SZ 139: AI in the EU Ft. Yaad, Gerard, Florence, Julie, & Lutz

Alright, welcome everybody to the 139th episode of Silicon Zombies, where you get the best brains from the Bay to beyond. We get to demystify emerging technology, and as you can tell, we’ve got some very delicious brains on the stage today. So without any further ado, we’re going to start with you. We’ll kind of kick off with some brief introductions, and then we’ll set the table and we’ll have some fun.

So yeah, please take us away. Yeah, so thank you. I have two roles in SAP. I just introduced my first one.

I’m going to repeat it in five seconds. I’m the managing director of the 18 offices we have here, the biggest product in the US. But I also have another role that you may figure out by the longer the beard. I’m a developer geek, right?

I’m running the R&D unit, the global R&D unit. So I’m going to introduce myself. I’m the CEO of the R&D unit. I’m also going to be led by John R&D unit, big global R&D unit, the people in Silicon Valley, Germany and beyond, building, surprise, surprise, AI, in SAP platform, called the Business Technology Platform, and also we have some couple of hundreds of developers looking into the future of technologies, advanced AI, quantum and other cool stuff.

Wonderful. Thank you so much. Yes. My name is Florence L.

I’m a new professor in France. I’m currently a professor at Stanford University. I’m currently leading at Stanford a program called Governance of Emerging Technologies. So we spent months working on AI regulation, and we’ve just released a report that provides a lot of views to various types of regulations that were adopted around the world.

So, of course, we studied the AI Act, but not just the AI Act. We looked into various types of regulation in China, in India, in Japan, in Brazil, and, of course, we followed also what is happening and has happened in the U.S. Super. Julie.

Yeah, I need a mic. Yes, hello. Thank you so much. Really glad to be here.

Thanks to you, Chairman. Of course, thanks, California, which is the French-American Chamber of Commerce. My name is Julie Owono. I am the Executive Director of a nonprofit called Internet Sans Frontières, or Internet Without Borders, and we defend freedom of expression and human rights.

So we love to talk about innovation, but always being the one in the room saying, okay, actually, be careful. There are people’s rights at stake. In addition to that, I joined in 2020 the Oversight Board for Meta. So we help Meta do its content moderation, I mean, we help Meta make good content moderation decisions.

And… Yeah, right? And we recently published a paper, a white paper called Content Moderation in a New Area for Automation and AI. So obviously talking about the benefits of harnessing automation, but also being careful, of course, about the downside.

So glad to be here. Super. John. Yeah.

Well, I’m Gerard de Graaf. I mean, I was recently introduced by somebody who’s, I mean, we got relatively close to correct pronunciation of my name, and he said, less said the better. So I thought, well, that’s a nice welcome. I am the head of the EU office in San Francisco and also the senior envoy for digital to the United States.

I come from Brussels. I mean, I’m from the Netherlands, but I worked for a long time at headquarters. I was involved in Digital Service Act, Digital Markets Act, and also, of course, followed very closely the AI Act and all the other acts. So I’ve been involved in a lot of these acts that have come out of the European Commission and the European Union over the last five years.

And now I have, after having negotiated and designed a lot of these measures, now I have to explain them to people here in California that have to comply with them. So that’s a, it’s a fun part. Excellent. If I’m sharing your measures.

Yeah. So, hi, everybody. I’m Louis Spangler. I’m Louis Spangler.

I’m . I’m on Zara. And I’m a member of the European Commission for the Development of Digital and Information. And I’m also a member of the European Commission for the Development of International and International Affairs.

I’m also led by John in Europe. So I recently actually launched an online course where I supercharged me as a professor. I replaced myself to go to the actual studio, so I became supercharged. But you can imagine what it meant for other professors because we don’t need them anymore.

But for the good of technology, I guess. Did you ever get in trouble by saying, just say AI version of the event? Oh yeah, like the AI version is actually live, so you can talk to my AI version if you want. But this is the real news, right?

No. Go on. Like I can’t even put a foot in the real news. Wonderful.

Well, we’re going to dig deep. Before we do that, let’s set the table and maybe talk a little bit about, in fact, we’re going to start with you, Jorav, because you gave such a lovely first introduction to kind of kick us off. Clearly, there needs to be some kind of rules. It can’t just be, so I’ll give you an example of this.

Europe has had cultural moments that give us a different foundation of how we think about these things, collective versus individual. Who should be making these rules? Well, I think that’s a very good question. And it’s a question that isn’t asked often enough.

I mean, these technologies can have real implications on people’s lives. I mean, there’s choices that have to be made how these technologies should be rolled out and should be deployed. And be used. And I mean, it sounds maybe a little bit provocative, but I have said before, like, who elected Elon Musk and Sundar Pichai and Sam Altman to take these decisions on our behalf?

I mean, these are decisions that I think collectively as a society we should take. And so this is like regulation or not is like one issue. And I’ve given like arguments why the European Union has chosen to regulate. And this wasn’t done like on a whim.

I mean, we started actually looking at AI for many, many years. But looking at like governance arrangements like in 2017, 2018. So it wasn’t also something that like we just came up with an overnight put out a legislative proposal. But the fact that these decisions are so impactful on people’s lives and that we, of course, I mean, the industry is doing an incredible job here and pushing the boundaries of innovation and bringing all of the benefits to us.

But I mean, I think it’s important that these decisions have like. Legitimacy that people feel that like these technologies, the way they work and also because industry has its own incentives and we see a race to the market. You need to be first. This is the Internet.

I mean, you need to be first in the market. I mean, at least in the Olympic Games, if you come in number two or number three, you still get a medal. You get a silver bronze medal in the Internet. You don’t get a medal.

You’re a loser if you’re number two or number three. So there is a tremendous incentive for industry to roll out the latest generation. Yeah. Okay.

