foreign but I would like each of you to answer um so um I think I'm one of those crazy people that uh enjoy taking responsibility and fighting for things worth fighting for so think this is and I I like I'm a True Believer in in European values um I left Apple in the U.S to take my part in like to do my contribution to increase the chances that we'll have a a seat as a table that we can shape our own destiny because I think not just for Europe but for the world in general the next era will be better off if there is more balance of power and if it's not just the US and China calling the shots you mentioned something very important the European values set I think we'll hear more of that yeah what about you um yeah if I wake up with my technology hard in the morning I follow the saying of one of my favorite video podcasts of what the time to be alive because we're living in
a time where you hear there are news and exciting news from the technology perspective every day yeah and it's it's even not nearly not possible to digest yesterday's news before today's news are are coming up uh so there are a lot of areas to to look after to work on and sometimes I feel a little bit dizzy so where to where to start first um and as the owner stated already then the major question is how how can we cope up with it yeah how can we avoid getting in another field of digital and sovereignty and uh cope also with the challenges on on a global level and and and stay sovereign we're going to be able to create our own models our own applications and respect the European the European value and and values and Quality Systems yeah on this point I think we are all rather aligned and I want to draw back on one word that yunus was saying balance so what we when I'm waking up in the morning um now
for one year I'm heavily involved in the AIX negotiations and there my boss and I um we are really fighting for this balanced approach because there's so much anxiety even yeah kind of Technology hostile approach uh within the European Parliament but in general and in politics and yeah just a lot of emotions ideology and so on and what we are missing right now also in the AI Act and then the whole AI discussion is to um really have a clear look on the ground how AI can also improve our life what are the chances and so on so back to your questions this is really keeping me up at night but also bringing me on the morning back on my desk or in the parliament to fight for pragmatic and forward-looking and more optimistic approach and to a technology and also maybe acknowledge that you can regulate too much and maybe you should also allow technology to prosper without trying to regulate every piece I'm saying something very important to strike a balance between what
is safeguarding and on the other hand Innovation is such a motor and to cut that off would not be so good however the AI Act as it is proposed right now doesn't seem to include the idea of Founders startups all that Innovation it it doesn't seem to have reached the AIX instead of that actually there's a ban on high-risk applications how does that impede the competitiveness also in our ecosystem um it's a it's a good question and um I need to start here really at the beginning um for us and luckily this is really a cross party um impression in the European Parliament the starting point of the commission is just completely wrong so um the idea of um building a horizontal framework that is applicable to every different sector to every different use case it's just not possible because AI is already a buzzword it's including a lot of things it's including expert systems that are existing since the second world war basically but also newly Foundation models and deep learning systems and so
on it's including applications in the transport sector in the health sector and so on and to build horizontal rules that are 100 or literally understood in every of those different use case is just impossible and this I think is the the biggest flaw in the commission's proposal and that again luckily we in the parliament we are now trying to um fix um hopefully tomorrow I was already asked tomorrow they will be finally um the the text and that we will vote on but there for example in the new article 8 we are really addressing now this point that we are saying those high-risk obligations that are following should not be understood literally but should take into account um sectorial legislation and harmonize standards and so on meaning that for example if article 14 on human oversight is speaking about a stop button the stop button makes sense for some high-risk AI applications but not for all and this again is one of the biggest improvements I think now by the European Parliament and the second
point is also that we thought a lot about the value chain and how to help European companies to in the end become fully compliant with the AI Act because unfortunately the commission um focused only on let's call it Downstream actors so let's say a Bavarian startup that is then completely left alone with becoming compliant and the big shots for example Google that is providing the data set of May is that is providing the foundation model would be completely out of the AI Act and again the Bavarian startup is being left alone this may be just two of the um or the main flaws of the commission proposal which make it extremely difficult to make the AI Act work and practice in the end again we did know I think some improvements but they are still a lot open and acts in my boss and I we are still not very sure if the AI Act is really a step forward and in the whole AI strategy of the commission it included an ex ecosystem of
