Thank you very much, Daniel. And , um, before we, um, launch some European AI rockets into the sky, I of course want to thank Veronica and Fabian for inviting us here, for the kind invitation and for the opportunity to, um, share with you some thoughts on how Europe, uh, can confidently soar into the future of artificial intelligence, which I think we're all shaping in one way or another. No. So the report actually starts, of course, with the central observation, which Fabian also mentioned , that, you know, the major developments in AI are happening elsewhere. Yes. If you look at this, you know, I submitted the slides back on Monday , and, uh, maybe it's a little outdated now.
You know, there's a certain risk in these things, but it's more of a general theme. You know, there's a lot of development going on in the US, if you just look at, um, progress in certain areas of AI. If you think about programming and software development support, a lot has actually changed in the last, I would say, six to eight months. Yeah, in other areas maybe not so much, so this rabbit hole, you know, is pretty deep. Yes. And we don't know where it will lead. Yes. And, of course, you know, capital is being invested very actively, particularly in the US, but also in China, to dig deeper. Yes. And of course, we also know that even if we wanted to compete at this level, you know, it would be impossible in Europe for various reasons.
Yes. Um, so even if we look very positively, you know, at the overall development, the question arises: where is the niche for Europe? Where is the potential for Europe? And I would like to try to convince you that there is huge potential in Europe, and I would also like to convince you that this year—this is actually the year of takeoff, you know, AI in Europe, this is such an optimistic view . Yes. Sorry, I was told it was transparent. So, you still see me. That's good. Yes. So, of course, if you look at the public, you know, among us experts, it's clear that we see this potential positively, that we also believe, as Fabian mentioned, that of course jobs will change and transform, which will be associated with certain challenges, but it will not be a catastrophe.
There is also a lot of potential for new jobs and so on. But if you look at the latest AI Index report from Stanford, you will see that public opinion differs, especially in the US. In the US, you know, over 70% of people assess the global development of AI quite negatively. In Europe, the situation is still different, and I think, at least for us in Europe, it is extremely important to make sure that this does not spill over to the other side. Yes. In the US, you know, you could say that tech companies aren't too concerned about this. You know, this has always been the way the US moves forward. But for Europe, I think this is a crucial moment for us to develop AI in a way that embraces human needs, is society-oriented, without losing much of the powerful potential that AI offers us .
Yes. And now, of course, if you talk to politicians or researchers, the answer is definitely sovereignty. That is, to have some control over the environment, over the infrastructure, as well as over the models themselves. This is what Daniel mentioned. I think that's important too, and it's that combination that will ultimately make the rocket take off, in my opinion. And if you look at it, there is even more demand for European AI now than there has been in the last few years. Yes. And this, in my opinion, will lead to the moment of takeoff. And I've already, you know, traveled a lot over the last few months, you know? And, honestly, it was the most stressful time of my life in a long time.
Yes, and the reason is that I feel this pull, you know? And I want to share a little about what factors are the main foundations of this attraction. The first is what Fabian mentioned: talents. And talents, we know that there are plenty of them in Europe; We have a very good education, excellent research centers. There are always complaints, and I see that it's true, that in the past we may not have used these talents very effectively in terms of technology transfer , startups, or scaling new developments. But, to be honest , a lot has changed in the startup ecosystem , and it also has to do with talent. This is due to the ecosystems that have developed around research institutions, like Atum, Nimatum -- that's a good example, and we're starting to replicate that in other places.
So we are in the process of trying to recreate functioning ecosystems and use these talents much more effectively. On the other hand, this is also what I see as a university lecturer: Europe is becoming more attractive today for graduate students and talents coming from outside. Again, you can look at the AI Index report and see how the US is starting to struggle to attract new AI talent. They already have many artificial intelligence specialists. So it may not be such a big problem right now, but they are having a hard time attracting new talent. And in Europe, on the contrary, we are increasing our turnover. Yes. So people go to Europe, you know? If I look at my undergraduate programs, at my students, I see some stagnation or even a slight decline.