Okay. If you don’t regulate, the decisions are going to be taken by industry ascension. If done correctly, and we obviously need to make sure that we’re doing that with intention. Julia, maybe you can talk a little bit about the Internet Confluence here.

Yeah, sure. How do you think about scaffolding the right kinds of questions to make sure these rules are thoughtful and serve everybody? Yes, maybe the first thing that I would like to do is to encourage and congratulate, really, the European Union for always at the very center of all these issues, prioritizing a multi-stakeholder approach, right, where it’s not only, you rightly asked, who gets to make the rules. And what we appreciate with the EU is that it’s not just the MPs at the European Parliament making decisions or the Commission.

You know, setting these rules. It’s the product of consultations, right, including with civil society organizations like us. And, yes, sometimes we are the ones in the room breaking the general agreement because, yes, there are some minor, which might seem as minor issues that you don’t think about. I have a colleague, formerly, I did a fellowship at the Bergman Klein Center.

I have a colleague at Harvard University. I have a colleague who created this term, designing from the margins, right? When you design technology, thinking about the margins first will probably get you ahead. Not probably.

Will certainly get you ahead of the risks because the marginalized communities are the ones that are bearing the brunt first of what doesn’t work with technology. So it is, indeed, a competitive advantage. And we see clearly. This regulation should test the EU AI Act.

But even on other subjects, we could say the same with the privacy, the GDPR, or even now with the Digital Services Act in another technology space. So, yes, it’s important to have this multi-stakeholder approach, to have these different voices in the room, right? Not just this dialogue between governments and companies because both have real high stakes when it comes to technology. And we saw and we see it very clearly, right?

Recently in the United States with increasingly relationship being built between the technology world and the political world. So it’s important to have those outsiders who have also interest in using technology. We at Internet Without Borders, we do encourage people to step into the AI age. We do encourage them to take advantage of it.

But we also encourage people to be creators and to participate in this economy where… You know, wealth is being created by the ideas you have and the data that you produce. And actually, I wanted to talk about this because it’s an EU funded project that we have. We are currently funded to collaborate with another organization, another big organization called Into News, which is present in the United States, in Europe and in Kenya.

And in this case, we are very interested in working with civil society organizations in Kenya to think about how can we as civil society organizations deeply care about fighting hate speech online, for instance. Is there a way for us to fight hate speech in an automated manner that respects privacy, that respects free expression, but that also deals with the problems created by hate speech? So, yes, I think it’s important to have a diversity of voices in the room. And in that regard, the EU AI Act is really an interesting piece of legislation.

How do you define hate speech? Uh-huh. Well, that’s a great question. What is right speech and wrong speech?

No, it’s not a question of right or wrong, right? I don’t think that’s the most interesting part of the conversation. You asked me what is hate speech. That’s a great question because there is no clearly agreed upon definition of hate speech, right?

There is one in your opinion. Except, of course, I was going to say that… Illegal, illegal hate speech. Illegal hate speech.

But I think to respond to that question, it’s important to look… Look at the diverse amount of platforms or spaces where you can express yourself. And even the most, you know, rogue ones will have certain rules about what, you know, the extent to which you can discuss freely, right? So, yeah, all this to say I don’t have a definition, but we can work on that, right?

That is actually an interesting one because you just said you don’t have a definition. You said the EU has regulated… I said I don’t have a definition. We see AI as powerful.

We see it can go wrong, right? It is impacting our lives and yes, if we don’t make it transparent, then questions will ask, ability to ask, who is doing what with this technology that is shaking our lives. To however say that regulation is going amazingly well in Europe is I don’t know if I would draw a map, I don’t know if I would draw a map of what is happening in Europe, I don’t know if I would draw a map of what is happening in Europe, I don’t know if I would draw a map of what is happening in Europe, and put the sizes of the countries based on their AI impact. I don’t know whether Europe would make it over the size of Liechtenburg.

Liechtenburg? I’ll give it to the Belgian man. He can do that as well. Europe at the moment is not really on the map for AI, but we are regulating it because…

We say that it’s dangerous, which I very much agree, and the question is whether or not to regulate. We should, but we have a middle name on a normal street. The question is how do we useful regulate? I think that’s an easy statement to make.

Europe is not on the map of AI and we need to test it, we need to contest it. We have European companies which are quite successful. Mistral, Alephant, and others. We have Alpha, etc.

We bring over companies to Silicon Valley. We brought over companies in healthcare and life sciences. 15, 20 of them. More than half had AI.

They were in AI. In applications. Are they AI companies? If you would be here in the US, you would say, yes, I’m an AI company.

In Europe maybe, because that’s easier, you can get an investment for a venture capital investor. In Europe they say, no, I’m a health company. Or I’m a mobility company. Or I’m a energy company.

Europe isn’t a backwater on AI. Not at all. It’s not a backwater on AI. To make a statement and to have that statement unchallenged, like Europe is somewhere like Liechtenstein.

No, that’s complete bullshit. First of all, I like that we hit the ground running with this. You wanted to be provocative. You told me to be provocative.

S&P is welcome to host such compelling discussion anywhere. You just told me to. No, first of all, maybe… Does S&P actually have AI?

We have a lot of AI. But first of all, maybe I can share with you in a few seconds our approach to AI, which I think is relevant, and of course touch the compliance thing as well, and regulation, regardless which vendor you are taking. And as I said earlier, I know it’s surprising many of you, but me with my accent and ponytail, I know how many US politicians want to learn from the European how you do AI right. But also where not to go.

And I think we have a formula which is very interesting, and we call it the 3R. First, R is relevant. We don’t talk about it too much, but I’ll be very short. AI is an expensive technology.

It’s also not always sustainable. And as a technology, you don’t need this nuclear power for anything. Sometimes it’s like shooting a fly with a cannon. To be honest, some of the challenges many people have with AI can be solved.

