:

The power you give to the AI that is with you personally and the power that is with the AI of a machine. So I'm working and I've described this in my new book as a concept to make a difference between a human robotic entity which is autonomous and has an AI system to be autonomous and to do the physical stuff and knows about his environment or its environment and can make all of these decisions, but should never, ever, come into a situation making a decision for you that will personally affect you. This is where you put in a governance level where the agent tells the robot, as a digital twin of yourself, this is what I want you to do, and the governance is so strong that the robot will never, ever harm you so keeping the control of the machine in some sense, making sure that we can keep control of it so my idea and what I described is so you give everyone a digital twin, which I thought this was true here.

Ingo Paas:

Yeah, so the digital twin is with you. The digital twin is based on the AI capabilities around you and the robot is just an agent and you really can control the robot with the AI system. That is your digital twin, so the robot will never, ever take actions against you that is not approved by the digital twin and that gives you a unique power individually.

Henrik Göthberg:

So, if I get it right, the distinction here is that this is what you highlighted the AI for the human as the digital twin taking care of your objectives, versus the AI inside the machine, which is a robotic entity. And here now, the way we can exert human governance on the robot is through our digital twin.

Ingo Paas:

Exactly, and the digital twin is actually based on your values. It's based on your experiences, where you are, where you come from, what you do, what you've learned, what you've seen, and it will really make sure that the AI systems connect in a very good way.

Henrik Göthberg:

I really love this I haven't thought it through in the way you have, Ingo, but I really love this that we need the avatar as a way to transcend into the machine world in order to steer them. Yes, we can't steer it as a human.

Ingo Paas:

Yeah, but I would be very clear that it is not an avatar, because the avatar is more representing a physical entity, while the digital twin is the digital version of yourself.

Henrik Göthberg:

Yeah, so this is a distinction we should actually make. Yes, it's very important yeah good, I love it.

Anders Arpteg:

So it's more a personal digital twin. So the twin is literally of you personally, not of the twin is literally of you personally, yes, of the factory, not of a society or a city, but actually of you personally. And by actually having that control what the other ai systems are doing, you can still keep yourself for the safe right, would you agree?

Ingo Paas:

yes, it's exactly what you say, anderson. I think that my idea was growing while I was writing the book and was thinking about human autonomy for people with severe disabilities and to help them with humanistic technologies, so I built this concept, but then I realized that, in fact, the more AI we're going to see in the world, the more AI you need to protect yourself individually, so that the AI is connecting to all of the agents that will exist around you. So I'm seeing this as a very visionary approach to protect individuals. I think I haven't seen or heard this.

Anders Arpteg:

No, I haven't either.

Henrik Göthberg:

It's a very noble and good idea, but the crazy thing is, I can draw a metaphor and I can draw analogy to what you're doing to what we are doing and working on with Scania in the simple topic how do you enforce policy as computational guardrails and governance? How do we enforce governance? And what we are now working on on a data level not at all shows a glimpse of the same thinking. So, in order to have governance of data even for GDPR stupid idea like this we need to have a way that we are codifying the policy. Policy is code. Now, in practice, then, it means you build data sets and you build data in such a way so you can have that embedded in the data, so the data cannot be accessed and cannot be used in certain ways.

Henrik Göthberg:

And fundamentally, what you're doing? You're taking computational governance to the extreme philosophical level. How do we have computational governance of AI? From my personal humanistic point of view, exactly that's what you're doing. So you're applying the same concept as we are, but you're moving into. How can I have governance as humans?

Anders Arpteg:

I'm a policy and Technically it works. It's not really policy as code for a whole company, then it should really be policy as a code, but for humans, and that's the big difference.

Henrik Göthberg:

Right, this is huge. I was just making the analogy that technically I get it, I get it, so it's completely feasible.

Ingo Paas:

Yes, and I think the more we see reasoning AI, the more reasoning AI can perform those tasks without you making any decisions about it and protecting your integrity.

Ingo Paas:

And also you can create a network of digital twins which could be your family, so you could not control your family, but you could make sure that you have this level of protection also on a group level.

Ingo Paas:

So my idea when I was thinking about you know this specific experience from my own life with my son. I was born 22 years ago and he was born with a syndrome that is very rare, so 180 million people is born with this challenge. So he is on a scale from 1 to 10 at zero on his cognitive abilities, his physical abilities and his social abilities, so he's very much depending 24-7 on support. So my idea was to use technology and to find a different way to create assistive capabilities around individuals that make the individual stronger and would protect the individual far beyond the time of the caregiver's existence, and that is one of the biggest problems in life. So I want everyone to be kind to my son and I want everyone to be kind to my son when I'm not here anymore. I know that I'm aging, so my situation is like everyone else. The situation is not getting better over the years, and the day will come where I will not be there for him anymore. So my energy went into this thinking for 10 years.

Anders Arpteg:

So it's not something that has developed overnight, but, um, it is actually a very strong mission for me personally, which goes beyond my um existence I like it as well because you know, one of the things I've been having is you know, are you afraid about the upcoming agi you know time that we're probably going to see? You know I'm more afraid about the upcoming AGI time that we're probably going to see. I'm more afraid about the stupid AI we have today, but in the control of bad humans, than I am of having other AI systems controlling other AI systems potentially, if we have some supervision of them, and I think in some way you're saying that we need to have a digital twin of machines that protects the specific human.

Ingo Paas:

Yes.

Henrik Göthberg:

And that's the only sustainable solution, and I certainly think that's true, I haven't really gone down this angle of thought, but the fundamental topics how do we exert our human values and our human things in a digital world? And we need that proxy, we need that digital twin in order to be part of the machine ecosystem, so to speak. Exert our power, our governance.

Ingo Paas:

It's very humanistic, so it's very essential.

Anders Arpteg:

It's very essential as well, such a great opening discussion here. And before we get even further into this upcoming very interesting discussions, I would like to just welcome you here, ingo Pass, and you are the CIO and CDO right of Cargo Green, green, cargo, green Cargo.

Ingo Paas:

Thank you. Sometimes we drive backwards, yeah, yeah.

Henrik Göthberg:

So the trains go backwards and you cargo green when they come by.

Anders Arpteg:

But you're also famous. You have written books and you have award-winning prizes, CEO award yeah right. And, yeah, very much looking forward to upcoming discussions here. But before we go into all of these kind of discussions, perhaps you could just briefly describe a bit who are you? Who is Ingo Pass?

Ingo Paas:

If you would ask my wife, I think she would say he's an empathic idiot. I was an empathic idiot. So yeah, I'm very much in a situation in life where I haven't planned to be, so that I live in Sweden was not part of my plans.

Henrik Göthberg:

Where are you?

Ingo Paas:

originally from. I'm from Germany, which part? Dusseldorf, dusseldorf, yeah. So I actually ended up after studying. I've done a lot of different things before I went to university and I was probably the oldest one sitting there and had no clue about what the others were learning. So and I found out that all of these civil engineers in Germany they're good, so they build great cars. I decided to not be part of this group because competing with them was difficult, so I ended up in a different business. So I joined Ericsson, and that was exactly in 1994 when the mobile Klondike rush started.

Henrik Göthberg:

Can I from wireline to wireless radio Exactly?

Ingo Paas:

I didn't have a clue really what it was, but I said you know, it's a Swedish company, it's cool. They speak a different language, it's cool. And it's also very international that's cool. And they bring new technology it's cool. So I better find out. And then that brought me to Sweden a couple of years later. So I've done an IT project and they found out we've done something right in that one. So I got a lot of global responsibility at Ericsson for five years and having a great time learning a lot. My first boss he was from Australia and when I entered his office he was saying something to me. I couldn't understand what he was saying. He was from Australia. It was like don't eat kangaroos when you speak to me.

Ingo Paas:

It was like it was the beginning of a good friendship. It's like it was really cool. So I went back to Germany when the dot-com bubble spread and, just you know, one year later telecom, I think, destroyed even more money than the dot-com industry lost. So we went back to Germany. We've got our two kids and I joined Adidas at a point of time in my life where we were completely out of track and so worked with Adidas at the headquarter for one year, learned a lot about this global German company and you know no one knows that Puma and Adidas are coming from the same location. I know Someone knows. You know, henrik, you're cool. I made my decision Okay, and you know I was with Adidas and was really proud about that because I was wearing Adidas all my life when I was playing football. So it was really cool.

Ingo Paas:

And then the CIO asked me one day to come to his office and I said he's going to fire me. And then the CIO asked me one day to come to his office and I said he's going to fire me. And I said no, ingo, I want you to be part of the new formed area Nordic and take the CIO role. And I said what was the CIO doing so? I didn't know really. So yeah, he said I can tell you Because he was the CEO of the group and we had some good talks and now my family and I we went back to Stockholm and worked here for a couple of years.

Henrik Göthberg:

CEO in the Nordic, the Nordic.

Ingo Paas:

And later on Central Europe as well, for 16 countries. And then, when it was time for me to move back to the Ecuador and we've done this once so we said do we want to do it again? And we said no no, let's go. We're swedish now yeah, we, we started to enjoy scatterbracket, and that's good oh, my god, when you, when you, when you fall in love with these kind of institutions. You don't want to move because you know how painful this is everywhere in the world.

Henrik Göthberg:

But you know how painful this shit is everywhere in the world.

Ingo Paas:

But you know you contribute with good stuff when you pay your tax. But in fact I became a consultant and seven days after I became a consultant, career Lehman Brothers, Lehman.

Anders Arpteg:

Brothers.

Ingo Paas:

Yeah. So I went down and I thought you know you could have, I could have chosen a better timing. Global financial crisis 2007, 8.

Henrik Göthberg:

So I went down and I thought, you know I could have chosen a better timing. Global financial crisis 2007-8.

Ingo Paas:

Yeah, it was six months without a job. Got fired but re-employed the same day and started one week later with my first project. That was IKA. Then this first this contract was only for four months. It was funny. Ica. Then this first contract was only for four months, it was funny.

Ingo Paas:

But then I got a new contract and till svideranstalt, only to leave just three months later, and got employed by ICA. So it was this funny experience. So my consultant experience was really bad, but I think I met you first time when you were were you head of innovation at ICA?

Henrik Göthberg:

funny, funny experience on my, my concern and the experience was really bad. But yeah, but it's I. I think I met you first time when you were were you head of innovation at ikea? What was the actual title of the cast?

Ingo Paas:

that was a very long time. We don't have the title, but you know, the funny thing is we've had some people. How can you be head of innovation if you, you know, and then you should contribute to innovation and you know innovations. Innovation is everyone's responsibility.

Henrik Göthberg:

It was a funny title but it was difficult to really contribute with innovation and the IKA is also. It's a very Swedish company and a very Swedish organization structure.

Ingo Paas:

Yeah, at the end of the day I think the company was great, the people were great. I learned so much, but you also learn that you cannot really do so much. You know. At the same time as you learn, you learn about the policy making in such a big company, you learn about the culture, you learn about the people and the network you need to create and the relationships you need to have and everything else. So I became CEO of the pharmacy business and enjoyed this very much, but 10 years later I left and then I ended up at Green Cargo and that wasn't planned, but it was.

Henrik Göthberg:

Was it 10 years? How many years at adidas and how many years at tikka? At adidas.

Ingo Paas:

It was seven years at uh, it was 10 years, 10 years yeah and then tell us, because I remember we having lunch.

Henrik Göthberg:

You know I just started there ducks and you were right in between. You know. You know what's gonna happen now and I'm not sure you have land. You had the Green Cargo gig when we had that lunch.

Anders Arpteg:

You remember?

Henrik Göthberg:

You know, five, six years ago.

Ingo Paas:

It was not a gig. It was more a big disaster decision in the beginning.

Henrik Göthberg:

But tell us a little bit, like you know. Why did you, what made you go to Green Cargo and what hooked you on it?

Ingo Paas:

Yeah, I need to say one word. I met Anders and I wrote a business case about a company and I thought you know large language models and generative AI is probably on its way, so maybe I could start and build a company to get Gartner out of business, which was a big dream. It's difficult if you don't do this overnight. But I was actually thinking about, you know, the concept of you know, using these models to create personalized research based on your profile, based on your company, based on the projects you run and everything.

Henrik Göthberg:

And now it's very close. Now we're there.

Ingo Paas:

We're close and I talked to some investors later and, Anders, you were also laughing at me, but you said the stuff is coming and that was really cool, but you didn't have the time to help me.

Henrik Göthberg:

But that's the same thing. You had lunch with me and then you had lunch with Anders at more or less the same time.

Ingo Paas:

More or less. Maybe it was the same day.

Anders Arpteg:

So funny.

Ingo Paas:

And we didn't know that, we didn't know that so cool. No, but I had to pay Anders lunch to get into it.

Henrik Göthberg:

So he was more important.

Anders Arpteg:

But please tell us a bit more about Green Cargo. What do they do?

Ingo Paas:

Yeah, so you don't want to talk about the lunch, right? I understand that I joined Green Cargo because they were looking for a new CIO and I was meeting the CEO of the company and it was like we did not talk about what I've done. We did not talk so much about what he was actually looking for. We just straight started to talk about this problem and that was a two long hour conversation and he said oh, I only had one hour, so and and and. Then they called me and said you're welcome, so join us. And I joined the company, not because of green cargo, because I didn't know so much about green cargo at this point of time, but I was. I was fascinated about the discussion we had and I really felt that I could make a contribution. And the company is at a reasonable size it's not that big, so maybe I can be myself and not be part of this huge thing that always you always have 5, 10, 20 people who have an opinion and you have to convince them.

