Mario, a good friend of this conference, of course, uh now uh senior director innovation and research strategy at SAP. Before that, of course, everyone uh most of you know, state secretary in the ministry of uh science and education. >> Yeah, >> thank you. Yeah, I I'm not saying BMF. No, no, no, no, no, no, no. That's really nice. And uh back into into the business world, how was the year um outside of the Berlin bubble? Um to be honest, it was quite nice. Oh yeah. Okay. Hello everybody. >> Hello. >> Um it was quite nice to have really a year of thinking about what happened in the last 8 years. So normally when a politician um has basically a time off afterwards, people question why why is it needed?
And personally to be honest myself we're saying come on they can just get a job. But it really I mean it's eight years and it was a tough end so it took some time to sit down and decide what you want to do what is a good next step and so I really enjoyed it and as well on the family side of course there were more minuses in my bank account so I could could collect few pluses >> I'm sure that this panel will not have a tough end I will surely that it's educational and everything like that so Mario the floor is yours. >> Thank you. Yes. So um again welcome everybody. So first of all the ones who are just checking the agenda will see that we have a bit of a switch in [snorts] our panel.
So we have the beautiful AA Maria Hemp stepping in because Raa Kemer unfortunately had to call in sick this morning. knows she is sick and so she can't attend and therefore we basically um switch the German bundes against the natural force of innovation against Nvidia. So Maria please come up to the panel and thanks your applause and then um second Tina so um we quite had um some different chances to work together. It started in the onet commission of the joint German bundes tag and now you are back with the AI nation. So you're the only one of us running a whole nation. [snorts] So um it's a pleasure to have you. So would you please step up and join our panel?
Okay. First of all, so um as our title says, serenity and the ones of you who've been here um until the morning. So many many discussions on serenity. So the first thing I guess is to give an idea what your type of serenity is or basically how you would describe it and in order not to make it too boring because I think I've at least heard like five terms what serenity is. Would you please connect it with what you're currently doing? So what is an AI nation and how does serenity play into that? >> Yeah, I'd love to. So uh for me if I talk about sovereignity it's mostly about technological sovereignity. So this is the topic that I've dealt with in the last years and also the topic of AI nation basically.
So a nation is an initiative by the German government to support more AI startups all over Germany and uh we have started beginning of this year we have some preprocessors in the different locations for example in Munich and also in Berlin and other cities in Germany and now our goal is to bring all those different initiatives together to form one AI nation program that can support uh founders and startups in the field of artificial intelligence with different um yeah different things that we can offer and I think it's a very important step for technological sovereignity because we urgently need more AI [snorts] companies in Germany and this is basically the seed where those companies can uh come from.
>> Okay. And handing over to not quite a startup anymore. for a while. >> For a while that's true. So but but still what are you up to at the moment and what would be your definition in that context of serenity? So I lead our public sector business across Europe. And as such, my job is to help the public sector adopt AI and adopt sovereign AI, which I define as having options, having choice, being able to sort of stand up from the table and and go if the offer is sort of not good enough. And we're thinking about sovereignity and about AI in Jensen called it the five layer cake. Um so we have the bottom layer which is energy and I think it's interesting because a lot of people still think Nvidia is just chips.
So but we are working across all of these five different layers. So the bottom layer is energy. So of course we're not an energy company but we are having conversations for example with energy companies. How can AI help stabilize the grid kind of forecast when are we going to have renewable energy? Um we are working with um governments as well like identifi like helping I guess it's not our business but we're kind of making sure that there is sovereignity around or options around energy. Then there's of course the next layer of the of the cake is chips which is our core business. Um and then we come to the infrastructure layer. So there is like who's operating those chips?
Who's turning those chips into actual tokens into intelligence? And again there is so much optionality there right now and I think that's the layer where a lot of people are thinking about when we talk about sovereign AI like is where's the data center or who's running the data center and our board approved definition is that sovereign AI is an AI which is um owned operated and governed in country but there's still a lot of different permutations on that and like I said it's always about options it's always about choices. It's always about a riskreward or like a risk assessment like what's my risk and what mitigations do I have if this is an essential enough risk.