And it’s solved with an Excel. Regardless of which vendor. But it’s education of the people. It’s a very strong technology.

And the relevancy, and SAP of course is the leader of enterprise applications. We know the business very well. And this is why we try to apply a business problem that justifies this. And have over 27,000 customers.

But for AI. But the second R is reliable. And I agree, reliability. The technology is not fully accurate and safe.

I’m saying it’s the technologies. There is a lot of things to regulate. There is a lot of protection to do as well. And the reliability of technology also goes to the vendors as the need for good data and anonymization of data.

It’s like cooking, right? You put junk ingredients, you get junk food. So now the data needs to be in a higher level. And this is something that all vendors need to work on the reliability.

Of course also policy maker. But without reliability, it’s just going to break. Anything regarding ethic, IP, and bias, the risk bias in the main models out of the box. I’m very proud of SAP.

It’s a very European company. We’ve always thought, even before any deals, any acts, also it’s our differentiator. We know, maybe SAP is not the most sexy AI jumping avatars and others. But if you want to sell for enterprises, you must have a lot of what’s happening on the iceberg.

A algorithm that protects the algorithms and even takes care of removing bias and others. And here I’m very proud, we have thought leaders. And this is what we do. So yes, great, I think, hope coming from European companies.

Because the mind thinking, right? And my boss is also, hold your feet to the fire even if you don’t have a sign or bill yet. On the other side, I just want to say one thing about the regulation. I think it’s a must, the need to protect the AI.

But I think a lot of clarity is needed. I think we are throwing a big blanket with some of the acts, like the definition of AI systems. It’s very broad, it’s no secret, it’s all over the news, I’m not part of Apple, but Apple intelligence, a very strong powerhouse is not released in Europe because of clarity. It’s not saying compliance, because they are afraid.

And I think sometimes there is a lot of work to be done to define it. It’s like a surgeon, it’s something very huge. You cannot throw a cap at a net and just say, I’m going to use AI. And I think now, and Julie as well, in Meta, Meta 3.2, which is a great model, open source model for AI, also not released in Europe.

And I think it’s because of clarity. And I will end by saying, and I know I’m going to say something, that it’s also the big equalizer. I’m not working for Meta, but it’s an open source model and it has open ways, I’m not going to geek out, but it gives you a lot of free stuff. If you are a startup, you want to use AI and again, it’s expensive, but they give you a lot of open source free stuff that you can use.

So the truth is, we need to regulate it, but we need to define the risks, and what is lower risk, and clarity, And this is where a policy maker needs to meet like me, industry leaders, to define, and it’s a lot of work. Actually, this is not really, I have a question from the thought process behind regulation. We heard earlier on the regulation of a hair dryer. I don’t think regulation of a hair dryer is a thing, because hair dryers have to be managed.

It became a thing after, unfortunately, people met a conflict with the hair dryer or dropped it in the baths at the wrong time. So then we started building on safety nets around it. Now, if we think about regulation, how do you see that? Yes, about regulation.

So first, what I could say, what I want to insist on, as I’ve said, is the fact that as you said by your heart, if you want to develop an AI model and a very advanced one, you need substantial resources. You need compute, you need data, you need to invest massively, which means that today, it’s industry and in those major AI companies that you see the long-scale transmodels being developed. And as a result, you have a very important asymmetry of information and of expertise between the governments, the regulators, even to a certain extent people like me, the academia, and the industry. So it makes it very difficult to regulate and to regulate properly.

And of course, it’s hard. What I think is super interesting in what the EU is doing with the AI Act is that the EU has started a process which is dialoguing with the industry to implement the AI Act. So you have those codes of practice that are being drafted at the moment that involve many, many types of stakeholders the industry, but also various types of experts so that you can find at the end of the day clarity, definitions, precise rules, so that the regulation is well thought, efficient, and you have, to a certain extent, a regular agreement. Of course, you will always face some pushback, but I don’t think this should prevent regulators from regulating.

I think it’s important to understand that the EU is not a government that is just a government that is regulating. As soon as you can have this dialogue that is a quality that you will have something that is really well designed. And my second point about regulation is the fact that in practice, when you look at what’s currently happening around the world, you can see that the AI Act is actually influential. Even here in California, you have a few statutes that were adopted over the past few months and weeks that are clearly influenced by the AI Act.

For example, transparency about data is something that clearly was inspired from the AI Act. You see there is a pending regulation in Brazil, there is a pending regulation in Canada, and my last point is that even in countries that for now said, we are not interested in regulating, we are afraid of stifling innovation, like the UK. This was the UK’s position, and it was the UK’s position in the last 70 years. The UK started this process with the AI Safety Summit that took place one year ago.

There was another one in South Korea, a virtual one in South Korea. The next AI Summit is in France, and all the summits are the occasion to discuss the risks. And it’s super important because you have in all these countries a new AI Safety Institute that are currently trying to discuss these industries and ways that they are counter-productive and want to have a better understanding of the risks and what should be done. So I totally agree on the transparency part.

It’s super important to actually have models that we can look into and how they are trained in order to understand what is under the control of the AI. So that we can actually open it up. There is a balance to strike, obviously, between… And this is something which in my view with Europe gets slightly off if we talk about regulation.

It’s not only about citizens. There are three entities, right? There is the state and Europe, national security, the South as the state, there are the citizens and there are the countries. And what is interesting is President Emmanuel Macron recently said we are over-regulating us out of the Margaret, I’m sure about that, as it came down to AI.

And the reason that he brought this up is that he essentially said we have created so many rules for things which we don’t yet understand. And since I’m working with the technology inside our catalogue, many of the arguments which were put down don’t necessarily… Transparency makes perfect sense. But other things in terms of making it explainable.

A neural network is built like our gut-feel. I mean, you all have been in this hiring situation for some of the three seconds you knew what you would do in the next 27 minutes trying to figure out how to explain your decision you just made. That’s a neural network. And now we are saying that it needs to be explainable super hard.