But if I look at master's and doctoral programs, I see a huge growth, especially among students from Asia and other parts of the world. So, there is a personnel reserve; it has always been there, but it is growing, and we can use it. And I believe that this is the most important factor at this moment. The second point is, of course, infrastructure. As I said, you know, even if we wanted to , we wouldn't be able to build such a huge infrastructure, like, for example, in the US or China, to be at the forefront of developing the latest big models and so on. But, actually, we need a certain amount of infrastructure. This is extremely important, so it's something like...
I can't say exactly how big this infrastructure needs to be , but I can say that it's not enough right now. Yes. So that's something we need to work on, something we need to expand. This is very important. Last year there was a kind of push, saying, okay, we're building this infrastructure in Europe. You know, I'll talk about that a little later. In my opinion, there is a certain slowdown now. This is what we must counteract. You know , this is important. We need this infrastructure here in Europe too. And the third part, of course, is data, which is, so to speak, the fuel for the rocket. And here too, if you look at the industrial data in particular, this is something that you are familiar with, because the US and China, of course, have built their big models or started building them based on consumer data.
Yes. And consumer data, of course, is quite "noisy ." Yes. So you need really large volumes to build stable models. An interesting observation about Europe is that the situation is somewhat paradoxical: at least here in Germany, we have the feeling that the digital infrastructure is very lagging behind, you know? Although many companies began investing in digitalization within their companies 10 years ago, in fact, at the beginning of the development of AI . So in fact, the digital infrastructure and data are there. Yes. So, there is data, there is high-quality industrial data that we can use. Yes. And this is again a small advantage that we have. Of course, China, in particular, is catching up with us quite quickly on this issue.
Yes. So, they are also starting to evolve, and this is part of their, I would say, holistic approach to improving technology, where data is used early in the production process. They use data, for example, from their robots operating in real-world conditions to improve robot motors, right down to motor wiring diagrams, etc. So this is an area where China is very actively catching up, but I would say that we still have a certain advantage here. Yes. And we must use this advantage. This is very important to continue, in fact, to support the flight of this "rocket." Yes. What we are observing now is quite interesting. Capital, of course, is mostly concentrated in the US, also in China, but it is starting to flow to Europe as well.
So, there is investment in Europe , and big companies are also starting to invest in infrastructure here in Europe. If you take Microsoft, for example, they are investing, it seems, 50 billion US dollars in infrastructure projects and opening a new research center in Barcelona. They will scale up the work of this center. Google has opened an artificial intelligence center here in Berlin. So it's not just about entering the European market. The point is to try to build this " rocket" together, using European resources. Yes. And they wouldn't do it if they didn't see the potential here in Europe. Yes. Therefore, we must also use this potential. I think this is extremely important. If others see potential, we should see it too.
Yes. This is obvious. Yes. That he is here. Of course, we have other investments. As I mentioned, Black Forest Labs, as well as Yun Long, who is returning to Europe with a large investment in his company that creates world models. He could have done it in the US, but decided to return to Europe. I believe this is no coincidence. And just yesterday it was announced that SAP is investing a billion euros in Prior Labs, a company that deals with fundamental models built on relational data, i.e. tabular data. So this is also an interesting development . It's actually, um, the CEO is German, Frank, a very smart guy, and I think it's a fantastic achievement in this regard.
So, capital is also starting to flow in. There are some developments, perhaps not on the scale that is really needed for a start, but the process has already begun. There is this push, yes, and. hmm, I think this will continue . I see this also, as I mentioned, I have had to travel a lot in recent months for precisely these reasons. And then, of course, we need, um, infrastructure. Last year, and it's amazing, you know, it was only a year ago, I was at the AI Action Summit in Paris, I was there and I saw firsthand the talk that Mo gave, you know. I don't know who else was there, some of you, I'm pretty sure, and it was quite an exciting experience.