But then there's s of two more layers which is the model layer like what kind of models are using alo models and data and then finally the application layer and sovereignity happens at all of those five layers and across all of those five layers and you will never be completely independent like I'm being recorded but still there is no country in the world which is sovereign across all of those five layers. So it's a question of strategic choices about what do you own what do you what do you build yourself where do you partner and where do you also accept dependencies which are ideally interdependencies >> yeah I may go first so I would like to um to strengthen that point as well because so serverity is not autonomy right so it's a that's a different concept and if we say we do have exactly like you said there is not one single country in the world that is totally autonomous and doesn't need anything from any other country and of course it depends a
lot on this balance then how much do you give to the table and how much do you need to get from others and there's a balance in uh technological sovereignity now that you have to keep in mind so if some um I would say like supporter um or country you work with or partner with a lot changes its political structure then there may be a direct risk for your sovereignity. So I think it's a matter of balance not so much a matter of doing everything on your own. So if you talk about balance then I like measuring so and you have a regulatory background so I had one and you work in so how would we like as Germany or maybe as the AI nation measure it?
So I fully buy into it to be honest that you need something at the table to negotiate. But if we follow the stack model, what should smart politics focus on or how can you somehow measure it or is it a feeling which you can never ever measure? Do you just feel sovereign enough? So what's your stand on that? >> No, no, no. I wouldn't trust your feeling. I mean yours, yes, but not everyone's. So um there's definitely like KPIs that you could have a look at. I mean of course it's super hard to come up with one the one KPI saying for example for artificial intelligence how so are we yeah so or is Germany it's super hard of course to measure that but there are some um numbers that you can use and I would always go for the economical uh KPIs right so how is our innovation activity how many companies do we have that build AI how many uh workplaces do we have that have AI how many um uh studies and
um possibilities do you have for education in AI? So there are many many different KPIs that you can use to get a picture of how good a country is um basically situated or look in the field of artificial intelligence and other technologies. >> I would almost say like it's a more theoretical than actually doing the math but I think you can calculate it. >> Okay. Like if you think about what is the risk of a like if you take a particular choice let's say I basically go with a local collocation provider. >> So now what would happen if this option is doesn't work if for some some reason I can't use them anymore. It's like how many days of work would I be losing?
What kind of business would I be losing? And then you can multiply that with the likelihood of your local colo center being shut down or not being supplied with energy anymore or whatever. And that gives you an expectation value mathematically of like what's your risk and I think that is also a way to compare risks along the different the different layers. It kind of is almost like what's your exposure and then you can decide is it something I am willing to accept or is it something I want to hedge against and go or is is it another choice which has a better value >> and how would that concept work when it comes basically to let's say the state or military where I so it's hard to change or swap my state so um but as a nation maybe not as the nation but as Germany if I want to be sovereign Is it possible to have that without a military dimension or is this interlin?
So I'm asking that question because at many panels before there was the the situation and the US and all these topics but personally feeling that the real push that we take this so serious and Fabian said it in the keynote is really because times have changed. So how much of the whole discussion is a a real technological rational one and maybe rather politically driven even military one so is is that a part of serenity as well or is it different topic so maybe it's diff it was a question for one year to be fair with a German bunak so it's not dedicated for Nvidia but I guess you'll have a stand on it as well um because I think that's an interesting topic because the push and the money is definitely coming from a new reality I would turn it around and I would say that digital sovereignity is essential for military sovereignity.
>> Yeah, I would support that. Yeah. So sovereignity, the technological sovereignity comes with a safety aspect. If you're not so at least in some way um a sovereign country then you will get problems regarding your military aspects as well for example quantum computing. Yeah. So if you are not able to be in the game of quantum computing the global game. Yeah. To have some aspects of quantum computing in this country as well. um some companies that maybe build parts of it or things that you are part of this whole value chain. Um if we don't manage to do that then we will get into big problems also safety problems military problems um as soon as quantum computing is ready and other countries have it.
>> I think it's also much more practical like our adversaries are using AI and are using those kind of technologies. So if we don't have them, we are inherently vulnerable. And I think there's like a great case study to the east where like we're fighting new types of wars which are like not so much like tanks are still there, but like if you look at what's happening there with drones and with like landbased drones and other system, it's a different war. So for us to be safe, we need to have this technological stack ourselves because that's what will keep us safe. >> Because you brought up quantum, let's maybe do a bit of then stay a bit with quantum because I guess that was a very good example.