It’s not really feasible for us to explain, but we already wrote it down into law. And that’s where I’m kind of thinking it’s not about should we regulate or not. Guys, we should regulate. This is a huge technology and we need to figure out how to work with it.

But can we regulate at this level? And I would agree with Emmanuel Macron that it’s probably a little bit wrong at the moment, indeed. Sorry, it’s interesting that France voted in favour of the EUAI Act. All member states did.

It passed unanimously in the Council. It has a very strong support in the European Parliament. So that, I think, is a fact. I mean, the EUAI Act isn’t kind of rigid and set in stone forever, because the way we also regulate, and I’m absolutely in agreement we need more clarity on how this is going to work.

And some of it we will simply find out as the regulation kind of ends. And that will then get us into effect or into application. There’s a number of elements in the way we regulate which is different from the way that the US typically regulates. We have more empowerment.

I mean, we call this Delegated and Implementing Act. And so in the legislation, the executive, which is the European Commission, is empowered to take certain measures if it feels that that’s necessary. Of course, this is well circumscribed. It’s not like unlimited power.

But if you take, for example, the model we have a threshold for models which is different from the threshold in the executive order of President Biden. There is, I mean, this gets very technical, but it’s 10 to the 26 flops, which is like bigger. We have 10 to the 25 flops. And this has to do with the fact that if we would have said 10 to the 26 flops there would be nothing within scope for a politician to regulate and then say okay, how it’s going to affect the industry so that there’s actually nobody who’s going to be affected by it is a little bit difficult.

So 10 to the 25. Now some of the industry says, oh, you’re going to be overwhelmed. You’re going to be inundated. There are going to be maybe 25, 30 models that meet that standard.

If that’s the case, the Commission has been empowered to change the threshold. We can go to 10 to the 26, even to 10 to the 27. We can go down as well. We have this opportunity to provide guidance, which we will definitely do.

And then a large part of this, I mean, we call this co-regulation. In the US, it’s a very strange, people don’t really understand when you talk about co-regulation. They understand self-regulation, they understand regulation, which they call guardrails because I guess regulation has a negative connotation. Guardrails is a little bit more like of an innocent term, presumably.

The co-regulation means, again, as I’ve explained, standardization. I mean, the European AI Act cannot work without standards. We have given dozens of mandates to the European standardization organizations to flesh this out in a way that makes sense. The standardization is not just done with the industry.

There’s also stakeholders. You can’t just work with the industry because, I mean, yeah? May I speak to it real quickly? You talk about empowering these legislators, but they’re not the ones building.

Don’t you think it should be the entrepreneurs who should be empowering instead? Why? Because they’re the ones who are building the technology. Yeah, but I mean, we talk with the entrepreneurs.

We talk with the regulators. This is regulation. Regulation needs to be enforced. Is that not mud on the racetrack that just slows us down?

Well, if you mark Max Verstappen I mean, he’s not typically slowed down by mud or rain on the racetrack. He actually gets better at it. And maybe you need to be a bit more careful when the weather is bad. So that’s not necessarily a bad thing to go at a really high speed when it’s raining or there’s mud.

I mean, that’s dangerous. You can get hurt. No, this is regulation. All joking aside, the regulation has to be implemented.

Regulation that’s not implemented is not worth the paper that’s written on. Who is implementing the regulation? Well, the governments are implementing it. I mean, that’s the role of government to implement the regulation.

Of course, we will listen to everybody. Like the industry is telling us, 1025 is probably going to be a problem because there are going to be too many models in scope. Okay, let’s see. Legislation will kick in the first of August of next year if there are so many in scope.

I’m sure my colleagues and this is also, it’s not unlimited powers. It will have to be checked with the European Parliament and the Council of Ministers, which is like the Senate and the House in the US. They will scrutinize whether we’re using that power. But what I’m trying to say is that we have flexibility built in to the legislation and that’s often the argument.

You can’t regulate because it’s moving too fast. Well, again, we are not regulating the technology. We’re regulating the use of the technology. We have regulations in place.

I was myself involved. The GDPR is going back to 1995 and then 2016. I was involved in the e-commerce directive, which was until the DSA was the framework in place for regulating kind of the online platform economy. We celebrated the 25th anniversary of that framework.

So the idea that somehow these legislations will very quickly be kind of overtaken by events is that it’s not again underpinned by evidence. I mean, give evidence of that before making these far reaching statements. And one of the reasons why we avoid it is to build in that it can actually be changed. You can change standards as well.

We can give guidance. We can change the guidance. It looks like Yad may have something to add. No, first of all, I’m honestly fascinated by this.

I just want to try to draw a bridge and maybe a question to you eventually, because I think, and this is an observation, right, that we are talking a lot about the ex-ante approach. Try to perceive, and Lutz, your point, what will happen. But actually, Florence, I thought the UK is a great mechanism of ex-post as well. When they gave some guidelines to the industry, they and I want to get your observation on this, they diluted a little bit the standard and the finding in the billing and gave the front runner, right, Nick, to your point, the one that invented technology in industries, guidelines.

Now you can say, wait, wait, you’re not regulating them, you are finding them. I mean, any business leader in any company, if he had the sword on his neck, would try to go as much as he can to maybe even touching the border lines to make value to the stakeholders. But I really agree with you, Gerard. Without the selfish incentive for businesses, even with guidelines, is to trust the technology.

Again, I’m being very honest, SAP is super successful. AI is not that sexy, but our big differentiator, reliability, responsibility, they trust the technology. Without this, nobody would use this. So I actually fascinated by the UK.

I know we have also the British American Business Journal folks here. I saw them on the slide. So I wonder if this approach of ex-post and guidelines can help not replace regulation, but balance a little bit, and maybe help to solve some of the problems we heard. Yes, well, yes, so I, the UK has a tradition to regulators talk to the industry, go back and forth.