Anders Arpteg:

So how big was Green Cargo at that time?

Ingo Paas:

2,000 people, 4 billion in turn over Sweden Krona, so very reasonable size.

Henrik Göthberg:

Reasonable size compared to Adidas. Oh yeah. So I'll use a funny anecdote. My brother worked for many years at Green Cargo and he actually left. Had nothing to do with each other, but the anecdote is that I had I've been following Green Cargo through my brother for many, many years and he left right about the same time as you joined.

Ingo Paas:

He heard that I was coming, yeah, he heard no, no, I called him.

Henrik Göthberg:

You need to get out of there.

Anders Arpteg:

But please for people that don't know.

Ingo Paas:

His coffee was still on his desk.

Anders Arpteg:

So Green Cargo is actually a spin-off from Swedish Staten Säger right.

Ingo Paas:

Yeah, you're well informed.

Anders Arpteg:

And please let us know, what do they do?

Ingo Paas:

We're running trains about 400 trains a day so it's a huge and important business. I think when you look at the rail cargo industry, we're about moving 30% of the goods in Sweden with a pollution that is much less than 1% of the total pollution.

Ingo Paas:

So the contribution for a greener future is amazing and this is a massive thing, and we do everything we can to make that business more efficient and to get our customers happy, which is difficult because there are so many limitations, restrictions and problems. So the way we are running trains is one thing, but then there's the infrastructure and then there's a timetable, so you need to really make sure you get your resources organized.

Henrik Göthberg:

It's an NP problem, the travel and sales problem.

Ingo Paas:

at the end, yeah, but we should not blame anyone else for our success or problem, but it's a complicated system.

Henrik Göthberg:

It's a complicated system.

Anders Arpteg:

Yeah, for sure, and tell us a bit about your role there. What are your main concerns? What do you do there?

Ingo Paas:

Okay, when I joined the company, I said it was a big disaster decision. The first thing I thought I should leave because you know we all love information security, cybersecurity and these kind of things and I asked the question what do we have? We have a policy and that was just, you know, written on used paper. It was not even a policy and it was the same with everything else I've asked for. So I could not find anything that was good or anyone that was happy, or IT was not respected, and there was nothing. And most of the projects that the company tried to implement for many years and there were strategies coming from McKinsey and Oliver Wyman and they all failed.

Ingo Paas:

And then I came in and we had a very complex infrastructure with two mainframe systems, one SAP platform that was completely out of control, the same for the mainframes, and then lots of ERP planning systems totally disconnected, and I thought you know, no one who had the courage to do change, no one who had the skills to do it and no one who was empowered, and I better leave. But my CEO said I trust you and you know you can clean it up and I could make all of those decisions on my own. So when I was meeting with SAP the first time, I say we want you to go back to maintenance. You know I could make that decision on that day in the meeting and I told these guys because SAP is a great company but it was not right for us at this point of time. So it's not about SAP. But I said if you have customers that love you, you better focus on them, because I will not buy your proposal back to maintenance. These are the kind of decisions that helped me to make a lot of decisions in a very short time and to help to get back to a foundation and instead of taking the responsibility which I was asked to do, replace all of these core systems, the two mainframes and the CP.

Ingo Paas:

I said we don't have the money, the time, the resources, skills to do that and if we do that we're going to be out of business because by that point of time, in five, six years, we start with something that we don't know what it should be. You will solve a problem that is probably the wrong problem and then we try to solve it in a way that we can never get it right. So it will be a disaster. So instead we tried to solve the critical business problems. We put a digital infrastructure into place.

Ingo Paas:

We did this based on core principles and the most single important word describing this architectural infrastructure is the decoupled architecture. It is decoupled to its very core. All of the platforms that we've built, all of the changes that we've done, we've done it incrementally and we've built that from scratch, and that was actually giving us the power to solve a lot of business problems. And during this time, the business has reorganized. We have a new CEO and now we have a very strong business strategy in place, and now we are replacing all of these systems. But now we know why we want to replace them. So there's a good reason and there's a there's a platform.

Henrik Göthberg:

Yeah, you slot news. All this money spent on you changing technology for when we don't can't even explain why we need to change it exactly, but now the business knows and they're better in, in, in in explaining you know what they want.

Ingo Paas:

We still have a lot of hobby architects in the company, so they make a lot of noise, but we're trying to deal with this.

Henrik Göthberg:

I remember this, and maybe it's also one of the reasons when you won the CIO award, you did something quite interesting in terms of containerizing or encapsulating the old technology, and you used a specific approach to this. Could you elaborate a little bit on this, because I think for many people who are stuck in legacy technical depth, they don't know that it's possible what you did. Could you just elaborate a little bit about how you went about containerizing the old in order to take it in? I think it's a very interesting concept.

Ingo Paas:

Okay, yeah, Now one more phrase. We talked about people from Oxford. People from Oxford know that word. There's two words. I want to use the difficult word first. It's ambidexterity, Ambidexterity.

Ingo Paas:

If you haven't been to Oxford and I haven't been to Oxford, I have Google, so you would not know what to do. Mb dexterity. So that word explains the difficulty of an organization being in two different stages at the same time. One is the exploitation, the other one is the exploration. And how would you do that in a balance to bring the organization forward and helping the organization to make that transformation and balance and integrate it move. So whenever you deal with your legacy, that means you have to use your innovation capacity to help the legacy to become better. And you need your legacy because it's always from the core of your business to do your innovation, because if you do your innovation outside the core, it's innovation that doesn't really give you a payoff long term. So, and then I came here's the second word, the one that is much more easy to pronounce.

Ingo Paas:

It's composable that's the one you were looking for right now, so we've built a composable architecture and that is that's really based on those principles I explained, without going into the deeds of the principles, but very much allowing you to build platforms, and I created this in my first book that I published in April 23, the definition of enterprise platform engineering.

Anders Arpteg:

So if you would Google, Perhaps you can go to that book as a separate topic. I think it's a great segue into that topic. And what was the name? You wrote a book called Digital Composable Enterprises, right, yes? And can you just elaborate a bit about the principles and what your content of that book was?

Ingo Paas:

Yes. So principles that you want to have in place and other principles that help you to describe what you do on your business when you want to build a digital infrastructure. So the principles you want to have. They're independent of the business you run. So I give you some examples.

Ingo Paas:

One is interoperability. So whatever you've built, it needs to be interoperable. Whatever you build needs to be robust. Whatever you build needs to be autonomous. Whatever you build needs to be robust. Whatever you build needs to be autonomous. Whatever you build needs to be in a, as you said, decoupled situation so that it's very strong. Whatever you build needs to be scalable. So whatever you build needs to be reusable. So you find those principles and then you build your story around those.

Ingo Paas:

So when you make your decisions, you always ask yourself is this a platform that I can stay in? Is this a platform that I can reuse? Is this interoperable? Is it working with all of the other things that we have, or is it giving me new opportunities?

Ingo Paas:

And when you solve a problem in that complexity, make sure, if you cannot solve the problem with the platforms you have, you either add a product to those platforms it could be on the analytics side, a database or something similar, or you find a different way to build another platform that you can scale. So the first problem that you can't solve on the existing platforms should not be a new application where you try to build a platform and what you can see is that you will find similar problems in most cases. Sometimes you just have to buy a system. But what we have seen new problems created new demand on platforms and then we created new platforms and the funny thing is that the more you see those platforms and from an enterprise perspective and it was my idea about the enterprise platform engineering thing is, when you start building the engineering capabilities around those platforms and they are in harmony, you can scale at any level.

Anders Arpteg:

And I have to throw in some Elon Musk comment here as well, but he you know when he's. We have this kind of Tesla agile DNA, and one thing he spoke about is very similar to what you're saying here, but he uses term like a very modular approaches, and if you have a modular approach, that is, you know, decoupled capital, capital that means that you can actually work in parallel in much more efficient ways. So you can continue to work with fixing a small like water pump in the car, but as long as it has clear interfaces to the rest of the car, it doesn't mean you get blocked or you can work in an independent and much more scalable way. Is that similar to what you're speaking about here?

Ingo Paas:

It's difficult to compare myself with Armasi. He's much richer than me, probably much more successful in terms of business Not these days, but absolutely. But I think that if I look at it from the way we have reasoned these days, it's pretty much that we said we want to build an environment where we have all of the digital infrastructure in the cloud, so when we orchestrate the objects in our infrastructure, they should be as strong and autonomous as possible. If one object is failing, it does not have an impact on the rest.

Henrik Göthberg:

No, exactly, and this is you can take this analogy with composability, even into the future. Now you can even argue principles like design to be replaced Right now, with the technology speed we have right now, with the innovation speed and even tech vendors coming in and out from Silicon Valley, dbt coming from nowhere and now you know they're not even green behind their ears and everybody's using their technology and we don't even know if they're going to be around for five years. So design to minimize blast radius is sort of the risk mitigation view of talking about these topics as well. So I think this whole modularity or composability topic I think is the only way, because it's the only way we can take this all the way to AI. And again, take AI and having agency of systems and agents. They need to be modular too. They need to be composable too, otherwise you're going to have different alignment on what an AI system can do and the agency of each team. So this composability, modularity topic, I think it's at the core of the future.

Ingo Paas:

Yeah, I think you can relate this to my favorite sport is football. In the USA, I call it soccer. I don't know why, but for me, if you play football and you have a team, you can apply the same principles. When someone is having an unlucky moment and you need to replace that person, you still have to have a team that performs on the same level, and the one that comes in would just support the team in being equally strong as before. So you should not see that this is making a huge impact. So that would be the perfect team, and I think that it's the same for for a digital infrastructure. So you need to make sure that the infrastructure is as autonomous as possible and and it comes into, it did not really.

Ingo Paas:

It wasn't really clear what we were doing these days, so it was more when I was writing my book that I was learning about all of the decisions we made and how we made it up, because when I joined Green Cargo, I did not have that plan. I did not know what was possible, and so it was an exciting time to learn from all of the people around me and listen and support the decisions they were making, a lot of decisions. I was only supporting the decisions and trying to get my arms around it, and while we were making progress, we could see that we could create stability. We could create and that's, I think, one of the most important words in these days resilience, and resilience in what we have done is the digital resilience. And if you would ask a CEO if your business is resilient and the CEO says yes, because we survived the latest global supply chain wreckage, then yeah, maybe you did, but how can you be resilient if you're not digitally resilient? Which company is digitally resilient? So that's not the standard. How do you create that?

Anders Arpteg:

Yeah, and I guess that's my second question, because I think a lot of people would like to have a composable architecture and infrastructure, but it's not that easy to do all the time and if you have some system that is dependent on another or some old legacy, I guess, especially if you have a legacy legacy like tech stack and some ERP system or whatnot, and if you can't make planning around this because it's dependent on everything else and that can't move forward, can you just elaborate or perhaps give some example how do you really create digital composable enterprises?

Ingo Paas:

really create digital composable enterprises. Yeah, if you look at the idea that we've created, we said we want to have an environment where the data is becoming independent from our systems without building the platform for it, by just creating the various capabilities on those platforms. And at the end of the day, we ended up there without really having that plan. We did not know that we would use a graph database from Neo4j to build the ultimate experience of data innovation. And while we were doing these kind of things, we found out that when you have your exploitation and exploration problem at the same time, so when you have to balance this, so if you build innovation into your business and you build innovation into your digital infrastructure, you could use that innovation for the exploitation and support the existing systems. So you can use your local platform and you can rebuild capabilities, you can rebuild complexity and then you can make your legacy more agile. So when we've done this, we've moved from running four releases a year on our mainframe to 300.

Ingo Paas:

One second mainframe On mainframe, we were having four releases a year. We've had this in an outsourcing situation and we've done insourcing and we went from four releases to 300.

Henrik Göthberg:

So you're deploying small things all the time, exactly.

Ingo Paas:

Because you need to make your mainframe agile if you want to do the exploitation and the exploration at the same time. So the innovation is not possible if your mainframe is your key problem. And we were told to replace the mainframe, we said we better make it agile. You can never really make the mainframe agile, but you can make it more agile. And that was the idea that what we have shown it was possible to use the integration platform to make better integrations into our mainframe. And these are the things. This is how they support each other.

Henrik Göthberg:

But what I hear also. It's very easy to talk about this topic from the technology perspective, but what you're essentially talking about when you're going from four releases to 300 is that you're innovating on your engineering practice. Yes, You're innovating on how you're looking at the problem and how you're tackling problems. So that is maybe what constitutes a platform and a composable platform. Is it the technology or is it the practices and principles that you live by?

Ingo Paas:

The principles yeah, the platform is nothing else than just a temporary, let's say, configuration of your principles.

Henrik Göthberg:

Exactly.

Ingo Paas:

It's nothing more than this, but the platform is very complicated. You need to be very good in doing the platform, so you need to have highly skilled people, and when you move from zero cloud to all in the cloud, this is a huge thing in the beginning, but if you have the discipline and understanding why you want to do that, you also find the opportunities in the cloud. Like you know, when you think about the opportunities from the large players and you have your past platforms in place so you can get so much innovation and whenever you have a new problem, you just add a product and you can increase the value of your platform tremendously.