So you said we need more AI companies with an AI nation pitch. Would that include quantum? And basically what's your stand? Are we at the tipping point where it's more about the scaling? Because we heard many panelists that we are too late in many things and and that we should stick to the stuff where we have chances. Personally believing that at the quantum scope or at this quantum stage we are at the point where it needs to go into industry scale. So it's it's nobody wants to build a new computer from scratch. There are some players there are some fancy demos but but personally feeling it's at the stage where it really makes a difference now.
So is the AI nation focused on quantum startups as well or is it more like still a side topic? So I would be really interested to hear your >> Yeah. So just for now we don't have a specific focus on quantum quantum computing. Uh although we do have some startups that we have supported that work in the interface between quantum computing and artificial intelligence. I mean the combination of both is super interesting. Um I'm not an expert in quantum computing. uh unfortunately so I'm not the right person to answer when is what coming but um I I would say the the potential that is between quantum computing and artificial intelligence is enormous and I think um we will see more and more startups and companies going in this direction of combining those together.
>> How about quantum? Are you scared new chips coming up? I mean I did a PhD in quantum optics for like a year before I decided to change tag. Um I think what we do work quite a bit with a quantum ecosystem and so we started building also systems which can help with quantum error correction which is one of the big issues we're having that right now these systems are just still super super noisy. So you need like hundreds thousands of physical cubits to kind of get to one logical cubit. Um I think what's really interesting is also some of the ancillary developments like right now we're still developing this technology and I think there is some technology coming out of that which is sort of useful in its own right.
So I think that's one of the things that makes pursuing quantum uh interesting even if it might still take a few years. Pick your own number. I'm not going to say a number um until >> how many years you guess >> until we actually see it um for widespread adoption but I think as with a lot of things we will learn a lot of very useful things and so I think this is also some of the ways how we are supporting the quantum system right now we're making progress as well with quantum with quantum algorithms which are still being um tested and simulated on supercomputers I mean I think we look at this it's just another chip.
So when I joined Nvidia which was two years ago we had one chip. It was really easy. I was like there was just one chip great. Now we're having seven chips. So our latest systems consist of seven different chips which are all interrelated and every chip has its job. I mean we didn't replace CPUs. We're actually now making CPUs ourselves as part of the overall system. And to me the QPU will just be a part of the overall system doing a particular job. In the same way a CPU has a job, a GPU has a job, a DPU has a job, bluefield card has a job. So they all have their own jobs. And that's how I think about quantum computing.
>> You didn't say a year. So how much? >> No, it's I wouldn't answer as well. So you want to make a guess? No. Um then then maybe stopping with quantum or is there some quantum quantum startup here who wants to pitch real quick. Um then we get back but maybe with the quantum thing um I can use it um as a bridge because the next question is again going back to our German landscape. So um surely we have the middleand we have all our thememes um and and that always is and was basically our powerhouse keeping the economy up but now we came a bit from the scaling problem with quantum but it's still true as well with AI.
So having that much talent in the field and that many bright minds and many companies, is it somehow in the scaling part of AI a disadvantage or is it still an advantage? We're just doing something wrong. So why is it so hard at the last tipping point to scale here and there's the risk that elsewhere is faster and maybe not always US maybe even be it France with a more centralistic approach or a more let's say politically driven agenda by macro or something what is why do we have that problem at the last piece of really skyrocketing what's basic and then >> yeah I basically agree with you that we do have a scaling issue Yeah.
So we do have um a lot of very interesting startups um generated created and then we don't have the big players. Yeah. So somehow we kind of seem to have not the perfect pipeline to have a startup growing into like anme or bigger company. Of course that takes a while. Yes. So um it's also a matter of time and I think we do have several reasons for that. one is definitely um still I mean we've heard that very often but still it's a lot of uh question of money so how much funding and how much um venture capital is available for even bigger rounds for startups to grow um but there's also an issue about buying software and solutions by from startups and that's maybe what you have mentioned culture and the theme yeah position that we have um but I'm not 100% sure because also I seen many big companies and corporates not buying solutions from startups.