I talked with various AI companies who told me, well, we get very clear information any time we ask questions and then we try to get information about the standards that we could apply. And yes, this is something that should probably be improved in the EU. There is this problem with GDPR and how GDPR applies to AI and specifically generating AI. So yes, it’s true, so we should be, we should probably have a better coordination between the various authorities that are in charge of implementing our EU regulations in the EU.

What I would like to add about this idea of ex post regulation is that I, I’m fascinated here because when I arrived here I learned this expression regulating through litigation, or which is something that I find so interesting as a French lawyer, as a continental lawyer who was trained with the city code, and this is what I’m currently experiencing because I’m following those claims and copyright claims that are currently pending here in the US. And so sometimes, you know, thinking ex ante can be super helpful because you have massive, massive costs associated with these litigations and I don’t know how much time it’s going to take to have clarity about, you know, the copyright rules that should be applied in this fair use exception. But I think that sometimes regulating rules ex ante can be super helpful. May I ask Julie this, I mean, new technology AI is in existence in 1956, we had other new technologies coming in between, so and we didn’t do well with them, just to be clear.

Facebook, social media in general, people died because of the usage. Like, they were unrest in other worlds. In other countries. So the question to you would be how do we regulate this and could we have regulated at a moment when Facebook just got started?

Would we have known how to regulate it and how do we regulate it today? Does it even work? Is it a thing? Let’s do it!

And oh yes, the EU will act very quickly. That’s not true. I’m sorry, I have to dispute your statements all the time. This is a new case, guys.

There is new technology coming on. We have seen in the past that this technology was used not in our best interest of society. But we all use Facebook. Or Snapchat, or Instagram, or whatever.

So it has value, but it wasn’t always the best. So when should we have regulated it? Do we know how to regulate it today? And are we doing it right?

I think that if you look at the history, the very conversation that we’re having today about AI, is probably the lessons that we’ve learned from social media. I think. When Facebook was launched and there was this there is still this famous phrase, you know, move fast and break things. We’re in the land of move fast and break things.

At the time, we didn’t think about the negative external externalities, which is innovation was central. Now, we are all in this room to discuss about a regulation. Which is historically, I think, a quite interesting and big step in the right direction. We may not agree, all of us, on the details of what regulation entails, whether lawmakers actually understand technology or not, and whether technologists actually understand the importance of protecting democracy, because that’s also another very important debate.

But, it remains that we are, and I would say virtually all around the world, there is agreement. I mean, you mentioned the Black Sheep Declaration. China signed it, right? Because, preoccupied it.

That’s not the question. I mean, the very fact that governments are preoccupied and want to make sure that humans remain in charge of the innovation is a very important step. And it’s also an opportunity even for entrepreneurs. I don’t think there is a dichotomy.

I think there is alignment, really, on the importance of building trust and using that trust to better perform and to, yes, and to prevent yourself from litigating, from spending hundreds and, I don’t know, lots of money on lobbying, on PR campaigns to protect your brand, because no, you don’t want to be a part of that. And so, I think, historically, we are in a very interesting moment. As I was saying, and I’m very convinced, the European Union is really has taken the lead on this. We might not see it right now because we are in a very difficult time in the world.

And so, we are very preoccupied by imminent dangers. But I think in the mid to long term, it will prove the benefits for societies like the EU, but also elsewhere, to think before you act, definitely. So, yeah. One point.

One of the hard things that we can measure by is the amount of capital that’s going into these things. If we think of a metaphor of the automobile, like the amount of gas, the capital that’s going into these, my understanding and you can fact check me here, is that the US has 5x the amount of capital in artificial intelligence. So, I guess if you can help us understand when you say that the EU is in fact ahead of the curve, can you maybe unpack that for us a little bit? Ahead of the curve in terms of innovation?

Innovating? Yes, I… Ahead of the curve because I think because of the history of the continent, I mean, technology is really great, but let’s remember that technology can also be very, bad. Like, I don’t want to, you know, I don’t want to have the…

Sorry. There is an expression, you know, a point, a good wind point. I don’t want to reach that point, but just giving the example of the nuclear bomb, right? It has so much, so many great…

It’s a progress, but it’s also a huge danger for humanity. I think AI is probably to some extent exactly the same, and that’s why contrary to what you said 50, 60, 70 years ago, we have nations that finally say, yes, innovation is amazing, but just let’s pause a few minutes with this but, and save us years and probably a lot of money and lots of problems that will cost us probably even more. That’s why I think there’s… Yeah, maybe coming back, because you…

That’s why I’m fact-checking you. I mean, you’re right. Europe is a deep tech continent. I mean, this idea that somehow Europe is a desert where you can hardly find any technology is a complete fallacy.

I mean, and where is Europe strong? Europe is strong in clean tech. Europe is strong in healthcare, in mobility, in energy, in agriculture, in food quality. Maybe these aren’t like the big tech companies, but they are there.

I mean, in Germany is a very good example with the Mittelstand. There is tremendous innovation going on in Europe. We have a scaling up problem. That’s a problem that we have.

Companies like NVIDIA works together with 13,000 companies around the world. Seven and a half thousand of them in Europe. NVIDIA wouldn’t be the company that it is today without the intellectual, the research contributions of European companies. Many of which are located in the country that you know best.

In terms of… And I think it’s important to realize because I don’t… I never accept Europe to be put down for no good reason. If you put it down for good reason, okay, I mean, good friends, but not for no good reason, not without evidence.

In terms of like Europe acts quickly, I’m not sure Europe acts quickly. If you take the EUAI Act, we started in 2017, 2018, because we’ve got these processes with like, we do white papers, we do green papers, we do consultations, we do impact assessments and everything else, and then we negotiate for two years or longer. So is that quickly? No, it’s not quickly, but it’s timely.