Henrik Göthberg:

But I think this is something that we almost this is almost too fast. Slow down a little bit, because I think I engineering practice exists and where the platform evolved Versus when we come to enterprise and we start listening to the vendors and we think the tech will solve the problem for us and we don't even sometimes reflect on the fundamental practice or principles, and I think that's maybe one of the key ingredients here. You know, how do you do a composable digital platform practice? You work really, really hard on the principles.

Ingo Paas:

Is that a summary yeah, I would say so, and I would say that the more you build your platforms by defining platforms as a pre-built solution, the less you have platforms that allow you to be innovative and explorative. So I would say that you know I don't want to mention the big sales platform, things that you can configure and you can do this. I don't call it the workday platform for HR or something like this. You know you can configure those platforms and they call it a platform, but it's not a platform. This is something else, because the only thing you do you configure in a given environment. Someone else is saying what you're allowed to do because you have your limitations. You do not build a platform. When you select such a thing, it's a big, big thing that's complicated and it's forcing you into something rather than a platform that is holistically designed, based on the principles and that you can replace or you can change.

Henrik Göthberg:

So the platform is what's giving you the maximum flexibility and freedom in designing and solving business problems yeah, but and and then we can conclude that the platform word and what we mean with platform is so screwed up in terms of how we've been using that word and what it means to different people, and I I concur with your idea what constitutes platform, but I hear everybody's the platform and every single vendor is the central platform. Doesn't matter if it's a marketing platform or is if the data platform, is it the erp platform? And it's just a name. It it's just a name.

Anders Arpteg:

Do you have a preferred definition?

Ingo Paas:

of what a platform is or should be A platform, should be a very individual, autonomous entity of multiple opportunities that you can configure at any time in any way you want.

Anders Arpteg:

It is a multipurpose.

Ingo Paas:

It should be multipurpose, it should not be a single purpose.

Henrik Göthberg:

And I want to add to that definition what you already said. I want to suck that out. That was difficult, by the way. Yeah, but the platform definition here is also about what constitutes a platform is not only the tech, but it's our practices and our principles that we have around any technology. We are technology agnostic and yet we have a platform.

Ingo Paas:

Yes, I think the thing with platforms in terms of business success is that you do not talk about platforms, you talk about the business value. So when we've done this journey at Green Cargo we never actually went out with a plan or a demonstration or a PowerPoint presentation we said look, we are building platforms. No one knows, no one cares, because building the platform is the engineering thing that you need to own as a CIO and CDO, and when you build that, it's not important for the business. When you run a train, you don't want to know about's not important for the business. Like you know, when you run a train, you don't want to know about the steel composition of the wheels. You just want to have the wheels to run your train. So don't make it too big, but make sure that you have the ability. You want to ask me about scalable investment strategies later on so that you can relate your platforms to such business value propositions. It's not, which is not easy, but then it's not necessary to explain.

Henrik Göthberg:

Sometimes you need to but that's also you're honest, if we segue back to the question, give you some tips and tricks. You know how to, how to do it, how to think. I test another angle here to jump in. I'm putting words in your mouth. I think you're talking also about being super vigilant about business use case or business problem solution. You have ideas and principles that are platform fundamental how to act and how to build. Platform fundamental how to act and how to build and then you viciously focus on a specific business problem that everybody understands and wants solved. So the business-led view on this must be a key point.

Ingo Paas:

Yes, I can only agree, and I can reinforce what you just said, that when you have the right capabilities and you have the right data, you understand the business problem. So you need to have the data and you need to be able to interpret the data and to make it available. And in this discussion you always come to the conclusion that you can get better support when you come as a business person and you have a problem, Because with the data perspective you can explain the problem better. And that's what our experts are doing on IT. They have a reasonable understanding about the business not so much, but they have the data and they can help explaining the problem differently and that helps a lot explaining the problem differently and that helps a lot. But it's only possible when you create this clarity and the integrity of the data when you have such a complicated legacy.

Anders Arpteg:

Yeah, super interesting. And you also speak about I'm sure we should move into this topic right now, but you're speaking about this kind of concept of you know, harnessing horizontal and vertical innovation. But what do you mean with that? Can you just elaborate a bit more with those kind of dimensions? Yeah, absolutely.

Ingo Paas:

First of all, it's very complicated, but on the other hand, it's not. So. The horizontal innovation is the layer of platform innovation that you can use. So this is where you build the enterprise platforms, the composable capabilities, where you can bring innovation from Google, from Amazon, from Microsoft. You can bring in specific products and drive innovation. So you build scalable capabilities that will help you to solve lots of problems. But you won't solve those problems on the horizontal layer because this is your engine. The engine is not about solving problems. The engine is going to to support the problem solving the vertical perspective.

Ingo Paas:

This is the scalable view on on having all of the business problems. It could be business processes. You know you should not always have problems, but I love to use the word problem because it's describing a particular thing that you want to solve and and it's not negative. So so if you, if you think about the vertical perspective this is why you can build solutions based on the platforms that would learn from the solution you build and would only be enriched and would also inform other problems that you want to solve. But here you can use distance. It will help you to solve your problem faster or better. So the vertical perspective of the platform is all of the problems you solve, which could be processes, which could be business cases, we could do everything else and it is a scalable moment where you contribute to the business, while the they are visible, they are out there, the people, people feel, know, use them, love them, hate them, whatever. But nobody knows about the horizontal part, as long as it works.

Henrik Göthberg:

Can I see? If I understand it right, you can also look at. Let's say, an example is that we have many different use cases where we need data and we need to share data between different business units. Each use case has a unique composition of what type of data they need to share to solve their business problem. This is the vertical view of solving one use case. However, in order to have extremely efficient data sharing, we might need to have fundamental investment in how we standardize on protocols or metadata or something that is in the backend in order for data to be shareable. So data to be shareable is not really interesting on its own. So this is then investing in platform capability in order to share data more effectively, in order to have that more effectively used, in order to have that problem solved more efficiently in each use case. Is that an example?

Ingo Paas:

Yeah, I can refer to this problem in a different way now, because I can only be the observer of the problem, because I haven't solved it, because Green Cargo is a much more let's say well-defined business, if you compare this with Sandvik, for example. So you have strong business units and they're getting too much involved in the discussion about the horizontal layer. You should disconnect that, because then the horizontal layer could be the same for everyone. They don't need to know, but they are always engaged in I want that Microsoft product, I want that Google thing and I want this and that, and they want to build things because they think that they need the autonomy about the platform.

Ingo Paas:

No, the platform is like the infrastructure you need to drive a car. Not everyone would have to have his own highway because that would not work, but major companies think that this is still the way to go. So if you disconnect the vertical perspective, then you can build as many cases on different business units utilizing the same digital infrastructure, decoupled and ready to run. You can have 20 business units with 20 different business models using the same analytics platform. I think you know it as much as I know you have probably seen the opposite.

Henrik Göthberg:

But you're pinpointing at one of the major enterprise challenges, especially in a little bit more decentralized, distributed enterprises, where I joke sometimes and say they hide behind the word autonomy but they want sovereignty for the full back end, for the whole thing. You see what I mean, right? So they are really arguing instead of having agency for their problem. But you should know where you don't have agency for the horizontal problem, that's someone else's problem and this becomes messy fast, right, and it's difficult.

Ingo Paas:

And I think it's much easier for a company such as Green Cargo for me as a CEO of a small organization to build these capabilities especially if no one gets involved Because we had that luxury so we could make all of this decision without having any discussion. I think the only platform that was up for decision for an investment with our board of directors was the integration platform, but it was a big investment because there was a compliance issue that we had to solve immediately. But it was a big investment because there was a compliance issue that we had to solve immediately, so it was a quite big thing. So we planned to do it in steps, but then we were forced to make a big investment, but otherwise we never really made this too big. So platforms are not important for the users. They're not, no, they're not.

Anders Arpteg:

They're not. I mean, I think it could potentially be easier to innovate in the vertical sense. So then you know those use cases is easier to define the business value for you can easily see. They're potentially easier to break down as well in smaller parts. How do you make investments on the horizontal layer?

Ingo Paas:

That's a very good question and we come to the other question Can?

Anders Arpteg:

you combine the answers, of course, the scalable investment strategies yes, exactly. Okay, please elaborate.

Ingo Paas:

I love that. I love that one Because the thing is that when you want to build a platform, you cannot go out and say I want to build a platform Because which CFO these days would come and love you for that, yeah, a platform. You cannot go out and say I want to build a platform because you know which CFO these days would come and love you for that, yeah, because the first question is you know a platform, what do you want to do with a platform?

Henrik Göthberg:

What's the value?

Ingo Paas:

Yeah, we have this business problem, we have this project, so solve it. So then you buy an application and then you end up with a very strong ROI. You have your business case, you do your TCO stuff and everything looks fine because you can isolate that investment. But when you've done it you may find out the problem was not really well defined, so you need to change it and it doesn't really help us to scale. We cannot use it in the business units everywhere. And then you find out there was a bad investment.

Ingo Paas:

The difference to a platform investment strategy and there comes the scalability is that you find out that the business has a problem. You cannot solve it with the existing stuff you have, so you build a new platform. So if that platform helps you to solve that first problem, you finance the platform. But you do it more. You don't talk so much about it. You just build the platform. It's part of the investment. So you explain that you need to buy these things, you need to have those capabilities, and you implement the platform.

Ingo Paas:

And then when the project starts, if you're smart, you already have planned for it because you need to first set up the platform. It's like going through an RFI and RFP. So you need the time to build the platform first. Never, ever, allow the first problem to define the platform. So make sure that when you build the platform, the platform is based on those principles that we discussed before. So you put the platform into place and then, when the platform is there, then you build the first case on the platform. And now we already invest into your platform. It's not ready because you have to do adjustments over time. When more problems come, you can do more investments on your platform, but they are much smaller and different scale. So you already use your platform to build new solutions. That is the scalable element of it. It's very simple, but if you have the luxury to make it work this way, let's see if I get it.

Henrik Göthberg:

If I go the application-centric way, where you have a use case project and you need to stick with that, you end up buying a point solution, you end up even building a technology, but you're letting one narrow use case define everything and your actual principles and framing requirement doesn't exist, it's only the business requirements for that use case. This leads to fragmentation and a lot of different point solutions, and going the whole monolith we need to build everything at once doesn't work either. What I hear is something like think big, think principle and architecture, what you're thinking about, but this is just a hypothesis. And then you allow your first use cases to, based on those principles, do the first pieces of the puzzle and then you have an evolutionary investment strategy and you're drying out the old by stepwise implementing production of use cases with a principal idea Think big. So the important thing is think big, start small Scale fast. Exactly, I agree this is think big, start small scale fast.

Ingo Paas:

Exactly. I agree, this is exactly what we've done. The thing is that we also built in the resilience and the flexibility, interoperability all of the things we discussed before. I can use the example of our local platform. So we decided to implement a local platform. The business said we have a problem Wagon damage reporting. You have to build an app. Okay, we have to build an app.

Ingo Paas:

We had, until this point of time, never developed applications. That was difficult because they couldn't buy an application for that. Okay, so we should build an app. Shall we go out and find someone to build an app or shall we look at this as an opportunity to build a platform? We decided and I was in business for about I think, eight weeks, 10 weeks, and someone came and said Ingo, we want to look at low-code. Okay, what do you want with low-code? Yeah, we want to solve that problem. I said that's not good. So that was the time when we started to really come to the point of making those more I would not say intelligent decisions, because my wife would really she would oppose Heavily. She would say it was just luck.

Ingo Paas:

And what we found out there was in our discussions and also the negotiation with the vendor when we were discussing the final thoughts and our approach. It was not in line with the vendor's contract and I said I want a scalable contract and you need to remove those burdens from the contract. Whenever I have to buy a package of application objects, when I want to do additional development, I need to come with new business cases on that particular cost. I don't want that. I want unlimited. That's the only way for me to build a scalable platform. You know you're already in the contract phase. You need to have that in your mind. But at this point of time I have not had the time to think about it because I was only running through incidents and problems. But this was the kind of principle of thinking that you need to apply in every step you take. So it's a contract. What are my principles? Are my principles aligned with the decision I make or not? And if they're not aligned, don't make the decision. It's hard, but it helps you to find better solutions and the local platform.

Ingo Paas:

So we're using our systems. We're very happy with the platform and we've built the modularity in a way that one of our tech leads came and said you know, I'll give you one example before I talk about this one. So we built the first application and then the business found out but that's not what we want. Not that we built wrong, but they did not know the scope. So we changed the scope, then we split the application and then we added another application that we built to the first type of the application.

Ingo Paas:

So we really modeled because it's so loosely coupled, right, and that was one thing. And we could develop other applications and we never, ever had a problem when we were orchestrating the objects into the cloud, while many other customers of our systems had problems because they did not follow that kind of decoupled principle, and that is a huge issue. Right, customers of our systems had problems because they did not follow that kind of decoupled principle, and that is a huge issue. But if you start wrong because if you want to have a platform to solve a problem but do not think scalable, you will run into major problems, performance problems.

Ingo Paas:

And then the second thing is that this tech lead came to me and said Ingo, we have now built a new application and that will be the last one we built. And I said, wow, are we now leaving our system? No, no, no, no, no, because of the principles that we've applied in all of the development and we've also done mistakes there. But the majority was built loosely coupled. So he said from now on we are terminating the idea of building applications and we are building loosely coupled services, meaning that when a user is starting his phone in the morning and logging in, the application is built based on the profile of the user. It's a dynamic application, so we do not define the application, because all of the objects are so autonomous and independent they can create the applications spontaneously. Super cool.