trust which is the AI Act and the ecosystem of Excellence which is good but the second point is completely forgotten Cai it is 82 articles focused on risk and two or three articles focused on Innovation and we think hopefully together with Industries society and so on in the next months we can make sure that the second pillar is strengthened and also a little bit more underlined and brought back a little bit in the discussion and not focusing any more just on risk even though it's important but it's not the whole story so we need a more holistic approach I'm going to take that word risk and ask you you know what is the exact risk for you right now if you look at the proposal the way it stands now without fixing it yet yeah um so the risk is that we're doing what we've done in the past we're pretty good at um is uh to hesitate and to be risk overs and to uh to some panels and shake hands and talk about
the future and then a few years down the road we'll kind of look back and say oh yeah man why didn't we miss this train right wouldn't that be great if we would uh be able to act and not just to complain not just to force the U.S tech providers to put buttons on it on like warning this is an AI that has been trained on this kind of data or do you accept all those cookies um I don't think that should be our contribution so I think this is and this is an excellent example like the fact that we're sitting here because here in this room they're our AI leaders the smartest people we have on AI in Germany what do they worry about compliance and they should be worrying about Innovation they should think about how do we position ourselves and our companies in the next Industrial Revolution how do we conquer the world with our great ideas what are they thinking about how can we be compliant what kind of forms do
we have to fill what are the liabilities exactly right yeah and and it's not like we have plentiful AI leaders where we can kind of spare a few that can worry about regulation I mean we are catching up to the US anyway so I think that's the risk not only for us we can only survive if we are part of a strong ecosystem we can the strategy that Microsoft is doing like monopolize everything that is of course not a strategy for of alpha our strategy has to be collaborative our strategy only works if we can Empower government and Enterprises to um build a strong position themselves then we can also survive if that fails we by ourselves impossible UC is a strong ecosystem what does that mean for you creating creating value um like we've we've got some of the best companies in in Europe as partners and I think they have a tremendous opportunity they have unique data they have deep knowledge and they can use all that and build new products or build
new kind of build new markets build new categories and they're starting to do that but the question is is it going to be fast enough because we are now seeing that and this is like a global consensus this is the market of the future there's tremendous potential everyone wants to win and if the US wants to win they're they're getting out their checkbook okay we want to win that 30 billion right and this is like 20 billion is just open AI the internal cost every every Enterprise we're talking with tell us that the Microsoft sales team was already there selling opening as they should right and that is completely legitimate Behavior we should be doing the same Jack how do you see this entire development because you were the the founding member of um the German AI Association you've been passionate about AI for a long time how do you see this View um first of all I fully agree to what has been said so far and uh coming back to to your first
question when when I heard the first time about there will be an AI regulation I asked myself so what for you so there are many regulations in in regards to uh safety relevant products and and all other uh kinds of of applications and products that potentially could harm people and and uh so why don't we focus on this and extend uh uh the the application of AI into these products the first version of the uh the first proposal of the uh the commission uh was at least well structured yeah so we were focusing on on uh on applications an application View and classification into risk levels uh but when I first read about uh from the uh Council proposal the general purpose AI will be considered as high risk in general I hardly could believe this because this is in contrary to the first proposal to the overall idea and that's simply not practical uh um and I asked myself why don't we do this like in a uh like in in any other areas
that if you build a high risk system then you have to make sure that the underlying Foundation model that you're using has some kind of certification or follows the rules of the overall AI Act this would be so easy and as as practice since since tens of years in in the market for other areas um because you you never if you if you build a foundation model you never know who's going to use it for what purpose yeah especially when it comes uh also to to open source initiatives or to startups yeah you cannot think about all potential usages and and applications and um uh and and it will be a huge burden yeah to to follow these rules and to to apply all the obligatory processes that that you have to do and this will from my point of view be a burden for the whole uh AI ecosystem in in Germany and and Europe which could be easily um made set up in a different way yeah and we have to be more