Because this is, to our view, I mean, the US always says, yeah, yeah, we need to regulate, but not now. Well, the European Union says, yeah, we need to regulate, we need to do it now. And that also has to be a bit to do with like our kind of… the way we…

maybe we think, we act, we think about risks before this point that you made earlier, like we only started regulating hair dryers when people got electrocuted in bathtubs. I hope not, but I mean, we are talking here about risks who are already… they’re manifesting themselves. I mean, the country I know best in the Netherlands, the government almost fell.

There was a scandal around AI used in social security benefits where 70,000 people were suspected of fraud, social security. Their kids were taken away from them. Some of them committed suicide because there was this algorithm that had bias in it. It was kind of singling out…

Can we fact check that? There was a human error and not an AI error. No, it wasn’t. It was a human error.

Yeah, maybe a human error in programming the AI, but it was definitely… it was an AI… an AI driven algorithm. And so, these problems, and you can pick them up, they exist.

This is not imaginary problems that somehow AI is going to be… Actually, so, the EU regulates, and we’re talking here a lot about, oh my god, all those risks. We talk about trust, and guys, who is using 3015, and who is trusting OpenAI? I mean, you can use without trusting.

But that’s a point, right? But we’re talking a lot about risk. The US is doing as well regulation. Will you sit in a Waymo if it was unregulated?

Would you go into a plane if it wasn’t FAA certified? I’m trying to make a different point. We’re talking about all the risks and how it’s terrible, and therefore we need regulation. The US is regulating as well.

But the US is regulating with a slightly different mindset. They’re regulating in order to protect their home market. You earlier on talked about market access. I see that the US is regulating quite a lot, and the latest regulation will all kind of push back from China to actually make their market safe.

You also mentioned that the EU’s regulation may not be fast, but it’s timely. That could be true. Historically, but artificial intelligence is more in fashion than any other technology throughout history. So I think maybe it’s important to reframe how we’re thinking about this.

Yann, maybe you have some thoughts here. I think it’s a great example for the hard work we need to do as an industry, right? The truth is in the middle. There are high risk scenarios.

I would never go to a plane if I’m not trusted. And there are lower risk scenarios where we can be more liberal and economy will fly over. I think there is a great example. We met in this office and if you follow me on LinkedIn, I also published it.

The Democrat senator was sleep, right? Bob Casey. But the point is, among all the other bills in the US that was mentioned, he tried to promote a great bill called No Robot Bosses. Maybe you heard about it.

And he came to learn, many other companies as well, I give it to many politicians who want to learn from the industries, many of them are humble. And in one manner, there is fear that humans will be hired and fired by the machines. This is wrong, right? You don’t want to get hired and fired by AI bots.

This is non-human. Wrong to bias. So now you can say, wait, it’s an AI system, human resource software. But let’s divide it.

Like, do we touch employee data to some extent? Of course for hiring and firing, no. You mean in the loop full speed, not even let maybe the algorithm have a secondary role if ever. But then you have also in human resource software education and training.

And back to the big equalizer. If you want to build a development plan for someone in a rural area like Paris, in the US, Germany, and you want to learn about their preference, what they would like to learn, right? I want to become a developer. I want to become a nurse.

I want to become something like that. And you maintain this information about developer preference. Of course, it can be very horrible for marketing purpose and others. But on the other hands, economically wise, it can be as I said, a great equalizer.

Now you create these micro campaigns, right? That are information about your preference, what you would like to learn. Not gender, not age, right? Maybe they are a little bit less risky.

And then you can create a lot of economical boosts with lower risk, relatively minimum risk. So the bottom line, I think it’s a framework of high risk, low risk. I know we try to define it also in many acts and also in the US, also in Europe. I think we can all do better job there, right?

And then I think we can opt in. It’s all about risk management. We will never stop, you know, it’s like AI is like water, right? A river.

You have pollution inside. But will you stop the water? No, you put screens, right? The screen is and this is where I think the truth is about framing high risk, low risk.

We have this clarity and much more surgical thing that needs to be done and it’s a lot of work. So we’re going to start doing Q&A, but so Valerie, if you wanted to, maybe hand the microphone now. And while she’s doing that, we’ll let you respond real quickly to that. No, it’s actually, I like the idea of that there’s a risk profile and you regulate around risk.

We also, however, should understand that risk and financial rewards are very much coupled. And our society, right? If you have an investment which is low risk, it is preferable to high risk. AI can predict risk.

Now, we as societies, we need to figure out how we work with AI. But that’s not a free decision. Because once we regulate AI in certain settings, we say this is all good, we need to reduce risks and so on and so forth. For us humans, we might introduce a more complex system.

And you know this about it. Like my example here is how do we in Europe regulate car insurance? So men are bad drivers. Sorry guys.

Men are worse drivers than women. So if you build an AI model to predict the likelihood of your accident, you would actually want to know the gender of the person driving the car. Now the EU said, no, we actually would not like companies to give information. It’s a totally fair decision for a society to take.

Because men are born into a great gender, they can’t do anything about their justice to wrong. And therefore the reason that they should be charged more for the accident, might not be fair. Society’s decision. But what happens is now obviously the risk you need to insurance, insurance company goes up.

So overall, the insurance market became more ineffective. And overall, we as society need to pay for that. Now, we are an international kind of collaboration. Other countries don’t do this.

Now, if you look at China, how they work with data, do I want to live in such a country? Definitely not. But there is a clear decision for us to make of how we as a society work with data. And we need a regulatory setup of how, not only for us, but how we deal with other countries.

It’s a little bit like labor laws used to be. As we started to see price down to a little bit. Can I maybe say something about insurance here? This is not gender based.

If you pay your insurance premium, it depends on how safe you’ve been as a driver in the past. And I know women, my wife is one of them, that I wouldn’t want to insure. And I know men that are driving really, really safely. It’s not done like, you do it on the basis of the history.