He was standing, actually, on an oval podium that was open on all sides , and there were about eight hundred thousand artificial intelligence experts around him, and he gave this 20- minute free speech about the importance of AI for Europe, you know, and then , you know, I really got the impression that there was a kind of "aha " effect. Yes. So people started to understand that it was critical to move forward, you know, to find the capital to actually assemble the pieces of this rocket. Yes. And, uh, the European Commission, of course, promised money to build this infrastructure, gigafactories. Yes. Um, and now we see, you know, this process with gigafactories, and it's a little frustrating because it's kind of slowing down .
I am completely convinced that we really need this infrastructure. Yes. that even if, even when we get gigafactories, we're still going to be behind in that kind of infrastructure , but it will allow us to , you know, get some baseline standards that we can continue to work from. Yes. So this is important. What we already have in Europe are artificial intelligence factories. There are actually two of them in Germany: one in Jülich and one in Stuttgart. And, um, that's already an achievement to some extent. You know, now the question of course is: how do we provide, um , low- level and threshold access to these facilities for companies, and also for research and so on.
And, and such developments have also led to European private, I'm pretty sure of it, this has also led now to European private investments in infrastructure. No. And that's important because AI factories have already been targeting startups and small businesses, businesses in general , and that's actually led to investments, for example, from... I need to go back one slide, wait, sorry. A. Okay. This actually led to investments in data centers. For example, German Telecom just opened one in January in Munich. Yes. With a strong focus on industrial AI. Yes. And with the idea of using this industrial data that we have in Europe, and they have strong partners. They have Nvidia as part of it. Siemens is part of this, you know, and I think that's important; these are important steps to build some sovereignty, to create a kind of our own infrastructure, to also develop models based on it, like OpenGPT-X, this is an initiative that aims to create a sovereign and open language model that we can rely on, where we control the whole
process, which is also important for attracting talent, so that talent can work with these techniques. Creating a large model is mostly engineering, requiring a lot of engineering knowledge, and by working on these models, you build and expand your engineering knowledge. So, this is important. But also Schwartz Digits, for example, Schwartz Group invested; They will also build a very large data center in Germany. I think it's very promising, but we have to keep pushing in this direction. And ultimately, it is also important, and it can be an advantage for Europe, to connect different infrastructure ecosystems: from very small ones, like our own at DFKI, we of course have our own small data center, regional ones, like the AI factories or AI service centers that exist in Germany, to very large ones, which we hope will emerge, gigafactories.
So if we can combine that and build ecosystems around it that take advantage of the entire talent ecosystem that we have, you know, the research environments, the startup environments, etc., the infrastructure itself, the companies, and also the startups. I think that's what we should strive for, and where Germany, where Europe has a great chance of actually providing a broad infrastructure base for AI. Um, and then of course the most important part , which I already mentioned, where we have an advantage—is data; in particular, data coming from companies. You know, there's a little problem with how to collect this data, how to put these things together. People are working on this, but the important thing is that this data is inherently more physical than all the other data currently used to train models.
This is data from factories, manufacturing processes, etc. , it consists of sensory information , information about timelines, etc., which can be used to build more complex models, and I think that's still a gap. This is why people start working on models of the world; This is why people are starting to change the architecture of current models, as there is an obvious lack of connection to physical reality, which, among other things, causes hallucinations as well as inaccuracies, and perhaps inhibits the models' creativity in finding completely new solutions to humanity's pressing problems. So, how do we stay the course? How should we proceed? What is the second stage that we need to deploy to further advance our AI rocket space mission.
Because I work at an applied research center, which is in a way located between universities, human resources, companies, problems in society, etc. I believe it is important to join forces. I firmly believe that this area of applied research, which has a long tradition in Germany, as well as in Europe, is what we can rely on in this regard. And finally, for DFKI itself, a few years ago, three years ago, we set out to define directions, you know, and it was a difficult task to think about, " What are the areas of AI that we think will be relevant in three, four, five, or ten years?" You know, in an environment that is changing extremely dynamically.
There are two areas I would like to present to you where I think Europe can make a significant contribution: one of them is the design of reliable AI systems, and this is not just about technology. This can apply to both technologies and ways of designing AI systems, but also to the processes themselves. So how exactly do you design these AI systems, you know, and that actually brings us back to this metric in the AI index. Yes, I think if we don't do this and don't remember this, we will actually lose the trust of people in society. We're not there yet. People in Europe are still quite friendly; Of course, they have concerns, we've heard that, but they tend to see the potential and embrace it.