Yeah. So it's maybe not just an issue ofmemes. So I would say we do have several things that we should tackle and we need to solve for those um technological companies to grow and that is definitely capital issue. It's um customer issues. So buy the solutions from those startups and then it's an it's also mergers and acquisitions topic because I see I don't see enough um companies German established companies buying startups yeah here so that they can um incorporate their tech in their own business model. I think that would be super important as well. >> I mean I think scaling is hard for everyone. Before doing public sector, I was responsible for our healthcare and life science portfolio >> and some of the large companies in that space and I'm not going to call names again.
Oh my god, like it's really hard to sort of scale within those companies and there we often had very clear use cases. It was very clear like what and every like yeah we need to do it but then just the the politics and the coordination and do we really need this? Oh do we now like do we really need that many GPUs? Oh no like we need much more. Maybe we need to do another round and need to look at this again and then another four months pass. So I'm not sure. So I think scaling is hard for everyone. It's hard for the big companies and hard for the small companies for different reasons. And I almost think that the having a diverse base and a lot of companies also being familyowned or private, that could even be an advantage because >> like some of these bringing in those technologies, it's it's an investment.
You have to kind of do things differently. You have to change. If you do it right, you basically don't just do what you've done before, but now with AI, but you use this as a chance to really rethink your processes, rethink your workflows. And those kind of decisions can be made often easier in a smaller organization with somebody who maybe also thinks in generations and doesn't think in quarters. So, I think it's like there's pros and cons for all of it. So I I think it's not like so easy to say, oh it's hard to scale in Germany because we've got a lot of small companies >> and but it might need some special kind of looking into like do we need what can we do to help overcome things like it doesn't need to be maybe money.
Maybe it's education. Maybe it is um exposure to already working um solutions because of course fear-mongering that Open Claw deleted all your emails is drawing a lot more clicks than okay if you basically run your open claw on a separate machine for a very specific task with regular backups and whatever or basically you box it in enough then this could be really useful. Maybe one more thought about it because what you just said also um that came to my mind um shareholder value if you are oriented just on shareholder value um that could also be hindering innovation right so the the typical innovators dilemma comes from very successful companies big companies that are only only oriented towards their current shareholder value and not about a longterm entrepreneurial real vision and then this risk comes up that you ignore a technology and someone else maybe startup maybe a comp competitor uh embraces it and then in five years has a different product and a better market share.
>> Yep. Fair. I would buy in. So at least if I'm allowed to have an opinion I would totally buy in into that one. Um at at one point if only you said um regulation. So I mean there were so many regulation panels and and I don't want to fully open up Pandora's box of regulation but since we have an interesting combo like with Nvidia and the AI nation being all the startups like to whom is regulation more harmful like a big corp big tech or the many small ones because everybody is complaining and I was a regulator so sorry for all we did but like can you is it is it different is it the same or is it in the end a necessarily who everybody wants to complain or even hide behind so maybe the different standpoints from Nvidia to the voice of the AI startups >> who wants to start >> with my previous head on in healthcare and life sciences I think sometimes regulation is a necessity and it's potentially also
a necess it's more helpful to small companies than big companies Because what regulation gives you is a a safe like a certain level of safety. Like you're basically knowing, okay, this is what's allowed, this is what's not allowed. If you're in healthcare, you're dealing with uh patient data. Basically having clear rules, what you can do and what you can't do is reducing your risk. And that's particularly important for small companies because they can't afford it. They can't afford to basically be c caught doing something that they then end up not weren't supposed to do or they can't like for them this dealing with this uncertainty is much harder than with a bigger company which has more buffer and where uh a lawsuit is not going to put you out of business.
So I think sometimes I'm not arguing for like overregulation but I think it's so much about AI is about trust and regulation is one mechanism for trust and I think we also said it before we came on the panel that I'm actually often quite glad I live in Europe where my personal data is protected where my keystrokes can't just be monitored and used to train some sort of model where there has to be transparency and see how my data is used and where I can opt out of my data being collected to train models or where I can opt in if I want to. And so I think a certain level is actually important like both to kind of have this safety to basically know okay I can do this I'm not going to risk my company but also to build the trust that this is actually something that also consumers patients citizens are willing to engage with.