Anders Arpteg:

So hyper-personalized kind of application. So it's actually changed to your needs and the context that you're in. Yeah, you get my Pages.

Henrik Göthberg:

So it's a subset of services and features and, depending on your access and what your profile is, your screen has these five widgets, whilst another one has five other widgets.

Ingo Paas:

It's mainly on mobile phones and it's basically on roles, so it's not we don't have one on different. It's a waste of looking at the same stuff, but it's like the same role has the same profile, but it builds automatically. So we've tested with the first part and now we are migrating.

Henrik Göthberg:

So this is what you meant. We are not actually building new applications for everything like now. We're having an architecture which is a fundamental GUI, where we then add the services widgets appropriate to the profile of the user.

Ingo Paas:

Yeah, and it comes automatically because it is connected to the Active Directory. The Active Directory tells which role it is that the user is being given and, based on the definition in the Active Directory and the application the last application we built it is getting the objects that build the application Interesting. Have you seen many other?

Anders Arpteg:

examples going all this way Because I think it is getting the objects that build the application Interesting.

Henrik Göthberg:

Have you seen many other examples going all this way, because I think it's super cool.

Ingo Paas:

And our systems have said that they haven't seen this so much. And they are. They are having thousands of customers globally.

Henrik Göthberg:

Very interesting, yeah, very interesting.

Anders Arpteg:

Hmm. Would you say that, like Netflix with their kind of shelves and even Spotify and others, are in some way at least a bit like personalized applications. Could it be a similar kind of idea?

Ingo Paas:

Not sure I don't want to. It's difficult for me to compare with these monster development companies. I think they're so good in tech and I'm not sure I could say something as an observer.

Anders Arpteg:

Maybe you could. I think it sounds awesome, of course, to have a personalized experience and being able to build towards that would be good, but I think still, you know the idea of how to scale investments both horizontally and vertically is still very difficult, I would say. But I think still, you know the idea of how to scale investments both horizontally and vertically is still very difficult, I would say, and, if I understand you correctly, if you can do it, if you can do the horizontal one as part of a vertical one, that's a good start and then you can try to scale from that.

Ingo Paas:

But still it's easier said than done, I would say, of course, and you need to have the lucky, the lucky moment that you know you are ending up in an environment where you have no other choice but, but there is.

Henrik Göthberg:

There is. I mean like this is a rabbit hole. I don't, we maybe should not open this one up, but I think what you have done and what you're talking about, in order to go this route, you also have to have the organizational construct and the mandate construct, where people respect the different agencies of horizontal versus vertical. If I look at very large enterprises that I've been working on lately, it's a political minefield. Minefield, even if, theoretically, people can, in principle, understand the ideas. In the end there are so many different agendas why you want to do the whole stack or why you know so they are not completely aligned as a complex system together, as an enterprise, together, how that should be implemented, which I think then is okay. A little bit smaller size, a little bit shorter, from the CEO to the CIO, to the VPs. So all these things makes this way harder. If I go to a global setting.

Anders Arpteg:

And we've been speaking a lot about enterprise kind of challenges and investments there and innovation. I'd love for us to also go into the more AI phase and I think we have a number of topics. I think you have a big interest in more agentic kind of capabilities and what that could entail for these kinds of solutions as well, so perhaps we can take that route shortly. But before that, goran, perhaps we could have a short news break.

Goran Cvetanovski:

Before that, goran, perhaps we could have a short news break. It's time for AI News brought to you by AIW Podcast.

Anders Arpteg:

So we usually have this short break in the middle, not always super short, but we're trying to keep it short and we just take a break from the discussions and continue to just have some personalized reflections on the latest week in terms of AI news. I actually personally been away on vacation and just came back like late last night, but I have potentially some topic, but I would prefer someone else. If someone, perhaps Ingo have you heard about some interesting AI news that you heard in recent weeks that you'd like to share something about?

Ingo Paas:

It's okay if it's a bit older, of course. Okay, good, I was thinking about if you use the word willow, people will think about it and say, if you use the word willow, most people would consider this to be a Walt Disney movie.

Anders Arpteg:

I think in Google right now. Please continue yes good, exactly so.

Ingo Paas:

I wrote down some things, because now I'm talking about something that I'm not really good at.

Anders Arpteg:

But it is one of my favorite topics, so it's a rabbit hole.

Ingo Paas:

Yeah, it's a rabbit hole, so I will talk about it and you will make it. You will refine it then. Okay, Anders, Can we do that Absolutely?

Anders Arpteg:

yeah.

Ingo Paas:

So maybe I should do this without my notes and then you make it really good. So the idea is that when you think about the computational power, that will give us enormous muscles in how we can apply AI in the future. So we need to find the use cases. They're not really there, but if you think about you know what things, what is possible. You need to consider quantum computing. So understanding quantum computing is not my thing. I'm not intelligent enough for that. But when you understand what Google has done, when they built Willow, they started 30 years ago, so the idea was to solve the problem of quantum computing and qubits, they are running into errors because of the logic of the unclear definition of between zero and one is indefinite, it's just you know.

Anders Arpteg:

Yes, it's superpositioned.

Ingo Paas:

Superpositioned exactly, thank you. Here we go the superposition, so that means there is an error rate and that error rate is quite tremendous and the greater the error rate, the less the computing power is useful. So you cannot scale on quantum technology. So what Google did and the research team? So they created a. It's not scale but it's an algorithmic solution to that problem. They solved that problem in a very smart way, meaning that the computational power of their chip technology chip technology which was done on a theoretical benchmark, um has such an enormous um computational power that when, when I tell you the numbers now you probably get impressed, um so yeah, yes, the fastest supercomputer as of today, um, would more time, much more time than the solution that Google built.

Ingo Paas:

So the Google said that it's about five minutes to run a mathematical problem. And then they applied the benchmark and it's an RCS benchmark, rca benchmark.

Anders Arpteg:

Yeah, the random circuit.

Ingo Paas:

Random circuit. Yeah, exactly, thank you. And when they applied that benchmark, they did a calculation and I said it was a careful calculation and they ended up with a number that is 10 at the power of 25, which is 10 septillion years that the most modern supercomputer would need to solve the same mathematical problem. To understand what that number means, I've done a calculation. I asked Chadky Patino 10 to the power of 25, divided by the age of the universe, it told me it is 725.6 times the age of the universe. So this age of the universe, so this is a massive thing. So I can hardly understand 100 years of living. And when you hear that number, this is the computational power that is possible with quantum technology. So, ray Kurzweil, you were right.

Anders Arpteg:

Okay, so you went into a rabbit hole. This is a rabbit hole where he will you know, it's okay.

Ingo Paas:

It's okay, I have, I have to.

Goran Cvetanovski:

I have to Now. This is going to be a three hour podcast.

Anders Arpteg:

Okay, trying to keep it short, but I actually have a big personal interest in quantum computing and I wrote my first program in 2014 or something quantum computing program. So, but I wrote my first program in 2014 or something quantum computing program. But I do think it's a big hype and let me try to keep that short. So, for one, the Willow invention that they did not. Innovation, using your terms, is trying to improve the error rate, as you say, and the normal problem when you scale up the number of physical qubits that you have is that they have deep coherence problems, so they get a lot of noise when they basically connect to each other. So they have the entanglement problem and then the superposition problem of the qubits and as you add more qubits, the entanglement rate is going down and they have increased decoherence problem to see. Now you have similar kind of problems for normal RAM memory that is not quantum at all, and then you can say I want to have super low error rate in the normal random access memory that you have in a CPU or a normal machine or even in a GPU. Then you add extra bits to it, like parity checking, to say that if some bit is flipped in the wrong way, you can detect that and correct it, and this is basically, as I would say, that they did in the Willow ship as well. So they add error correction qubits to logical qubits and by adding more error correcting qubits they can actually see that the error rate is not going up anymore as they add more qubits. So that's actually very good potentially. But still what they did, they had a 100 qubit system. They still had a single, more or less a single logical qubit. So if you truly want to do something that is of use with a quantum computer, you probably need 100,000 or even a million qubits logical qubits, potentially, and still you have one now. But then potentially you can see that by adding more error correcting qubits you at some point start to not go down in error rate for the whole logical qubit and you shouldn't. I mean, as you add more and more error correcting qubits, you should at some point stop reducing the error rate but you start increasing physical qubits instead.

Anders Arpteg:

So in short, what I'm trying to say here is a bit, and the example they gave with the RCS, the random circuit sampling, is basically trying to simulate a quantum computer with a classical computer and I think it's completely misleading that that would be any kind of useful application. So in theory there is not a single or sorry. In practice there is not a practically useful. You need to be able to scale up to at least thousands or tens of thousands, preferably millions, of logical qubits and then you can start to use it for some practical things. And we haven't seen it yet. And then you know this is just the skyline.

Anders Arpteg:

I can go into this kind of discussion of why I think it will be just increasingly harder as you increase the number of logical qubits. I have my own theory, which I call the conservation of complexity, but let me not go into it. But I think you know as you scale it will just be increasingly harder in an exponential way to keep that rate down. I can't prove it, it's just a theoretical thing. But even if you could, if you listen to Demis Hassabis he got a question about quantum computers and Demis Hassabis is the AI lead in Google and previous founder of DeepMind, and even Jan Lekun is saying the similar things here but even if you could scale a quantum computer, you could scale it to a level where it should be able to have practical use. Then the question is will it have a practical use. And then Demo Cesar is compared it to AlphaFold.

Anders Arpteg:

Alphafold was this thing that he received a Nobel Prize for in end of last year, where they could basically fold proteins so understand the 3D structure from a sequence of amino acids, and that's super useful and they did it like faster than anything else possible, instead of having five year for a PhD that do a single folding of a protein. They now have done it for 200 million proteins and basically made everything available for free. Super cool stuff and it's certainly well worth the Nobel Prize for that. He said. Basically this is exactly the type of computationally complex problems that quantum computers should be good at Optimization problems. It's like traveling salesman problems in trying to find out how this chemistructure should look like.

Anders Arpteg:

And to say that we can't use AI to solve this problem is very strange. There are patterns in the data that potentially AI is better at in identifying than have to do the super stupid way of trying to check out every permutation possible. As for the random circuit sampling, you would never build a solution like that. So he believes also that even if you were being able to scale it, it's questionable if it ever would have a useful value. So, from two angles, I think it's super hard to scale. And even if you could scale, it's questionable that you find a value. And it's not just me saying this, it's a lot of other people.

Henrik Göthberg:

But it's cool technology, but then you have to balance your sort of pessimistic, a pessimistic angle. We can also argue that actually we are going into quantum physics, so we actually don't know the math. The math we know actually. But what I'm trying to say is that there will be points that maybe we get to revelation, how this is solved that we haven't thought about. Of course it should always be. So to research it, yes, to work on the problem. For sure To see it in the near future in a PC near you, maybe not.

Anders Arpteg:

I mean, if you take them from an investment since we're speaking about doing strategic investments here and then we think from an EU level, what should you invest in? Should you invest in quantum computers or in AI? And I heard a person I met, a member of parliament in EU, once, receiving a question of you know, why did you make you know these billions of euros of investing in quantum computers? Why didn't you put it in AI instead? And the answer was so idiotic. I must say so I get a bit annoyed even thinking about it. But they say, basically, since the tech companies are investing so much in AI, let us do something else that they're not investing in. But then Google are investing in. But still, it was the most stupid argument I've heard for a long time, saying we invest in something else just because no one else is investing in it. We could just invest in astrology as well. I mean, that's not an argument for doing it.

Henrik Göthberg:

We had this conversation with people on the pod how much is invested in AI machines, how much is invested in HPC in Europe and a fairly big price tag on very experimental research for quantum around the corner, and we've been even arguing well, it's interesting, but maybe we can go to neomorphic computing. We can go one step. You know, if we are here now in the von Neumann architecture, the traditional architecture, what is the feasible leap? Maybe new morphic? That might be a 10 year horizon Quantum. The argument here is what is the timescale to solve this? I think this is underestimated. This is what I hear.

Goran Cvetanovski:

By the way, what is your thought on China Chiefs quantum supremacy claim?

Anders Arpteg:

Google even made a supremacy claim back in 2019-ish.

Goran Cvetanovski:

But what do you think about this one? This is from this week.

Henrik Göthberg:

This is all the same story, right. Of course I haven't read this, so I shouldn't say anything, yeah.

Goran Cvetanovski:

I was speaking today with Christian Goodman was here. We were talking about quantum. So it is very interesting if this is true or not true, but if it's true, it's actually much better than Google is doing.

Anders Arpteg:

Yes, yes, very questionable, but I haven't read it. I shouldn't say anything about it.

Goran Cvetanovski:

Exactly so. I will put it in the list there, and then I can do it.

Henrik Göthberg:

It's actually fresh from this week. So what is it called?

Goran Cvetanovski:

China. Basically, researchers develop a quantum processing unit that is one quadrillion times faster than the best supercomputers on the planet One quadrillion times faster than the best supercomputers in the world.

Anders Arpteg:

This is the Dennis Hassabi argument. This is such a stupid argument. It's just assuming that we would do the classical computer implementation in the most stupid way we could, which we never would. But if you do it in the most stupid way possible, then potentially these numbers would back it, but that's never how you do it Marketing-wise is great.