flexible and really more specific on on these types of of questions and the development is so fast that you have in the future there will be threats and and potential harms that we don't foresee right now even in the regulation yeah so if the if if these potential harms occur we cannot wait another four years to have another regulation yeah so we have to have a framework to react quickly and it could be done for example what I think is more more or less logical is that there should be some uh some rule to Mark artificially regenerated artifacts like like text chatbots or or images as artificially generated and this is somehow already done in in for example in Norway where there is a law that uh even manually uh photoshopped images from people in in social media have to be declared yeah to to avoid uh misinterpretations in kinds of beauty ideals for for younger people so this is this is easy this could easily be done and I I think this is the
way we should uh handle this now but all three of you agree also that one of the most important things for Europe will be European value set and a strong ecosystem now yeah I know you have two projects Liam is one of them how can that help with strengthening our ecosystem um yeah so the one project is really uh bringing in some ideas or comments uh in in regards to the to the the AI Act uh and the other one is is more fun yeah which is uh uh thinking about what what what can we do know in order to cope up with uh the US and and and China because uh we have the talent yeah and we have the data yeah and the the very best example is what what Jonas is doing with with aleph Alpha uh but we are not bold enough yeah we don't have the money uh as as for example Microsoft and the other big internet Giants have um and and therefore we have to uh to cooperate
yeah in all areas the um competition is not within Germany or not within Europe the competition is overseas uh and that is that is exactly exactly uh the goal of of the project that we set up with large European AI models to to build a platform that provides on one hand the expensive compute power yeah to a research institutions to startups but also to the industry and to build a catalyzation uh crystallization point in order to be a center of this ecosystems and that we each help with each other support each other so that we are not that we can face the the wave of of applications and and AI that is coming from overseas so one thing is the infrastructure as well what about um Jonas how does your company need to deal with the future and this competition how can you position yourself in Europe to make this way forward so we've started building language models even before gbd3 was out in 19 one of the first things we built was a very
similar to gpd2 type model that we're speaking like Chancellor Merkel uh that is quite fun um and we invented multimodality two years before an lgbt4 has it we now are the first team that has explainability so we can provide safe and audit proof results and that's giving us a lot of traction in critical Industries like Health legal Finance um government security so I think we're well positioned but the what I worry about is that we won't be able to move fast enough because we don't have the advantage of calling Daddy Microsoft to kind of pay for like a massive acceleration we basically have to bootstrap of course we're Venture fund we have some incredible investors that kind of realize the potential of this technology also pretty early but the the real bottleneck is the adoption like the work of partner with Partners I think kind of this is what will enable for the founding rounds this will enable everything right if we can apply the technology fast enough and this can only be done through
Partnerships and of course public administration is a big puzzle piece here for me as well but overall just the kind of the speed so what I worry about most is are we going to be fast enough and what I think is the kind of the bottom connect there is not necessarily the availability of compute that helps as well but also the courage and the the partners and the ability to move fast which is not necessarily a German discipline um so I think that anything we can we can do to accelerate and instill some courage into our AI leaders would be helpful I got that message um I actually think that our environment is changing um I'll tell of a project that we just started also in in Berlin called the testing experimentation facility Health which with the tooth for the first time um shortened the certification so that AI applications and Innovative projects can be brought onto the market faster so I think we actually are making the moves as you said we just need
to get more fast or more courageous about it what would give what do you think would give players the more more courage what do they need do they need the backing from from the EU or is it finding other partners who think like like are like-minded in in Europe or how do you think when regulations certainly uh is not helping but because it is introducing uncertainty and it is um focusing our Creative Energy on things that are not necessarily value driving so I think that's yeah certainly it's not helping I like your example on um what are we actually doing here we are regulating fun foundational technology because it might be used for something that we don't like which is to me something like it's almost like at the invention of the computer we would say oh look at that computer that's pretty complex right nobody really understands what it's doing and you could write a nasty email with that or you could like look at a evil website so let's regulate all Computing and