But we also have to make sure that the technological proposals we put forward take this into account, and that's what we do at DFKI. We have a center called Certain, where we try to not only suggest ways to develop reliable AI and what it should be based on, but also conduct research in this area, because it's something that, you know, there's not a lot of interest on the other side of the ocean . Yes. This is something I will come back to in a few minutes. You know, this is what I think is important. I know many in the industry believe that we need to launch our AI initiatives and so on more quickly. We don't have time for this.
I think it's risky. You should at least think about it, you know, and integrate it into your designs, and that's important because, particularly if you have AI systems that operate in the physical world and potentially do a lot of damage, you need a foundation that you can rely on. Yes. You need systems of systems, AI systems composed of other AI systems. Yes. Which are designed so that you have a certain level of guarantees and a certain level of predictability regarding the actions of these systems. Yes. And so that you also have the opportunity to learn from your mistakes. Yes. It is no coincidence that we make enormous efforts when a plane crashes to retrieve flight recorders, for example, from the depths of the ocean.
Yes. Although often you learn little from it , perhaps only a little. Yes. But it's about trusting the system as a whole, so that there's a way to learn from mistakes, so that you at least try to learn from them, and so that you have systems where you can trust both the technology and the processes. I think this is important, and, of course, these are the examples we are guided by. Another dimension that I am deeply convinced about regarding the development of gigafactories in Europe is infrastructure and so on. This is important, but we also need to move forward with bilateral initiatives to simply get this process started. Yes. That's why we work very closely with France.
Others work, for example, with partners in the Netherlands. I don't know if Holger is anywhere here. Holger, I think this is important. You know, we have cooperation with Scandinavian countries, also, for example, with Italy, and building on these types of cooperation to get things moving and then expanding them to the European level. I think this is important. But it's not just Europe. This, of course, also, I would say, there are other medium AI nations that are in the same position as us. And one of them, of course, is Canada. Canada, to some extent, if you talk to Canadian politicians or AI experts, this concept of "trusted," which seems to us more of a nice addition than a necessity, is extremely important to them.
Yes. This is the main topic they mention when we want to agree on joint initiatives. How to design reliable systems running on sovereign infrastructure. And this is important. And the second, uh, very important topic that we at DFKI started researching a few years ago is what we call TwinAI. This is a kind of very close connection of modern AI models to their knowledge of the world, semantics, and physical environment , and it can take many forms. Of course , autonomous driving is one example, but if you think about humanoid robots in factories, Chinese robots are actually reportedly already doing useful work in their factories. I have rarely seen anything like this with my own eyes.
Yes, I say this to the point that if anyone has seen a really convincing example of a Chinese robot being used in a factory, I would be happy to hear about it . And there are many reasons for this, the main one being that they are not integrated into the complex physical world of production processes. At least for now . This is something we can still influence. Yes. And I believe that this is a chance for us to change the situation in this particular industry. Uh ... Yeah, and the other direction, of course, is that instead of getting bigger and bigger, getting deeper into the jungle, we should focus on what works—smaller things that are complemented by physically generated datasets that help solve specific manufacturing and industrial problems.
There are many opportunities here for Germany and Europe that are worth taking advantage of. Ultimately, AI sovereignty requires the unification of different players and stakeholders; none of them can handle this alone. This includes politics, the scientific community, research, but also industry, startups, and society as a whole, and you can't start with just one piece of the puzzle. We need to work on this comprehensively. And if we do, then perhaps a bright future for AI on the Moon awaits us . Of course, when it comes to DFKI robots, all the robots you see are developments of our space department that have already proven their usefulness in a lunar crater. With that, I would like to thank you for your attention and look forward to the discussion during the breaks.
Thank you. Thank you very much, Antonio. Thank you. It's very nice that you were here. What a great start to the day, ladies and gentlemen.