Anders Arpteg:

It's quadrillion times faster than the best supercomputer forget about 10x I want quadrillion x actually just a few weeks ago had another ship, and quantum ship as well, called the mayurama or something yeah anyway, let me take a lightweight news.

Henrik Göthberg:

did you see blue? Did you see the robot blue? Yeah, yeah. So disney, what at the jensen at an nvidia launch, launched the collaboration.

Henrik Göthberg:

Disney, together with the google deep mind and nvidia, has been working on what is actually the newton physics engine for robots, and and the. And the core topic is about to make robots more expressive and act more human-like or to interact better with us humans. And, of course, if you can show the video, it's interesting to see. I mean it's full-on Star Wars blue, it's full-on like RTD2. Yeah, and what we're looking at here is the expressiveness of the robot. That's what the Newton physics engine is doing, and what is it? Disney is going to use this in their theme parks to improve the experience, having cool interaction for the kids at Disneyland. But the innovation here, the technology here, is this engine. It's called Newton, it's a physics engine for robots and they're going to you know. So let's see. But any remarks on this? You know, any reflections on? Why do we need to have a physics engine so robot is moving more human-like? Why do we need that? Why is this so robot is moving more human-like? Why do we need that? Why is this useful, or is it only gimmicky?

Anders Arpteg:

I didn't read the details about all the GTC announcements, but I think Newton was actually the physics engine. It's actually similar to the digital twin you spoke about. So in some sense you need to build robots, have the ability to build a virtual digital twin for the robotic physical environment, and to do that it's useful to have a physical engine like Unity or something else that you have for games. But now they have something similar. I think what they actually did also was they open sourced, I think a model for robots as well, and that I think actually was a really interesting robots as well. Yeah, that one, and that I think actually was a really interesting. Let me see if I can recall it now. But they they called Groot or something.

Henrik Göthberg:

Yeah, right, yes, groot, groot. There's Groot and the Newton physics engine, and it's literally trying to build an open source model. I mean, like in order for you to, for robotic applications to be able to, you know, train them more.

Anders Arpteg:

I don't know how to explain this, but I think it's interesting with Groot and I'm paraphrasing now and I haven't really read too much detail here, but I think what they did in that one was something really interesting they actually went to the latent space kind of thing.

Anders Arpteg:

So you know other robotics solutions like this is interesting for Jan-Le Kun as well. They have their solution for moving into energy space for doing this, and we have Optimus Bot also working with this in different ways. But what they did here they actually had the system one and two. Yeah right, that was the thing. So they had a system one part and a system two part, and if you remember the Thinking Fast and Slow book on Kenman and others in the past, like the 70s, system one was basically the reaction thinking fast kind of part of the brain that we have. Then we have the system two is doing the more long-term, deliberate kind of conscious decision-making planning that we can have, and they did something similar. For this you had inputs, which was a transformer as well, combined with text, if I recall correctly, and then they moved first to the system two part for the planning of how they should move the robot, for example, and then the second part, the tokens basically coming into some latent space after the system two and then the fast part.

Anders Arpteg:

The system one takes over and decides what actions to take, given the output from the system two. This is similar to like a self-driving car you have the perception, you have the planning and control. So the perception then of course takes. Have the perception, you have the planning and control. So the perception then of course takes the sensory inputs. You move it through the system two part and you get some kind of latent representation where you can do the planning. Then from this one you do the control, which is basically the decision on what actuators to move. So this is actually starting to move towards what we've been preaching for a long time to do reasoning in latent space.

Henrik Göthberg:

I think it's super cool and maybe another way to reflect on how impressive it is what we're watching. Everybody's seen the dancing robots from Boston Dynamics as an example, right, and here, of course, that has been built with a different type of machine learning or coding that is more business-based, like the traditional robotics, so not so much machine learning at all. So business, you know, coded robotics control system, you know.

Anders Arpteg:

It's hard-coded rules.

Henrik Göthberg:

Hard-coded rules right, and they're becoming more and more and more sophisticated, and we're now what we're seeing Not done that much, but what we are seeing here is a fundamentally deep neural network learning how to do things, and an engine that learns faster and then and can also not learn to walk but also work in an expressive way. So it's it's. It's a deep neural network, it's not rule engine, it's not instructional in the same sense. And then the way it's built, it doesn't only act with precision, it also acts with expression. Kind of cool.

Anders Arpteg:

I saw some video of not the blue but the other like humanoid kind of robots. You can see especially the hand precision in being able to grasp or grab objects and if you compare that to the Optimus bot that Tesla has, or the figure AI or this Unitree other Chinese robot companies, I didn't think it was very impressive, I must say.

Henrik Göthberg:

But I think the one thing that stands out here is the fundamental expressive the way it bubbles his head. Is he stupid in some ways? It looks cool but it. But if you think about what it goes into making that happen so naturally in real time, I think it's quite cool.

Anders Arpteg:

They didn't use like a, not even english language. It was some kind of weird I don't know the details anymore.

Henrik Göthberg:

And then we need to go deeper. I you know. We need to go deep, deeper, deeper into the Newton engine. What that is all about, I don't know.

Anders Arpteg:

Anyway, cool that they are moving in this phase, cool that they are open sourcing things.

Henrik Göthberg:

Any reflections on robotics? And you know the expressiveness. Or does this give you any creeps? Or is it fun, or is it the right way to go? I?

Ingo Paas:

think it's kind of frustrating in some respect because you know we're working on it for so many years and we haven't really. It's kind of frustrating in some respect because you know we're working on it for so many years, so many years, and we haven't really seen this kind of breakthrough. But maybe, maybe it's coming when I, when I, when I go back into my history when I was seeing the first Star Wars movie I was quite some years ago, so it was pretty much the same image, but it was a. It was a. I was not really a robot and not so much intelligence in that machine, but you got the impression, you know it's possible.

Henrik Göthberg:

So what they did in the seventies and in the expressiveness, this is kind of where we are now in movement. Yeah, anyway, yeah.

Anders Arpteg:

It was actually I also had the GTC kind of thing what they did there. But yeah, let's not go.

Henrik Göthberg:

Let's continue. We have so much to talk about. Let me round it up.

Goran Cvetanovski:

Okay, round it up, this is like a very rounded up because I think what is happening in a part two of this interview is very much connected to what I think is the future of, let's say, the next war within these things, and I think that is actually for sure. It's robotic and it's no wonder or strange that actually, like, jensen is coming on stage with a blue and I just like, sit here and you know the robot comes there. We know that Tesla, we know that NVIDIA, we know Intel, amd, boston Robotics and Google now with the latest Gemini family right, so it's Gemini Robotics and Gemini Robotics ER. I think it was so. More and more organizations are actually focusing on these models for robotics and we can see that actually everything is going there. You remember a couple of seasons back, we were arguing where Tesla's money is going to come from by 2030. And we were discussing that approximately 70% of that money will come from robotics, and NVIDIA has been pushing hard in robotics for some time now, actually for two, three years. It's super interesting coming back to the topic that we're going to have in the second part, but I want to just like I cannot hold to roast them a bit.

Goran Cvetanovski:

So I was expecting that there's going to be a very big rise in NVIDIA stocks after this GTC because they announced that I think it was more than 3.6 million orders of the Blackwell computer and et cetera. So I expected the stocks to rise and of course it didn't. So it's still on 180. There was a little bit of spike. As you can see, this is in five days. So usually around GTC you have quite a lot of instability depending on when he's speaking and how he's speaking. So there is a little bit of upswing there. But let's see how it's going to go.

Henrik Göthberg:

You can track Jensen's talks, yeah.

Goran Cvetanovski:

But the worst thing is that. So, for example, I think that Intel has like a new CEO right now who wants to revamp the entire organization because they have been very late with the AI journey, etc. Right, so I think that they will be putting a lot of effort in this one and their stocks is actually like plummeting. So should I buy Intel or not. I don't know what to buy right now.

Goran Cvetanovski:

It's not enough to provide anything at this point of time. But this is actually what is happening. If we're looking at Intel, for example, for a year, you can see that it's been declining quite a lot. But let's finish with the banger. I think that is the most important, because Tesla.

Ingo Paas:

Because, Anders loves Tesla so much. I don't want to see this.

Goran Cvetanovski:

Here it is, welcome to 70% down. 70% down, it's not 70. It's from 480, I think it was the maximum to 234.

Henrik Göthberg:

So they are coming to the flat line, which is so zoom out a little bit so we see, because it's the whole thing that went up this year has come down again.

Goran Cvetanovski:

Yeah, so the maximum was actually on 444,. I think, 45, right 48, and now it's on.

Henrik Göthberg:

So if you look now, so we are down to levels that's sort of been the average line since, I guess, 2021.

Goran Cvetanovski:

2021, yes yes, that is actually when you find it. It doesn't mean anything because it's going to bounce back, but um, you know the the it's very strange thing.

Henrik Göthberg:

Is it a buy request on Tesla or is it sell as fast as hell?

Goran Cvetanovski:

well, if they focus on a, robotics is a buy.

Henrik Göthberg:

If not, if you look bigger than the car and fuck the next Model Y and talk about Optimus. Maybe it's still a buy.

Ingo Paas:

Without making any statement in terms of buying or selling. I think that if you do not trust a brand….

Henrik Göthberg:

That's the point I mean. So maybe this is a much bigger topic.

Goran Cvetanovski:

This fall is representing. It's representing… I mean, I had like the Nazi cross on my car the other day. Somebody came and wrote it on my. I have a Tesla right, so I'm picking up the car. I'm going to go buy some groceries and somebody put, like the swastika, on my car and I'm driving like hey hey, hey. And we're going out. Somebody put the swastika on my car Like in the dirt.

Ingo Paas:

Yes, In some countries you can get jailed for that.

Anders Arpteg:

Yeah, yes.

Henrik Göthberg:

Yeah, true, it's a deeper question, because the way to predict this now, up and down, what's the potential in the company? Have you fundamentally lost trust in the company or in the management or leadership of the company? Then no one knows what the market can do.

Ingo Paas:

No, and I would not comment specifically on any brand here. But if you want to build robots and you want to have a robot in your household, you really want to make sure that you trust that device. Awesome.

Anders Arpteg:

Well, let's get back to our awesome discussion no comments dingo and then jumping a bit more into the ai context here as well, and we have so many things happening now with generative ai and agentic capabilities and reasoning starting to improve, even though they are worse than humans still to a great extent, I would claim at least. But perhaps we could phrase the question as what do you see for potential business value potentials here in these kind of capabilities that we're starting to see in AI?

Ingo Paas:

I think that most companies are looking at AI more from a Mickey Mouse perspective. So you know how can we use chatbots to be more productive in writing reports or these kind of things. You know. I think the problem is that the traditional way of looking at AI you know you are a data scientist and so much more so the traditional way of looking at AI you know you are a data scientist and so much more so the traditional way of building complex things is not what we might want to try to do in the future. So when looking at AI, I really see the opportunities by using it solving business problems differently. Using it solving business problems differently.

Ingo Paas:

We talked about the green cargo case. So I'm not a specialist in the field of AI, but what I'm seeing is that AI can help us solving problems that we would have never been able to solve, helping us to solve problems that we cannot even specify. Ai will help us to solve problems without us building major applications. So, as you said, we do not have to define the rules. So a robot that works based on rules is just a machine. It's a very stupid thing. It can act, but if you have a learning capability reasoning then you start understanding things differently. So at Green Cargo, we have complex problems and it always starts with data, because without data there would be very little AI. So in the perspective of applying AI, we are using AI, buying it in built-in product solutions from a US-based company. It's called Cedar AI and they help us to do good stuff, but it's all built stuff that we just apply to our data and then we run the algorithms and pay a lot of money for it.

Henrik Göthberg:

In my opinion, that is for you is the first step.

Ingo Paas:

Yeah, this is what everyone could do. You know you do it with a pricing engine, so we've done this at the pharmacy at Apotheke Yardhart when I was at ICA. So we've done this on pricing optimization. So this is just. You know, you buy an engine. It's not really.

Ingo Paas:

I would not talk so much about AI then, but the thing is that, when it comes to innovation and scalability the things that we discussed and now I'm getting back into this kind of thinking the idea is that AI needs to be part of the scalable environment. So if you consider composability as important and ambidexterity as the fundamental thinking of your enterprise, the way you want to do business, it helps a lot to see it as a platform and you need to find the components of the platform, and the components are, as always, data. It's a good intelligent database technology, such as Knowledge Graph, so we are looking into Neo4j as a very capable tool. But then we thought about the business problems and what kind of generative AI model we should use, and one of our architects came up with the idea. You know, look, we have these problems that we can't solve in traditional ways, so we don't know what to do.

Ingo Paas:

He looked at Neo4j, he looked at OpenAI's API. So we used the large language model, trained it against the data that we put on the Neo4j in a sandbox environment, and suddenly we found out that you know well, we didn't know that it was possible. So the combination of empowering with Knowledge Graph and generative AI was really giving us opportunities to solve business problems that we would never, ever be able to solve and it would cost us a hell of a lot of money. We would never be able to solve and it would cost us a hell of a lot of money. We would never be able to specify the problem. So so so that was that was. We've done this in the sandbox of mine. Now we are in going into production and I can talk a bit more about this in a minute I'm super interesting and, just for you know, people that may not be familiar with neo4j.