This is how the insurance industry works. It’s a risk based approach. You can’t just say women pay this and men pay that. That would be unacceptable.

But you can say, look, you have a good driving history for the last ten years, and you have a bad driving history, maybe you’re a woman, you pay more premium because you’ve got a bad driving history. That’s the way it is. This was a real European decision. This is like the first three years of your life where you don’t have a driving history.

They use gender. And the EU says don’t use gender. Use age. Let’s start over here.

Yes, hello. My name is Christina . I am a EU licensed attorney. I have an office here in the Silicon Valley called Side Law Firm.

And we’ve been accompanying clients here in the Valley and worldwide on EU regulation and making it easy for them for about 15 years. And I would like to highly and with all respect disagree with what I’ve heard. I have clients come to me from the AI corporation and industries saying they don’t care about EU regulation. They come back to me the next day and say, oh, wait, I didn’t understand.

This was not about regulating the technology. This was about regulating the use. And I want to sell my technology to people who use it. And now I understand why I need to address this.

Because all of the AI companies here in the Valley don’t do AI for the sake of AI. They sell. It’s transactional. And the use is what the EU regulation is after, not the technology.

And to that extent as well, as a point that Gerard has made as well, the EU regulation is one of the most extraordinary one that has a two-fold side. Protecting against the risks, but also encouraging the growth of the industry. It has two frameworks of components. And same thing.

The clients come to me and say, okay, I don’t care about EU regulation, but I want to benefit from the effects of the sandbox, from the data state high quality that the common factories are setting up. How do I do that? So there is a lot more interest than I’ve heard in this discussion that is expressed from the industry, not because of the technology, again because of the use. And then I’d like to make a point also to Florence’s point.

When some clients also come and say, I don’t feel heard in this regulation. This is all the politicians, the MPs. When is the industry heard? I mean, I’d like to remind that the codes of practice have called out for stakeholders, specific and multi-sectoral industry inputs for the codes of practice.

The call has been open for over a year. It closed in April. If you were there before and you answered that call, fine. If you didn’t answer that call, tough.

Right? And then last point, I’m sorry, to Julie’s point. I think you need to go back in history and I agree with Julie in saying we drew the lessons of what we’ve done on the regulation of the society of information. Let’s remember that the AI Act started before that in 2020, but mostly from March 2021, from the Conference on the Future of Europe, where there were a call for citizens to make AI a tool for access to information.

That was a marker before Thierry Breton did the first act. So initially AI was conceived as a tool to favor and promote access to information, drawing from the lessons that we had from the past on regulating on information. Thank you. I love the idea of regulating use cases.

I actually think this is the right approach. We should regulate use cases. We should not regulate what data sets to use or how, like, in the weeks we should regulate use cases. We should regulate the data sets that we use.

So let’s look at Europe. Let’s say we want to make an algorithm or a company that helps you predict fall risk of elder people or avoid hospitalization. I actually love those schools. You’re not that lucky if you try to do this in Europe because it’s very, very hard to actually get to the data.

Because of all the regulations around privacy, those are good regulations. Don’t get me wrong, I don’t want my data to be used, but the fact is if you want to build an AI tool to do what we probably all agree help elder people to live a better life or not get to a hospital so early on, then we fail. Why? Because there is a lot of regulation hindering us.

No, I mean no. No. I couldn’t do that. I was almost hoping I could let you finish in one hundred percent.

I wouldn’t have to intervene. My first comment, before I started talking about the EU AI Act, was about data spaces. We have lots of data in the European Union but they are distributed. They’re everywhere and therefore nowhere.

Health is the first place. We have a lot of public health data in hospitals, in health systems, in insurance companies. That is now being made available. It’s GDPR compliant, it’s cyber secure and you can have access to that data.

We need to do that in the US. I spoke with a health company in the US. They have to pay seven, eight million dollars in order to get access to that type of data. So here, unfortunately, you need to regulate.

You can’t just expect people to voluntarily say I have all of this data. Here it is. Use it. They won’t trust that their data will be safe and everything else.

So you need to build the infrastructure. You need to regulate. Sometimes you need to regulate in order to innovate. You need to let the industry do whatever it likes.

No, the world is much more complicated than that. But hold on. Like, just, I think we should regulate. Transparency, we should do.

We should use case regulation. But we should be careful to not regulate the actual tools because we don’t actually have a plan. Would you like the idea of regulating data spaces? Do you like that idea?

So that there’s health data all in one place that can be accessed and can be, value can be extracted from that. Do you like that? Do you think that will happen spontaneously? You need regulation for that.

I believe you’ve read the 459 pages of the Act and it is not about regulating the tool. It is nothing about the tool. It is about the use. The use that have risks on human rights.

The risk, the application of AI. Nobody, nowhere says that AI is not allowed to exist anywhere in the world because it’s so bad. It’s what we do of it. Again, the transparency is there to regulate or to give some transparency to the use.

None of those pages are dedicated to the technology as is. It’s not an AI. When you talk about explaining about AI and the whole world sits there and the research part and kind of thinks, what are they talking about? Because we don’t really know how to do this.

It’s like, yeah, you are regulating the actual, not the use, you are regulating the actual tool. Sorry, I have a question here. I understand, of course, you talk about GDPR and the old, the grossness impact and now it’s starting to go over the US. Now we are looking at the AI act which I think is more an ethical perspective and I think it’s much more difficult to have an international approach to ethics.

To have the same ethics around the globe. Some people will just say like the Chinese might have a different approach to AI surveillance and all these things. So if they are using a different approach that for us means we have this accurate world, even more because it’s not aligned in the approach, doesn’t that in the end mean we have to have separate AI products and we have to show the markets of things that have been developed elsewhere because they do not comply with what we understand, right? Where does that lead to?