So let's see. I'm very excited. There was the sentence that regulation is better for the smaller ones and I was checking the audience and it was absolutely mixed feeling in the faces down there. So okay, that's fair. So everybody has a talking point for hunts. You should pick that one up. No. So >> I mean definitely regulation has a fair point, right? So we we don't want to be endangered by certain systems we want we we have to to deal with. But on the other hand uh I would say the safest position that we can the safest situation would be to don't do anything right and this is basically where we are currently heading from my perspective.
Yeah. So u if maybe GDPR already wasn't very very big issue for startups and young companies. So I basically built up my own company in the time the GDPR came up and it almost killed us because we were 10 people and five of them were dealing with GDPR. So, and that has to be something that you can also finance, right? Um, basically it took us a long long time and I'm pretty sure that we would have been faster on the market without GDPR coming in in this aspect and that time on the other hand what we just see with the um AI regulation the AI Act also I would say if the AI Act comes in the way that it's currently planned although we have to say it's currently discussed again right so but the the current plan um I'm pretty sure with all the things that are in that regulation, we will not see any new AI offering, not using offering, total new offering, a company building a new AI emerging from Europe.
We will not see it. >> Perfect. That's what the panel is for. So, we get we're getting close. So but and staying there a little because u now we have a controversial discussion. So um trying to be neutral here but the panel is about serenity. So um we tried to define serenity with layers and KPIs and technical wise obviously um as an ex regulator um there is a standpoint as well from a lawmakaker's perspective. So is there a danger that the whole serenity thing turns something into something hindering or from a company perspective the whole serenity thing is basically a boost for sales if you've got something to sell which seems to be so somewhere at the layer.
So what's if we look basically into our glass ball what's will it be a push in terms of products and innovation or will it in the end be another layer of scare uh which slows us down. So what's your guess on that one? Well, if we have a look at other regulation that we've already had in the last years, I wouldn't say learning from those examples, it didn't came up with the new market opportunity. Yeah. So we we all I mean there are many arguments about there may be options specific sensible areas where we need products that are um especially uh according to some regulations but in the end look at what we have nobody is using European products yeah so >> that's tough >> I mean we everyone is using the the the digit tools of American companies.
I mean, this is the situation. Yeah, >> I do have Mistral in stock. >> That's great. And I really >> Oh, not anymore. I know. But on sunny days, I try to >> I really appreciate that. >> Yeah. So, I I would say this big um marketwise um potential that we maybe have also expected from GDPR to arise has not manifested itself yet. So maybe it's the same with the AI Act, but of course it's something that we will only know as soon as we have it. >> I'm not an economist as we established already. But in and if [snorts] there's an economist and I say something wrong, please speak up. But in economic theory, the most efficient system is where everybody is specialized on the one thing they do best.
So if you're decent at baking bread and you're excellent at making pasta, we're all better off if you're making pasta and buy your bread from your neighbor. So that's economic theory. So an in an ideal system, we would all basically focus each other on just the one thing we do best and collaborate on everything else. But that's just not the way the world works. And sovereign AI is about okay what as I said before like what do I want to own myself build myself where do I partner and where do I go for dependencies interdependency so that system is by definition not the most efficient system so it will create sort of additional demand in some places which just by the inefficiency because you're just not creating the most efficient system whether it's a good or bad thing We can argue about it, but I think what I also said before, for me, so serverity, it's a it's a trade-off.
So I'm willing sort of to make certain trade-offs to be to have more options and that will in certain places create opportunities that will in certain places um like not create opportunity or kind of close opportunity but there is something there that I basically yeah trying to kind of mitigate those risks which I'm seeing and kind of trying to keep options for myself at a price and I think that's what you were saying as well like it's an interesting question also between stated preferences and acted preferences like what people are saying they want sovereignity and are they actually willing to pay the price in whatever way shape or form that comes that could be monetary that could be performance that could be lack of features and we will see about that like what people are really willing willing to do but I said in the end it's about options it's about >> yeah It would be super sad if the only position that we have is to say, "Hey, we do have do have this
European product solution here. You can use it, but well, you have to deal with it. It's the performance is performance is low." Yeah. >> Yeah. I mean, so that's basically also I don't think so should be the only selling point. >> Yeah. Yeah. It's it's not possible. You can't force people to use a bad product only because of patriotism or something. I think that's a different discussion which we we should like and which I think I also heard at different places. It's like okay, what's our USP? What's our unique value? Where do we have a unique advantage? And to lean into that um and yeah I mean whether sovereignity is feeds into that advantage or is hindering that advantage is another discussion >> but then let's let's so um when you mentioned bread and pizza my brain went like so let's stick with that.