Anders Arpteg:

It's like this swedish company to start with, so should be proud about it. I proud about it, I think. But it's basically a graph database company, right, which actually, I know is investing a lot in AI now and see how they can leverage their knowledge graph that you can build on Neo4j by also taking advantage of generative AI in different ways. But perhaps you could explain, could you give some example, a more concrete example, of what did you actually use with Neo4j and OpenAI to build some kind of solution?

Ingo Paas:

Yeah, absolutely. So we built a solution which is track and trace. Sounds like well, track and trace so that we can see where the trains are moving and how they're moving. That sounds like well, that is something. That is something that you can do it on a package. So what should be the problem with a train? It is so much more complicated.

Ingo Paas:

So what we've used? We've used the Knowledge Graph database and actually built a solution that our CEO, 20 years ago, promised to deliver, which was impossible since then. And it was possible because we used the knowledge graph and we retrained the large language model against the database. But there was a simple problem. But when you see the richness of the information, the details presented in that visual solution that we built on a local platform, so combining the graph technology, large range model we trained against, and then also our local capabilities with the analytics platform, so we brought a number of platforms together and solved the business problem that was unsolvable for 20 years and we are the first in Europe having that capability. So that's cool. But we also use that to solve we call this our operational problems when running trains.

Ingo Paas:

So when a train is delayed, it has an impact on the number of other trains and all of these trains will be impacted. So there will be a delay in the network. That is like a snowball effect Compounding effects, yes, and now the other solution that we built is by using the large language model. We trained it against the data in our graph database and that is giving us the opportunity to see the impact and decisions that we have to make first the impact and then decisions of this delay on this problem that is occurring and seeing the implications of that single problem into the network and that is giving us so much flexibility.

Ingo Paas:

So we built another application to visualize this for the end users to make it easy, because they should not think about using an AI tool, about using an AI tool. They should not think about using a graph data base, because it's not really sufficient for someone being out in the field or the one who has to make a decision in a second. Instead of using seven to eight different complex applications, including mainframe systems, to gather the data on one particular thing, they now can do it instantaneously and see how it's spreading into the entire network. So this is the kind of evolution from spending about 30 to 40 minutes to find the data, in best case, 20 minutes and then just knowing about a problem that happened 20 minutes ago, rather than getting it immediately and understanding the decisions you have to make and getting recommendations and also learning from the decisions you've made before. So this is one example we're now looking into.

Anders Arpteg:

Can you just perhaps elaborate a bit more here? I mean, for some people you know how to train a large language model. It's kind of straightforward and you have like pure text or even images to train on. But from a structured data source like a knowledge graph or the graph data that you have in Neo4j, it's not as straightforward. Do you know any more details about how you took the data from Neo4j and actually used it to train a large language model from that?

Ingo Paas:

Then I would have to call Anders, so it's probably it has to do with the name. But what we've done is we loaded all of the critical business data, operational data, so the entire infrastructure data that we can get for all of Europe. We loaded all of the information about our wagons, our locomotives, the people working on our rail yards, but also the people driving the trains, their competence profiles. We loaded the customer data, contracts, the customer orders, the information from the Swedish traffic record, weather information and you name it the cost for electricity, whatever you need to run the business our timetables into the database, and this information is helping us, when we were running the LMM against the database, to find out how the relations are actually impacting operations.

Anders Arpteg:

So you've probably encoded the data in some way. So that's it. I just wanted to learn from it, okay, and the way you used it is basically you can query it. Given the information about what happened in the last 20 minutes or something, you can make it predict what's going to happen in coming minutes and how that is potentially going to be propagated throughout the network.

Ingo Paas:

In that specific business case. Yes, we can see how the disruptive event is having an impact on our situation in the rail yard. We could even in different cases, looking into you know what capacities do we have in terms of the resources needed to rebuild the trains. We could also see you know how much space is on an incoming train so we could rearrange the trains. We could also see how much space is on incoming trains so we could rearrange the wagon. So there are different kind of solutions that we are going into the deeper questions as we learn. So we are just about moving away from the sandbox environment. We've put the first case into production and the business is now using it. But we're also giving this um at a later stage to our customers, so it's like but they will not be in a situation to write, cipher and, you know, do prompt engineering in a complicated way.

Henrik Göthberg:

So we are building the applications with the other infrastructure on platforms that we have can I untangle this a little bit, because you started off saying well, we see a lot of mickey mouse ai right now, and my take on Mickey Mouse AI is that we have consumer grade chatbot style interfaces that people can use for task augmentation, for personal productivity Like I can get some help writing or whatever you know and then we say we start implementing them and we even have an enterprise license for open for you know, but it's not a fundamentally strategic understanding for what we need to do as a compound AI system to make this work, and what you are now describing is that you are trying to sort out the fundamental building blocks of AI compound platforms.

Henrik Göthberg:

So, okay, so how will we manage data at scale and how will we make different data relatable to each other? We think we need maybe Neo4j technology here. How will we drive this into a fundamental or green cargo LLM, which is sort of a brain for us, and, based on that, how can we now take it out into different types of front ends where the data is presented in such a way so it's useful within the decision data workflow of a certain person? So this is not simply a chatbot interface, but it's like what are the key components that we are now adding to our platform philosophy.

Ingo Paas:

Yeah, it's true. I think you've described it very well, because this actually like we're not building application, we're building interfaces for the users to make the data, to give the data a certain meaning.

Henrik Göthberg:

Yeah, and if you now imagine you're trying to take a point solution on each of those everythings you would forget about. Oh, we maybe would use a graph vector or vector database used for some simple rag, for one application, and you would build an app on that. And then you would do it again and again, and again and again. It's still Mickey Mouse, because now you're trying to build something that becomes an ecosystem around this.

Ingo Paas:

Yes, we call this solution a digital twin because it is a digital twin and it's actually representing the entire network, including all of the train operations, in real time. So real time is the definition of the frequency of how we are updating the data, the data streaming part, in our analytics platform. But in this perspective, the digital twin is a digital copy of our business model in real time and it's telling us everything. It is going so far that in the visualization part that we built for problem solving, which is not implemented we are now testing it and verifying it you can see the driver of the locomotive and the telephone number of the locomotive. You see the exact position of that locomotive.

Ingo Paas:

So it's things that were completely invisible and impossible before to communicate. So even small things follow and make everyday easier for the user. It's not just you know the big innovation, it's sometimes just the small thing that makes it much, much better in the future. So it's a huge thing. And there's another thing that we will see. It's like in Swedish it's called huge thing. And there's another thing that we will see. It's like in Swedish it's called root compatibility. So to make sure that all of the conditions for running a train in a certain track, in a certain direction and between two destinations is approved in line with the rules and regulations, which is very difficult because there are so many parameters and each wagon individually, with the load and all of the other technical details, needs to be considered in that calculation.

Anders Arpteg:

This system can spot it out like this and it's not hallucinating when doing so, I think.

Ingo Paas:

We will see. So it's like we're gonna have to do all of the risk assessments, we have to look into it. What we can see is that we will have risk, we will have issues, but we know that it will be so much as the risk will be different. There will be new risk, but we will eliminate a lot of risk that we have today without knowing.

Anders Arpteg:

And, if I may just challenge your question or your presumption here, a little bit about Mickey Mouse as well, because I heard someone phrase it.

Anders Arpteg:

I'm paraphrasing now but I think that the big innovation comes when you really disrupt the way you're working. You're changing, perhaps creating new business processes you didn't even have before or redid the existing one in much better ways. I know even, like in Spotify times, we can say that the Discovery Weekly was a disruption in terms of how you build a product, because it's doing something that was completely impossible without AI. But then he said also that, yeah, well, it's much harder to do the disruption part, it's much harder to build completely new business processes and it's much harder to do the onboarding and change management necessary to really change how the business works today. So the best way would really be to start with optimizing the current processes, because it's easier to have people become more productive potentially with the existing way of working without having to change all of the current processes. So it could actually be good to have a two-step process where you start to optimize existing ones because it's easier and not requiring the same change management, but then later do the big disruption. What do you think about?

Ingo Paas:

that I think what we're doing right now is we are optimizing existing processes, because the only difference here is that we never, ever, had control about processes. So we were managing every day, every hour, every minute. But we've done the same process, but very manually and very, let's say, not with the intelligence that we now could use. I can give you another example when we are running trains and the train is late, we are getting a notification from Traffic Record and then they identify the reason for this delay. So this is defining who has to pay for the delay either it's Traffic Record, packet or green cargo.

Ingo Paas:

So we built an application before with low code, to give two people the opportunity to look into those cases and organize them With the graph technology. The graph is actually understanding all of the delays in real time everywhere in the network, organizing those delays and prioritizing them and giving us the opportunity to give all of the data that's related to this specific case two traffic workers in real time, more or less, if we want to do that and that is completely different to having two people sitting there and then he's calling. What was the weather situation on the 25th at 8 o'clock in the morning at this specific location, what was the cost for electricity? But now you get all of this prepared, prioritized and organized. I used to follow up on what André said here prepared, prioritized and organized Package ready to sign.

Henrik Göthberg:

I used to follow up on what Anders said here because we used the word Mickey Mouse, and I think the distinction here is actually you are improving on existing processes and you're reinventing them in small pieces, and maybe what I put out when you said Mickey Mouse was more that it is not enterprise-oriented. It's like more personal productivity.

Ingo Paas:

Yes exactly.

Henrik Göthberg:

So that was the difference and that might work for some stuff, but it's not really changing the company. It's changing me as an individual, like I'm using a laptop today.

Ingo Paas:

Yes, you have more fancy images in your PowerPoint presentations. Suddenly, it's a train.

Anders Arpteg:

I think you're underselling it.

Henrik Göthberg:

You're underselling it, but I like it's also built with personal productivity right. But I like the very much distinction between personal productivity let's not call it Mickey Mouse, let's call it personal productivity versus enterprise grade. I think that's the distinction we're talking about here Exactly versus enterprise grade.

Ingo Paas:

I think that's the distinction we're talking about here. Exactly, it's right. So Mickey Mouse AI is not negative, but it's on a completely different scale.

Anders Arpteg:

So I think that as you say, anders, imagine the 2% productivity gain for all the employees in a company.

Henrik Göthberg:

It could be a huge saving right yeah, absolutely so you're not allowed to say Mickey Mouse, you need to say personal productivity. I will not say Mickey Mouse, say personal productivity.

Ingo Paas:

Look, I think I fully agree with what you're saying, anders, because you know if you've raised the productivity, but you know you need to look at the kind of stuff you have in the company. So if you have 2,000 people and 200 work with administrative tasks, you know you would only have those 200 using it and most of the productivity you can do with AI.

Anders Arpteg:

I would actually recommend if someone would like to look more into this. There were actually a study coming from Stanford just the other week, like two weeks ago. They looked into the current way of working and different processes and how much generative AI could help them in terms of gain and it was actually much more than people have seen before. And people before have said like, ah, 5%, 10%, but they were speaking about like 60% improvements in a lot of these cases and it shouldn't be dismissed too easily. I think no, no absolutely not.

Henrik Göthberg:

But to curb your enthusiasm. Because if I go to very large enterprises and maybe you have such a complex system, so you also have what academically is referred to functional stupidity. So when you have processes where they don't have the fundamental feedback loops needed, so you are making decisions based on a subset of data and you're not respecting how many different feedback loops goes into a decision referred to as functional stupidity or the unaccountability machine. If you read another book on these topics, this is a fundamental problem that even if I raise the productivity of each person and I'm fundamentally then improving productivity and I'm fundamentally then improving productivity, I'm not really improving intelligence, so to speak, in the enterprise. So I think there is a huge case for the personal productivity gain, but the personal productivity gain will not fix the broken underlying lack of intelligence in some decisions.

Anders Arpteg:

Everyone wants to have the big disruptions, yes, so I think that's the goal in the end, of course, but I think it shouldn't be problematic to start in an easier way and get people used to using AI and actually having them adopting that technology by itself.

Henrik Göthberg:

So, even though you want the big disruptions, perhaps it's not best to start with only that no, and I think this is a profound comment, because in order for get people into seeing the broken processes and all this, they need to understand the art of possible and literally, by becoming more data and AI literate in your own bubble, you start seeing the enterprise opportunity in a different way. So this is a big, big point. I agree with that, yeah.

Anders Arpteg:

And I think you know also, it can be useful to just have people understanding the possibilities of AI and thereby providing more innovation. So I think back to your point either we have a data science team somewhere that came up with you know, very advanced solution, pushing that into business in different ways, or you learn the business to become AI literate and they came up with innovations and the best way to create new processes and that potentially that could be even better. Yeah.

Ingo Paas:

But I think that you know, looking at the things that I was talking about and the way we look at the digital twin, the digital twin is nothing that is owned by IT.

Ingo Paas:

The digital twin is a democratized capability that everyone can use and when the business has a problem, they explain the problem and someone helps them to reorganize the data to solve that problem.

Ingo Paas:

And if that requires because there are hundreds of people involved and they want to have a simple interface, then we add a local interface to it. But they all have access to the data. They can also work with Cypher, they can do prompt engineering, but the difference here is that it is not about the individual working with email or working with a report or working with the next mathematical problem. So we are turning the entire business model with all of the operational data into our AI core and giving everyone the opportunity to use the interfaces around that core. So when you digitalize and that comes from what you said, from the composable idea that you build all of these capabilities, so you make it all accessible, and that is one of the key things in the analytics In the beginning we were building the reports and then we said we stop doing that we do a data engineering. That's the only thing we should do at IT.