now if you want to build like a watch with a computer with a calculator on it that all falls under the compute computation General computation regulation which maybe is like a little bit hyperbole but I don't think it's too much of yeah um came back to you you said um you work very much with positioning the idea of trust also in the regulation how do you go about that because I think this is where Europe can really make its Mark I think our brand here should be about trusted AI how do you see that how can we fix um something that the regulation that is coming now so that it includes that so that we are actually on our path to creating that brand yeah so this is actually in the European Union one of the biggest Point uh behind the AI Act was the biggest motive that they really want to create a second Brussels effect so after the gdpr which they'll see a little bit as a Golden Boy of European legislation um
and maybe and now I'm getting already a little bit negative not seeing things that didn't um go so well with the gdpr and and yeah I think you we should really learn from the mistakes especially when it's about enforcement about legal uncertainty with the gdpr and so on but also and now going back to your question focus on things that work well so for example in general even though I'm not supporting all the principles built in Article 5 of the gdpr but maybe it's a good idea to have some general principles for AI systems like yeah you were talking about trustworthiness or transparency for example in general I think human oversight makes sense or unbiased data sets and so on so I think if you manage to to have some guiding principles um it could really work if we manage to make them enforceable and also let's say logical on the ground going a little bit back to my first point and I think a lot will depend on good harmonized standards on good guidance
and also fast guidance not like with the gdpr where we waited for years or even the certification just at the beginning of this year exactly so this we definitely need to avoid and this is I would say one of the biggest potential bottlenecks with the AI even if we as Parliament and the council are fixing most of the problems it's really all depending on good enforcement but if we if we manage all of that I truly believe that um that it could create a Competitive Edge um like you are saying we could then maybe sell it as trustworthy AI made in Europe and people know that then certain safeguards are implemented are in place and so on um and yeah then maybe we will be able to catch up with U.S players and so on and but again it's it really depends on a lot of details and especially on the enforcement we need to have legal certainty and I would completely agree with my co-panel is that right now there is not a lot
of uncertainty and I would also be afraid that we actually cause the opposite a lot of legal uncertainty and this combined with what Jonas was saying with the lack of investments just one example there's an AI Lighthouse which has a budget of 800 000 Euros for and this is the Z key EU project even the Chinese city is investing I don't know 10 times as much and so on and if we if we are trying to compete worldwide with such low Investments we will never make it and Jonas was also making those examples of May I they are getting so much funding and I guess a company like alif Alpha probably if they are going to a city will often be confronted with a lot of skepticism and fear and so on so despite all of that that we are talking about we really need a change of our mentality and be more bold like York was saying and really try out and again make really use of this technology yeah to open up the
floor for questions from our audience if there are some there's one at the back thank you so thanks for the interesting discussions my name is Christian Goodman um so I'm just wanting to say you know regulation if it's imprecise and ambiguous it's probably a bigger problem I think it's inevitable to have regulation but maybe my question I'm not quite sure if trustworthy AI like AI made in Europe is an actual selling point to other countries uh you know why would they buy a European or German AI made ethical AI let's say why would that be interesting for them maybe you could expand now I'm happy to because I'm selling to the US currently um because they worry like I mean we don't have to that's like always a little bit odd to me that lawmakers are trying to force us to do things that and what that at the same time make sense right so um like I don't know like this is going to be too expensive so we Outlaw it right if it's
going to be too expensive nobody's gonna do it right and um we are working with some of the world's biggest banks and of course they care about the trustworthiness of course they care about um if they want to use AI that it's gonna be trustworthy and it's gonna be correct and it's gonna be like the human machine collaboration is set up in a way that it empowers their experts and not opens them up to risk and what they want to have it auditable and I think this is I like without any regulation I would have built the same piece of software and it's not European they're not working with us because they know our address is in Heidelberg they're working with us because our mission is to bring sovereignty to the Enterprise to bring safety and and trustworthiness and build this into our stack so for for me and of course maybe like European helps the brand in a sense right kind of there's this German company and of course they're gonna focusing on Enterprise