So I guess the German bread were cars or manufacturing and we were very good at it. So then what is the new German bread we should bake as products to become that sovereign. So if we remain in that model because obviously then it was not whatever I don't call any names but it was certainly not LLMs in and in that game it's obviously not not chips. So what could be our future bread the bread and butter topic to build? So what what should we focus on? I mean nobody knows but what's what's the guess? So and I mean you probably see it from the trends the startups who arrive and what they want to build.
So what do you think is worth focusing on to basically become superior again in that domain because I guess we all agree that fully owning everything that's just maybe in a political speech it will not work. So it's about being called when there is a decision for whatever reason. But what are the things the bread or we stick with the bread thing for the last 10 minutes. So what is the bread we can bring to the table? And we have good bread actually the Germans that that we still own. I mean maybe the Austrians no offense >> we can bring the bread itself. No, as for my perspective, there are a lot of options that we have because we have uh very strong industry and we do have industries and we do have a lot of research in different technological aspects that we could combine together.
So I see a lot of potential in robotics. I see a lot of potential in pharmacy and medicine in uh different use cases and I also I personally would love to have like why not building on our fundament industry and come up with a smart car and autoban. Yeah. communicating with each other. Um controlling cars to be exactly the u the velocity that is able now hindering and providing um better um sustainability less um murders. Yeah. And a lot of different other aspects that we could build on. So this technology combining the autoban and the cars with artificial intelligence and sensors and we just love that. I mean, that would be so great, right? It's >> that's that's how you energize Germans after eight hours like on the bar.
>> There's still a game in cars. >> It's that easy. >> And then we could export that. Yeah. Not just the cars, but the streets. The street. >> It's tough for you. You can start to sing millennial songs or something. So, how can you want to hear me sing? Believe me. Um I think what we learned at the beginning of the day is how important context is. Y >> and to me that is the USP which Germany and Europe has. We have a lot of context and >> AI becomes really powerful with context and a lot of that context is in >> manufacturing data. It's in processes. It's in how you do certain things. So I think leveraging that to have very like bringing the AI into the real world and then having it do really useful things grounded in that rich context.
I think that is the actual USP and we have other data sets in Europe which are very contextrich and which do not we do not find elsewhere like another great example is healthcare data. I mean across Europe we've got some of the best healthcare population data sets because we have centralized government in some way shape or form run health care systems which have often decades of data longitudinal from cradle to the grave which for example in the US you don't have you only have um like a hospital has billing data and then when people turn 65 they go into Medicare so they have you have all those breaks in the history of a person and so I I mean that's another another aspect like and there's I think a few more I mean I think the European data space has got a bit of stick at some point but I think from the idea it's great to have like a mechanism as well that allows for access to this data and there's also some work
done around it used to be called federated learning I think somebody called it something else today um sorry I forgot what it was but also basically being able to use those data sets and keeping protected though, but still being able to train AI on it. And I think that is the opportunity is to take our existing industries into the new world because in the end we will still need modes of transportation. We still need houses. We still need food. Um I mean all of this doesn't go away just because we have like our phones in our hands and our computers in front of us. I mean there's still a lot of needs we have in the physical world.
And I think that's where the opportunity lies. >> Um so would absolutely agree that we underestimate um the value of our context and all the data we have. So that's fair. I just because we only have five minutes maybe a little loop back when you were saying we we can't just buy products because they are patriotic but they suck or I mean it was it was not a quote it was a summary something like that. Um which which leads me to the last question. Um so talking about the context and the data that was realistic but today we often and it's a serenity panel we heard the thing about the values and so that the values are basically um with what differentiates us and what's so powerful.