Anders Arpteg:

Good, because I think I also like the way you speak about platforms and saying actually these kind of you know Gen, ai possibilities and reasoning technologies can be a platform for other people to use as inventions and innovations on top of it, and that is really when you scale the value. I would say if you come to that point.

Ingo Paas:

Yeah, and this is the plan. So we are now looking at this. So the combination of the generative AI capabilities with our Afing Neo4j graph knowledge database, in combination with the other parts, so this is a new platform that we can scale, because we talk to HR, we talk to market people, salespeople, they found out they can sell more capacity with it. So this is the same platform, it's just, you know, we're using the same data, we're using the same logic. Everything's the same. So they can sell more capacity, they can look into pricing issues, so they can inform the customer, they can give customers, it's a proper platform.

Henrik Göthberg:

Yeah.

Ingo Paas:

It's multipurpose.

Henrik Göthberg:

But can I stretch this a little bit there's been in the. Sebastian, the CEO of Klarna, made a nice LinkedIn post where he sort of made a little bit more explicit explanation why are we leaving Sauce? And the backdrop is that a couple of months back it came out in US media that Klarna was getting out of Salesforce and Mark Panihov had to sort of defend that on stage. So Sebastian got super as he said himself embarrassed right, and what he's arguing for is like exactly these topics. We are starting to rethink our architecture, ai first. So basically the way we are, we do so this. So when people say platform that we lose them directly if they are not on your idea of platform, but if you. If we are starting with defining platform from principles and something evolutionary, then now we put the fundamental of data and how data can be accessible and how we can now leverage that in different ways. Now we are starting to build an enterprise architecture which is actually designed, composable in order to then do whatever we want on top.

Ingo Paas:

Yeah.

Henrik Göthberg:

And this is different architecture than an application-centric architecture, and it's slightly different even to a data-centric architecture.

Ingo Paas:

And I fully agree and I think what you can see, we actually hate SaaS, but we need SaaS as well. But when you look at, when we look at SaaS, we want to integrate it as much as possible and we make very strong statements that we never use their integration engines. We never use their data platforms, analytics parts. So it's only if there's pre-built stuff that is, just general things that we could easily use. But we limit the impact of the ecosystem as much as possible and really try to restrict it to the SaaS skill, like the applications built around many, many decades of knowledge in an industry, so something that we don't have. So we use it because of that knowledge is built into a solution. But then we integrate this and make sure we use the data. We only use the data and then support the business with the process. So SaaS is not good. We hate SaaS, but we make it work in our environment.

Henrik Göthberg:

But this is super important now because I think we now need to, maybe for the listeners, distinguish the difference between PaaS and SaaS, potentially right Very much, I agree.

Henrik Göthberg:

Because here we're talking about cloud technology and, depending on how we use it, it's infrastructure as a service or it's platform as a service or it's software as a service. And when we are saying software as a service, we are basically going into a pre-configured view or point of view of a vendor how you should work. But if we have principles, our own platform principles, what is our fundamental beliefs and values that we architect, engineer stuff, we are having a past view of which technology focusing our evolutionary platform, it might be that we're using the same technology, but we're not using it in a SaaS way. We're using it in a PaaS way, Sometimes to build a very niched application with a lot of domain knowledge. Fuck it, let's get this SaaS on top, but we want to decouple the data still, stuff like this, yes, Am.

Henrik Göthberg:

I summarizing it, because I think this is really power. It's a very strong statement to say you hate SaaS. That's what I'm trying to say.

Ingo Paas:

There's also another reason why I love PaaS Platform as a Service because it's my second name, paas PaaS as a Platform as a Service.

Henrik Göthberg:

You go.

Ingo Paas:

Platform as a Service. Yeah, exactly, I PaaS Infrastructure Platform as a Service or IP Internet Protocol. So my name is very much digital. That's the reason why I'm into PaaS. I cannot do it any different.

Henrik Göthberg:

I'm not a SaaS guy. I'm a SaaS guy, so that's why I hate SaaS.

Ingo Paas:

It's not that SaaS is bad. I hate it because of my name. But I have to be proud about something in my life, so that's my second name. But the thing is that you know, in terms of the definition between PaaS and SaaS and I fully agree a platform as a service from our perspective is what someone is offering us and then we select it from the components we use from that platform and we can refigure and we can rebuild it.

Henrik Göthberg:

You own the practice. You decide on the practice. Yes, not them. The practice.

Ingo Paas:

You decide on the practice, yes. And then the SaaS comes into the game. We have to integrate it into our platforms and we prioritize our platforms when it comes to integration, analytics and these kind of things. We talked to one vendor and I was saying we give you that PaaS platform, we give you the analytics platform, we give you that integration service. No, no, no, because it's not as good as ours. I was like not being arrogant, but we looked at it very thoroughly.

Ingo Paas:

It's the same with these SaaS vendors. When they come, they have the industry knowledge, but they are probably not good in building applications around it. So, the mobile applications when we look at their capabilities the business first well, it's everything we've ever wanted. And then we say but look, none of your users would probably use it because no one would understand that user interface. So we build it this way. Isn't that better? It's a different logic and at the end of the day, we're using our low code. So because we can build those applications better than what they are. When you buy the applications, in some cases it's not necessary because it's not a business priority or business value, and better could simply also mean in order to fit, be usable and adoptable in our purpose.

Henrik Göthberg:

We actually don't want to go down left here. We need to visualize the information in this way, because otherwise we need to redraw the whole map of how we work.

Anders Arpteg:

Yeah, the time is flying away here and we have some awesome topic that I'd love for us to be able to cover, and I think especially the one about your upcoming book potentially. I love to talk about that. Yeah, so please let us know it was called. What was it? Empowering Human Autonomy, was that the name of it?

Ingo Paas:

Yes, and if you can make it, fake it. This is my first book, but I put a new envelope on it. So, this is the new. This is the new book, but it's going to. The release is planned for and it's not a sales pitch, it's more. I'm going to talk about it very differently in a moment from now. It's going to be on Amazon, but it's late April.

Henrik Göthberg:

Yeah but your first book is on Amazon. This is called the Composable. If everyone is interested, what is your book? If they want to find your first book, what was this?

Ingo Paas:

Yeah, it's called Digital Composable Enterprises, and if you use platform service and you look for paths, or you add Ingo, you find it.

Ingo Paas:

But the thing is that this book is something that is based on an experience that I wish very few should only make, and that's back 22 years ago when our kids were born. It's always a very heartful moment to talk about it, but it's like the experience when you prepare for the greatest moment in your life having kids and then something very disruptive happens. And that was a moment where we've gone through hell and since then we are in 24-7 care for our firstborn sons. We had twins and it's a very special situation when you realize that there is no hope for improvement. It's only probably getting worse and without knowing what it is, and waiting 18 years to get a diagnosis and, as I said, it is a very rare disease, 18 years to get a formal diagnosis.

Ingo Paas:

Yes, and it just came in the beginning of the pandemic. So I got a telephone call from Kaolinska and I said we found it and they were really working hard on it for very many years. And my son is completely unable to communicate. He is unable to move, to do anything on his own. He learned to hold a spoon. It took years to teach him this and he can get the anything on his own. He learned to hold a spoon. It took years to teach him this and he can get the spoon to his mouth. And if we put food on it, this is what he could do. He would do it also without food, but chewing is not possible for him. So it is very specially prepared food, but it is the same quality of food that we have, so he could experience it in terms of everything else. So this is just an example of what he has learned in life. This took years for him To be in a situation.

Ingo Paas:

As a father of a son in that situation and his twin brother, you also need to make a lot of decisions, and my wife and I we had a lucky enough a very good discussion in the very early beginning to treat everyone equal and make sure that no one is left behind, but no one's left behind and having this kind of situation where you cannot support your children and you can never communicate, never exchange a word, it's very difficult to handle. It's very difficult to handle and then making all of the decisions, fighting for all of the small things every day and trying to be there 24-7 to support, to handle diapers. And you know, this is my reality and I'm working hard on being there with love and dignity and everything that needs to support us as a team. But there's one thing that is particularly hard. This is the idea of leaving one day without being able to support my son anymore and at that particular point, knowing that he might not be treated well, because I know how much it means for him to be there with love and to be close to him. I've learned how much he appreciates this and how he reacts, and I hope that this is going to be like this when I will not be there, when my wife will not be there anymore, and solving this problem is impossible.

Ingo Paas:

And I found a way and I was working on that problem for very many years and I'm not technically advanced in terms of you know, you have that great advantage of knowing AI, but I've used the knowledge I have to work on bringing love and technology together to find a way to think different. And the way I look at assistive technologies as of today. They are pretty much like enterprises react they solve individual problems. But they are pretty much like enterprises react they solve individual problems, but they are they are fragmented. They are not this. They are disconnected and they are often built around the knowledge. These people have meaning that the horizontal and vertical layer does not exist for those technologies. So this was this was my starting point.

Henrik Göthberg:

I had to solve that without knowing how so you're trying to figure out or you've been something like a 10-year thinking process, thinking deep on what happens when I'm not here with my son.

Ingo Paas:

It's like what can technology do for him to make him more autonomous, even if so many things is not functioning in this sort of physical approach, yeah, which is the same as of today, because you know he needs to be protected by the ones being with him, and that requires that they love him or at least appreciate him very much, to be there and to make sure that he's doing well, because he will not be able to communicate, share and say no, or say yes, or tell me what has happened or what hasn't happened.

Anders Arpteg:

We spoke about this in the beginning a bit and I thought it was, I think, a very profound solution or way forward here that you spoke about the personalized digital twin. I guess this is what you're referring to.

Ingo Paas:

Yes.

Anders Arpteg:

Can you just elaborate a bit more? How would that work If the horrible day comes where you cannot be there for him? How could he potentially, or how can you help him continue to have his preferred way of living?

Ingo Paas:

Tell us the vision I can tell you. In my book I looked into the. I was describing the thoughts in my mind to share it with others, to explain how I think about the opportunities with assistive technologies and turn them into more responsive, humanized, personal digital assistive technologies. So the idea is that, instead of building fragmented solutions, I want to create a digital twin. Building fragmented solutions, I want to create a digital twin. The digital twin is the vertical part of the innovation and that is connected to the individual. The digital twin is part of an agentic network with agents. The agents are creating a relation could be a robot in 15 years from now, in 10 years from now. That is a connection to a healthcare system, but that is mainly the connection to the caregivers and people around this individual. So the digital twin is connected to all of the caregivers around and building that ecosystem that is protecting the individual from all of the harm that could happen. Protecting the individual from all of the harm that could happen. I have developed around 10 global use cases. They are very specialized and detailed and explaining in different ways how these kind of capabilities could help in the future.

Ingo Paas:

When you look at the power of reasoning, ai and the contextual ability, the contextual situation that the AI could understand and interpret and also translate into actions. This is where sensor technologies come in. So biodata could be measured, so you could measure the state of the individual. You could find out and continuously repeat and measure the well-being. That could be any kind of data that you could use, so the data would give you an image about the well-being of that person.

Ingo Paas:

Imagine you would have a caregiver with him that is not doing well, so you would see that in the data, you would see it in the patterns and you would see this in a kind of report. It could be a dashboard and you could have a discussion with that person. Pretty much as you would have a discussion with a person if he would serve you and you would not be happy, so someone else could take that dialog. So that is one way of measuring his well-being with this individual, which could be very different to his individual, and well-being with someone else and you you used the phrase earlier, I think it was before the pod started.

Henrik Göthberg:

that I that, I think, is a way to you can, if you can elaborate that, to even frame what we're talking about exactly now. You used the phrase the ai for the human versus the AI in the machine. Yes, because what you're highlighting here is that something is very close to the human and we are trying to set up the policies for how this AI works. As the digital twin, it's an extension of the human, but in order for this extension of the human in this particular case, where the human is incapable, needs to be able to interact with machines, like systems, healthcare systems, whatever. And now is the distinction between we build AI for enterprise, so to speak, and here's the human digital twin that then can interact in machine to machine, so to speak. So this is the AI for the machine versus now the AI for the human. Did I get it right?

Ingo Paas:

Yeah, that's correct and I mentioned the healthcare system, but they are just, you know, a very small part of it. But it will also be important to you know to understand all of the examples of the details of the use cases. That could be like if you are in a room and it's too cold, you would change the temperature. The digital twin could actually do that for you. If you are in a room and the noise is too much, the digital twin could adjust the noise because it will recognize you're not doing well, it must be. The noise twin could inform you, as a caregiver, that, without the individual being able to say or identify that I need to go to the restroom because you can read the biodata. So you, you, you have diapers, but you will avoid to to to. You don't need them in reality, but you should have them so. So the digital twin could also do a lot of other things. The digital twin could be your lawyer in terms of taking over the administration of the processes where a caregiver when I'm not here anymore, my son is going to be very much dependent on the people who would take over this responsibility.

Ingo Paas:

Imagine you have someone who is not good in arguing, who does not understand law. Because this is what we need to do we need to understand law before we send an application, so to really make sure he's receiving the help he deserves according to law. So the digital twin would be the lawyer, his lawyer. He would monitor these kind of applications and, ideally, in a couple of years from now, you would continuously automate all of this. So the digital twin would be the lawyer, knowing all of the law.