grade robust safe and explainable AI naturally so well kind of everybody else is kind of doing the playful b2c stuff maybe that helps the brand but I think overall um that's just what we're good at um I think German companies are very good at these complex processes at manufacturing uh things that require a lot of care but it's not necessarily that we have to be forced to do that yeah maybe to complement also um there is for example or there are several U.S companies one in New Jersey I think that is basically building on the gdpr using it to um yeah build up a Competitive Edge and it's they are really really growing growing growing because in the United States apparently it's a selling point to say we are following and are fully um compliant with the gdpi at least to those parts that make sense so again if we if we construct Cai act in the end in a good way I would completely agree that it yeah it could become a selling point
and with the gdpr and a few other legislation out there we see or we see empirical data that it's indeed working but it needs to be done in a good way so yeah you are right maybe if I've just quickly extend I mean I guess the question was a bit towards um the values that other countries have in other cultures I think Hansen elaborated on a couple of points in the morning and then it might just be incompatible right when you have systems that are giving answers which are not necessarily in line with the values of that country or that culture right so then it doesn't really change well this sort of argument that you have this type of let's say European valued AI right it may just be incompatible it's a bit different to gdpr perhaps maybe maybe uh double clicking on that I in in that we do have like somewhat of a different design philosophy as I don't think that I we should be building an art system that is outputting the
world as it should be um and then basically enforcing our I don't know like ethics Chief ethics ideas on what the world should look like onto all our customers and maybe we're kind of a little bit off the hook here because we're not Building b2c Systems I know that for b2c systems responsibility is slightly different but um our our customers and partners are world's best Enterprises and they can take damn good care of that themselves and so we're focusing on transparency and control rather than Shifting the output of our model to some some ideal output that we feel is is kind of represents our values the best and as Han stated in regards to also these ethical questions there are many areas when you follow this route at least you will have to answer who's going to decide what the truth is yeah and and this is a this is not an AI question this is a general very complex question uh but we which in within this discussion somehow right now is transferred into
the AI sector and the AI is guilty for for being unethical or or causing problems but the the real problem is on a different side and to actually also um include them when we make the laws but we don't want to put hardship onto our startups so that they have the liabilities or they they have added compliance we still want that Innovation there's a question one more question Joanna please we have three we have exactly two minutes and 58 seconds I'm confused about what to ask because I feel like you guys are are kind of having both sides of the cake or something you know uh the I strongly agree that um the duty so I'll tell a quick gdpr story I was doing a lot of work with Google um for pro bono for some reason uh on their on the policy and ethics and when when the gdpr was coming in and they were all saying oh by the way we're pulling out of Europe when it comes in they just lock up
and try to tell influencers complete BS right and then but I kept here working with them and she moved on this adiac thing they happen later anyway they after the gdpr came in and not only did they not they not pull out but six months later they're like this is great this is like an API to 28 countries and I'm like if the EU is a trade line and so this is why I was so angry that you guys are saying oh you're not thinking about the you know everyone is thinking about smes that is core to the EU values trust me I've been in a lot of meetings about the AI Act that everyone has always mentioned that they've always mentioned the the positives and the negatives of gdpr right we have been thinking about all these things and so a lot of what you guys are saying is great and I agree what the what the council tried to introduce with the you know generative AI it's nuts right there's this late minute
shoe warning that could disrupt the whole thing don't play into the hands of the American companies that are again trying to disrupt something that will probably help them as well as us right that just don't do these talking points and don't undermine our process do contribute to it but don't just make weird critiques thank you are there any more questions I think so here we're becoming engaged I love this yeah because I have to thinking about gdpr and also the air actor for a long time right I want to provide a very pragmatic solutions for you I think gdpr gives many companies the excuse also government not to AI data-driven approach therefore I think on the one hand Define the regulation on the paper but I think very important EU provides also two links standards for a person to identify personal data and also provide methods to a data privacy don't only deliver regulation also deliver tooling standards and then every company say this is the current standard technology which the world can achieve we