So I I've been to US and I never had somebody coming in the company in the morning say let's build something super unethical. So and then they start working. So is the whole discussion about this value thing just because we don't have decent products or did it just not yet fly to bring them into superior products to win that thing because I personally think what you said was right it was honest um but nobody would use or maybe we would use so like me using mistral to some extent but as it comes for business solution nobody would buy I don't know the sec or healthcare the second best system just because it's super fair. I mean, in the end, my accounting is wrong and I go to jail, but it never felt so fair.
So, what what's the thing with the value? Is it an excuse or can we really leverage it and we just didn't do it until now? It's something philosophical for the end, you know, something tough. >> Handing over to Hans, he comes up with AGI. So, it's getting rougher, guys. There are people who do make decisions. There [snorts] are yesterday I talked to somebody and were like, "Yeah, I basically sold shares in that company, not because the company wasn't performing well or because I don't believe in the um prospects of the company, but because there are some things I just don't feel all aligned with my values." And that's completely fair. I mean, that was the whole discuss discussion we had on on ESG.
I mean, values is a a tricky one because values are very different. I mean different communities have different values and you could almost loop it back to regulation because in a way regulation is a manifestation of values like what do we think is worth protecting? Do we think and sorry going back to GDPR but do we think the privacy of where do we stand on the privacy of our citizens versus economic growth and if you look around the world you will find different countries which answer this question in different ways also different cultures which answer the question of the value of the individual versus the value of community and society at large in different ways.
So I think it is something that I technology we can show what's possible but I think the role of society of government of regulation is to have the decision of what do we want and that's where it comes together. >> Yeah. I'm I mean of course sometimes there are situations in which values may really be some buying argument. Um it depends a little bit on the situation but I doubt that they can be the one main buying argument for all time. So but what I can see is the different aspect of values is more like if we come to LLMs for example trained on different data they contain different values of course and um that's an issue where I think there where sovereignity and values interact very closely with each other because if we don't have like solutions that we want to build here um but we have to use solutions from other parts of the world they may contain different values um from to ours and that's actually an aspect of sovereignity that becomes
quite important also from yeah companywise um viewpoint because sometimes it may generate errors right so that you don't want to have so in this case it becomes a real hard um market figure >> it's basically also interesting when we said before does a product suck or not but like also It can be a question like if it I might be willing to take a product which is a bit less good but I for example have transparency into what went into training this model. I mean I think open source is a really interesting concept when it comes to models like what is open source when it comes to a model and very often people then say oh yeah it's open weights okay fair I basically I but what's a weight it's like a it's a matrix with like some value in the matrix what's what's the weight what does it mean and so actually another step is then to say okay no I don't just want models that are open weight I want and open weight
models have the advantage I can run them locally So I basically don't have to send my data somewhere. But I basically can have it sovereign under control in my server. But if you take it one step further, you could also say no open actually means I want to know what data went into that model. And then I might have a model which is a bit less good than the best I could find in the market. But it gives me a better feeling. It gives me like the confidence that it's for example aligned with my values because I can inspect the data set that went into it and I basically know okay there is nothing in there or there's nothing missing there which is important to my value set.
>> So since we are running out of time so maybe as the the quote of the evening regulation is just the manifestation of values. I mean that's I should have had that as a state secretary when everybody was complaining it's like think about that. So I like that one since Oh yeah, he's already standing up. Maybe one last question. So uh today it was serenity and when we all started with AI it was a car then it was a chatbot taking over the world. So if we are back in 10 years sitting here what do you think would be the title of the panel? So maybe just both of you one shot and then we leave the stage.
So in 10 years what will be the topic? with your guess and you can all prove us on and I actually just that's our invitation for 10 years back. >> I have one in my mind although I don't know why but it was like what came to my mind was 10 years later and we still have radiologists driving to work in their cars like I mean I I don't think it's what we're going to see. I think I mean I worked on self-driving cars 10 years ago I think and back then we were close but I think this time we were really close. So, I hope that in 10 years maybe the surgeon or the radiologist will drive to work in his self-driving car.
But actually, I think we will still have radiologists in 10 years. But it will be interesting to sort of see how the world of work is changing. >> Radiologist in the car. We have to speed up Hans. I'm setting the stage for the future stuff. So, what's the next one? >> Maybe it could be something like uh the German autoband technological future. >> >> So then thank you so much for the panel. Thanks for your attention and um I guess we leave this stage. Thank you so much.