Ingo Paas:

It would know his status, status and what you could do instead is that you would have a digital passport that would explain to all of the authorities that that needs to approve those applications for parking that could be for for for having a transportation services or whatever. They would always have the updated information. They would never have the updated information. They would never, ever to send an application again, because you would be in a situation that DigiTwin is owning that information. It's very personalized, it's very much in integrity of your data that you protect, and then you could send that status information to another agent digital agent and you would not have to do those processes anymore. So there are many more cases.

Henrik Göthberg:

You can take that into the whole. How do we have our identity back on the internet Even today, right?

Anders Arpteg:

Yes.

Henrik Göthberg:

Do I really the way you know? So there are so many layers on that. There is an argument for the digital twin.

Anders Arpteg:

Yes, and you mentioned a very personal, important one for your son, of course, for more physical and mental well-being or health issues. But it's easy to see potential other use cases right, which you're saying, a large number of them, which I'm not sure what the proper term is, but I think you phrased it well, with a digital twin for the person, like hyper-personalized kind of experience Someone really understands. I guess it goes back a bit to the Butler kind of metaphor, for a while back someone tried to phrase what is the optimal AI for a person. They said it should be a Butler. Why a, a butler? Well, a butler knows you very well, he understands your preference, he will review your mail and only ask and interrupt you if you really need to and want to. And having someone that really knows you very well and wants you very well would be an awesome experience in some way.

Ingo Paas:

I agree. I think it's a wonderful description of what is coming from my heart, because it's a lot of technology, but it's coming from my heart. It's very deep, it's very important and it's very sensitive. So the idea with the DigiTune is also recognizing the need of having a kind of very personalized experience for everyone, because you cannot set a certain definition of this is what the digital twin should do. The digital twin should do what you need, so it needs to be very, very Customized individual.

Ingo Paas:

It's extremely individual, so the digital twin should also see. What is it that the digital twin could do for you in the long term? So I think that the contextual data that the digital twin would own, it would own your agenda. It would know where you are. My twin son so Julian is his name he would be in a very different situation in in a couple of years from now if that would work, because he would always know how his brother's doing. He would always know and he could take actions on it without being physically on site. And we need to be there on on site every time have you followed any technology progression?

Henrik Göthberg:

Because I mean like the Butler idea or this, I mean like we even see the technology as R1, coming out, as the assistant, like the apps that owns all apps. You know we are getting tired of having thousands of apps on our phone, which is really highlighting that we cannot have more fragmentation, more so at some point when this is exploding in productivity and scale of information that we need to deal with. We need to have something that is us. This argument. I've seen different arguments on it, but have you circling back to the question have you been looking at technologies or are anyone working on this problem, building the digital twin or going in the? Do you see examples of fragments of this thinking?

Ingo Paas:

Yeah, I think that I was just recently talking to someone from Cote d'Ivoire and there are some attempts on particular problem areas to build kind of something similar, but the holistic view has not been addressed. So I've done so much research. Maybe there's someone else out there doing similar stuff and having similar thoughts. But if you look at the system as such, it's like enterprises. They're having many divisions, they all work against each other. They all have their different agendas. It's the same with the investments that we can see in assistive technologies. The doctor who is specialized in the field is building a solution that is solving a problem in his field. He is having a budget that is supporting a certain thing. He would never have built a horizontal infrastructure because it is not his problem and that means that it doesn't matter how good the innovation is on the individual. They are practically not scalable. They're fragmented. It's not negative. It's a very positive experience, because every improvement that you can do is a wonderful thing, but the thing is that we do not do as much as we could.

Henrik Göthberg:

So you're highlighting if you go into the industry of assess I mean, I'm sure there are numbers on how big the industry for assistive technology is and then you could take it, you simply into health care. You know you could find out hundreds of use cases, but they are very application centric. That's what you're saying yes yes, and they are platform centric, yes, where the individual is the platform and they're not digitalized because they all are connected to physical experience.

Ingo Paas:

so you have a lot of you know stuff around you and and it's it's and, as I said, the the most important thing is that we have to respect every single innovation, and every single innovation is great, and so every single thing that we can do to help people with limitations to have a better experience and to improve, it's wonderful. But it's not criticism, it's the need to do more and it's possible Also.

Anders Arpteg:

I think one interesting when they released ChatG, gpt or GPT 4.5, which is like three weeks ago or something, it wasn't really that much better in some benchmark, but it was supposedly much better in emotional intelligence and I was thinking first, this is connected to what you're saying, but I actually don't think it is, because you know what you're saying is more we need to fix the data engineering, we need to have the sensory data, we need to have the sensory data, we need to have the way to really understand and want to do the best for the person, and I think what Chatti or GPT 4.5 is potentially good is making you believe it is doing something good for you, but not really wanting you the best, and it doesn't really have that information.

Anders Arpteg:

It doesn't have a policy, it's just trying to do the best it can to doesn't really have that information. It doesn't have a policy, it doesn't have a. It's just trying to do the best it can to make you think it wants to. It's not bad in itself, but I mean what you're saying. I think you could find easier solutions than this that simply have the proper data engineering in place to be able to understand what the true feelings are, so to speak, or the true state in some way is, which is a very different thing.

Henrik Göthberg:

who would be the ideal um startup or incubator to run with this idea. Would it be google deep mine? Would it be donyalex? You know, I can see prevent, preventative health care and now individual I, I, you know what. What. How would you dream about you who should be, you know, invent, inventing. How should that be realized?

Ingo Paas:

I think potentially everyone who wants to invest is great to invest. But I think you know if you do not understand the purpose, you need to have a humanistic perspective. You cannot come with a technology perspective, you need to have a humanistic approach.

Henrik Göthberg:

This is what I mean. Who is the perfect innovator, incubator that has this humanistic perspective?

Ingo Paas:

I would never say no to Daniel Eich. I don't know him. I would never say no to anyone because no one ever talked to me about it. This is the first time for me talking about my book in this way, so but I think that having a humanistic and approach to to make this world better for those who are in need, I think that that would help a lot. And then, if you come with the ability of you know, you build a wonderful thing and you build a global capability, it helps a lot, because what I'm very much into is that I want to make that technology affordable, accessible, and it must be personal, affordable, accessible, personal. Yeah, it's very important because you should never, ever set the objective based on the technology. The technology should help you to set your objective.

Henrik Göthberg:

Well put, well put, but it's an interesting topic in so many ways.

Anders Arpteg:

What can people expect with reading your book? What will they learn?

Ingo Paas:

They will learn that I'm a dedicated father. I'm very proud about our family my wife, who was fighting a lot while she's the risk-averse person, I'm the other part of the story and my son, who is there with his love and dedication and never, ever complaining about a life that he wouldn't have chosen. And likewise my son, born with this severe problem, smiling every day and having fun, and he's laughing.

Henrik Göthberg:

He's laughing most in our family, so he's impressive and I want to make sure he will continue to experience life this way, um, but I think the book title um empowering human autonomy is actually very timely when we have this whole argument and we have the narrative of ai replacing humans or engineers or whatever it is, and actually we want to have ai agents.

Henrik Göthberg:

We have AI that makes us all geniuses or augments us, and I know, for example, sana, the AI Sana startup. They are doing a very strong they're launching in America right now and they are pivoting in terms of how you build AI knowledge agents for human genius, like use the marketing branding, and I think this is so important in the whole narrative of how we want to understand and steer AI. We want to steer AI to you know, not to replace us, but to make us shine, and if that is, you have an extreme edge case here, if I use tech jargon to describe your son, but in reality, it's a profound message how do we build AI and how do we tackle this so we end up winning and end up benefiting from it?

Ingo Paas:

Yeah, I think my title. You don't know how many hours I spent working on my title. Yeah, I think my title, you know. You don't know how many hours I spent working on my title, but I think it's like the title doesn't say so much, but it says so much Exactly as you just, you know phrased it, henrik. I fully agree, because it is an expression for every individual to live and embrace the opportunity with AI and to grow with it rather than to suffer from it.

Anders Arpteg:

Yes, and it's given the interest of time. I think we should move to potentially the last, more philosophical question and I think the title actually is very apt when coming to that Perfect segue.

Anders Arpteg:

You know, if we can go at some point have AGI that is better than most humans and potentially is on par with an average co-worker that's Sam Altman's basic definition of AGI, and we're far from it yet, but it's moving very rapidly and we could imagine, in X number of years, us having AI that is on par or even above humans.

Anders Arpteg:

And then we can think about it in two extremes. Either you go exactly opposite to your book title, which would be the Matrix and the Terminators of the World, where the machines try to control humans and even kill humans to a rather large extreme extreme, or potentially the other one, and perhaps that is really where AI is really trying to empower human autonomy, as your book title is where we actually are having control of the AI, and AI is working for us and our beliefs and perhaps is helping us to solve your kind of mental issues. Perhaps is helping us to solve your kind of mental issues we have curing cancer and other disorders and problems that we have, even climate crisis that we're starting to see and, who knows, perhaps fixing the energy crisis, fusion energy driving our society and so many more things. What do you think, ingo, the future will be? In which of these two extremes do you think we will end up in 5, 10, 50 years?

Ingo Paas:

You mentioned Sam Altman and the perspective of reaching AGI Whenever this is, the commercial AGI is probably happening faster than the global AGI or the universal AGI is probably happening faster than, say, the global AGI or the universal AGI.

Ingo Paas:

But if we I think he said it this way yes, we will reach AGI probably faster than we think, but until then we have great companies, and so that's a very, very funny way of saying until the day we have as much fun as we can.

Ingo Paas:

I look at it differently because I don't have his wisdom, so my view of looking at it is like you phrased it very well, so you both should continue writing your book, and it's getting really good, but the thing is that I believe in empowering autonomy. The thing is that I believe in empowering autonomy, and to understand how important autonomy is, you need to spend time with my son, because then you learn about what is needed to leverage someone who does not have autonomy, and if you understand what autonomy is, it goes all the way through life. It's nothing that is only a challenge for people like my son. Autonomy is also what's happened to you when you go through education and so on. So I believe that AI, when coming to AGI, it should empower us individually, collectively grow, but with the values that we want to share. And if this is possible, I don't know, but if it's a commercial AGI, it will not be this scenario.

Anders Arpteg:

That's a big risk for that right?

Henrik Göthberg:

Yeah, so the commercial AGI or the AGI that is not inclusive or, basically, is profiting the few. We have talked about AI for greed, ai for control and AI for good, so to speak, and if we want AI for good and AI for good has something to do with empowering human autonomy what are the things that we need to really start working on to go in this direction? Or how can I vote I say it sometimes vote with my wallet. How can we all try to push my wallet? How can we all try to push principle driven, value driven, making decisions here and now. That nudges. How can we nudge in the right direction?

Ingo Paas:

I think it starts with respect, integrity and the values that we need to agree upon, and it's not so much about the governance, because the governance is always the first thing people mention, but maybe you need to understand the why Deeper, yeah.

Ingo Paas:

So go, go, go, go deeper, rethink before you set governance, because governance is nothing else than the consequence of the acceptance level that you set how much you want to protect yourself or not. Maybe this is, this is where the, the interface is between you. So when we talk about it's my last chapter in the book the fusion between machines and humans, so when this fusion is happening, you need to make sure that you have a very clear and, let's say, working relationship between those two, and that is very much what governs. You mentioned another example, which is also very interesting to look into, but I think that if you would own a digital twin, and the digital twin would be the direct connection to the agent, the robot that you might have in your household that would take care of my son, that would be the governance that would come into place. When you've set the boundaries in terms of learning and values.

Henrik Göthberg:

That's where I would yeah, but because I think it's interesting. You talk about something here that you know how practically to get from governance into code, into the policy of your digital twin. At the same time, I know that you're also in some ways a little bit skeptical how we over-regulate in Europe. So on the surface it seems like, oh, you're in two different camps, and I don't think you are, because the problem is when we have rushed into regulation for regulation's sake, without understanding the deeper whys and the deeper you know, what are we? Starting with the respect, because this is really tricky, right, in one way it's not governance. In one way it is governance.

Ingo Paas:

Yes, now we talked about a lot of things that you know are very personal, very deep. It's emotional stuff, it's very difficult. It's 22 years of my life that I dedicated to create a better life for my family, but you know this is a question that I could, I think, answer again with football. You know, you can never get a team to play in the Champions League and win the Champions League if you don't give them a ball.

Henrik Göthberg:

No, that's it.

Anders Arpteg:

Give them a ball. Yeah that's a very symbolic measure. You need to have the freedom to do it.

Henrik Göthberg:

We have so many regulations so they don't get out of the dressing room.

Ingo Paas:

We have so many regulations so they don't get out of the dressing room. Yeah, but you cannot regulate the ball away from football if you want the team to play football, because then they can't play football.

Anders Arpteg:

Such a great ending metaphor. I think we need to have the humans having the ball in some way. Thank you so much, ingo Pest, for coming here. I hope you can stay on for some off-camera kind of continued discussions in so many interesting topics, but I also very much hope that we will have the vision that you have, I think, described very nicely, of a digital twin working for humans coming true, and I hope, especially for your son. That's not going to take too long and I got a lot of thoughts and inspiration from you, so thank you so much. Awesome, and also the episode.

Henrik Göthberg:

Awesome.

Ingo Paas:

Thank you, thank you.

Henrik Göthberg:

Thank you.

People on this episode

Head Shot

Anders Arpteg

Host
Head Shot

Henrik Göthberg

Host
Head Shot

Goran Cvetanovski

Co-host