can only do as far as we can data protection because technology are not there but we don't have the excuse no don't have the excuse don't do Innovation I think this EU also who you act right I think about it only risk driven classification is too simple-minded if you're good to your kitchen you study the devices most of them are high risk but under which condition you are allowed to use your knife if you don't use your knife you will not have food right this this is also don't make the things too simple right I think this is very complex because it defines the future of our society not only one dimension risk also potential and which condition these use case can contribute to the innovation of the society this is by proposal thank you I would like to give the panel the chance to Now respond to your questions yeah very quickly and so I think we are completely aligned and my boss and I we are really trying to to push for that
we are all the time underlining again what I said AIS password you need to consider that there are a lot of different use case is not everything should be banned or is a big threat to human Humanity there are a lot of positive opportunities and so on so this we are really pushing and also the the SME point in the parliament yes everyone is talking about it but there are of um some ideas for example there's now a fundamental rights impact assessment brought in by um by the greens and social Democrats and so on I understand where they are coming from they really want to focus on the deployment of a high risk AI system and maybe to some extent it makes sense but there are things in it like you need to make a public consultation before you are deploying an AI system if this now every German or european company needs to do no one will use AI because it's too expensive it's too burn so and so on so this actually when
we are talking we want to help smes and startups we need to make sure that we are not um yeah putting in certain things that will definitely um unable them to to use Ai and so on but um the points that you were also or both of you were making again enforcement will be a key point and definitely we need guidance we need audience as a national competent authorities or now they are called National supervisory authorities to really be there prepared and have guidelines have common interpretations not differences between Bulgaria and Belgium Germany and again have a fragmented internal Market if we manage that I'm not 100 sure I really hope we are learning from the gdpr mistakes because otherwise and now this you both mentioned we again strengthen um Google Microsoft and core because they have the means the financial resources see um human resources to be compliant after six months even if they are not comply client say it pays a fine no problem but a German startup will struggle and in the
end if they are facing the risk of paying I don't know now after the parliament seven percent of their Global income as a fine they will think about there are no standards there are no guidance there's no clear interpretation of those 10 keywords no AI is too risky I may not invest and then we have again a situation that we are making big Tech even stronger and uh kind of achieving the opposite of what we want to achieve with the AI Act because it will we will fully depend on U.S and yeah non-european companies when it comes to Ai and we will be basically just the end user of those final products this would be a terrible outcome so I really hope that together we can avoid that unfortunately we have a run out of time I still want to offer Jonas or you'll get the the possibility to answer if you still want to add something to what Kai said no I I agree with everything just maybe wanted to add that um I
agree that it's not just about complaining and although I may seem to have complained a lot on this panel this is not the only thing I do to help out with a European regulation let's say the same for us when entering into discussions uh we mostly we come also up with alternative suggestions yeah how could we do it better and I think one one thing in the discussion this is frequently missing is the complexity and and and the uh the gray area yeah the uh it's really hard to Define what really is meant yeah and and I often enter into a discussion where I say if if you if if this article is is defined this way you can't do X yeah and the counterpart says no that's we don't mean this yeah I mean our motivation is a different one a good one I say yes I understand your motion motivation it is good but if you read it really a word by word you have this gray area and a CEO or a
VC who wants to invest or is responsible for these types of applications and wants to avoid risk he doesn't get into this and and this is is one one one one major construction issue yeah that we are trying to regulate something that isn't clear yet and we have this extremely large gray area yeah gray area Affinity though I hope unfortunately we need to finish our panel discussion today I want to thank you for your valuable insights I take with me that we need to lean forward more engage more be positive about it and grab it and form it to what we think we need in Europe thank you so much thank you for my audio