Counter globalization? But that is the reality. That’s what I used as the reference to market access. There are certain applications of AI that we don’t want to see circulating on the European market.

And social credit scoring systems is one example. So if there’s a Chinese company think that it can be successful in the European market selling that type of AI, we can give them like a clear answer. No, that’s not what we want. And yes, technology is never neutral.

There’s values. AI is there’s values behind in AI. But there’s values behind in many technologies. And so yes, you will see different approaches.

I mean we are democratic societies. We have like fundamental rights, human rights and that’s how we regulate. And then you have societies that are based on like different principles like China or Russia or other countries that I think we know what we’re talking about. And they will regulate in a different way.

And that’s just, I mean we can deplore it but that’s the, I mean we don’t want those products in the European market. And they don’t want the first thing that China regulated was this, I mean Florence he studied it extensively was, I mean these large language models. I mean try to ask the large language model in China what happened on Tiananmen Square in 1989. Well you can’t.

Because they see the large language model as a potential threat to the communist system. And so this is how they regulate. I mean if you think that the European Union regulates heavily, go and check out China. You cannot have a large language model without a license.

So there’s, but indeed, I mean you’re right. I mean but that’s also with the internet already. I mean Meta is not available in China. I mean Google is not available in China.

So that’s a reality we’re already living with. Okay. Yes, thank you very much. I’m enjoying this conversation that we’re having here.

It’s a great discussion which I like to hear whenever AI is being discussed. So touching on two of the gentleman just now speaking about and how the global nature of AI and particular policy and regulation is. I’m from Uganda. And a lot of the discussion here has been about how AI is as it relates to the global north and the regulation around that.

When we consider that in 15 years a quarter of the planet will be African. One third of the labor force will be African. What does the EU start as it relates to regulation and policy in that part of the world? Is it something where they feel that they’re going to take a position of leadership in it?

Or do you think that you’ll be able to resolve some of the things that we’ve so passionately discussed today so that the Africans don’t have to adopt imperfect or substandard forms of regulation? Or is it a situation where it’s going to be we’ll just leave it to the hair dry kind of thing and the real African stuff? Well, I mean it is absolutely critical that this technology benefits the whole world. And we talk about regulation.

I wouldn’t start with regulation. I would start with infrastructure. I mean you cannot be. There’s only 36 countries in the world that have the capacity to develop large language model and really sophisticated AI.

If you look at Africa, Africa has one super computer. It’s in Senegal. It is a second hand, I mean it’s a used model. It’s almost like a second hand car super computer from France that France donated to Senegal.

I mean you cannot develop like a thriving AI and data centers and data and all of these things that like the ingredients of AI if that’s not available and it’s not just a problem of Africa it’s a problem in Latin America, it’s a problem in many Southeast Asian or Asian nations you are going to see a divide. You’re going to see a massive divide and a divide that’s going to widen and so regulation is definitely I mean also Africa I think you have 54 countries in Africa. I think if all 54 countries develop their own AI regulation it’s going to end up as a mess effectively. So you will need some kind of coordination but that’s not where I think here and we have tools and I mean we have like digital gateway etc.

We are the number one developing aid organization in the world. This is where we need to bring and also the connectivity and everything else to Africa. We need to help them with the skills. We have to help them with the administrative capacity in a way that allows Africa to develop to its own full potential extract the value from AI in its own in its own kind of context in its own economy and society and there we are not succeeding yet.

I mean I heard some of the statistics yesterday on kind of even like electricity that large kind of I mean 50% of Africans don’t have access to electricity. And so the basic necessities are still considerable and huge. But if we are going to make this a technology that can benefit the world and of course I mean we’ve seen in some instances in Africa I mean Kenya, Uganda for example with leapfrogging I mean when like with the mobile I mean medical services, education services kind of provided over mobile communications. I mean Africa has huge potential I mean a young population, I think 50% of the population in Africa is like under 25.

Give them give Africa, make sure Africa has access to the technologies and you will see magnificent things coming out of Africa. And that’s I think that’s an obligation we all have as a world. In 10 fastest growing economies in the world or in Africa? How do we drive accessibility and make it fair?

Yes, that’s a really great both and your question is absolutely great. I think I would I am one of those, provocatively who want to position the continent not just you know as a consumer of product that’s come from outside but also as a huge contributor. One of the commodities of AI is data, right? On the African continent, yes there are connectivity problems but yet you have huge huge billion amounts of data that are disseminated all around the world that are feeding the language models that we’re using including the multi-language models that we’re thriving to develop more.

And so you respond to your question and your question also the gentleman was referring to the ethical approach of the AI act. I think it goes way beyond ethical approach. It goes on the very question of human rights which have nothing to do with ethics. China has signed declarations for human rights.

Of course it hasn’t signed the fundamental charter sorry, charter for fundamental liberties and rights in Europe, of course. But nevertheless we all have an understanding that virtually the whole humanity including the United States and I’ve learned that through the work I do at the oversight board has a common understanding of a fundamental right and how it should be protected. Yes, I think so. And to get back to your question about the approach taken by the EU and other regulations, I think that’s where the strength is going to be.

If you look even for companies, if you look at a company like Meta, I’m not here to, I don’t work for Meta I should say. We are an external body to Meta. So we are I have absolutely no interest in them developing big language models. I don’t care about that.

But what I like really much about the approach they’ve chosen and others is that they focus on their role in respecting and mitigating harms against human rights. Right? If we as a company harm human rights in any way how do we repair this? How do we mitigate this?

And this is also a central approach I think in the regulations that we’re talking about. And for me, this is probably one way forward, right? It’s not going to stifle innovation. On the very contrary, it’s going to encourage and support innovation because it allows all the stakeholders who are contributing to the AI economy it’s not just entrepreneurs, it’s not just technologists, that’s not true.

When I spent hours on Instagram, okay, commenting my information is being used and Florence was reminding this of us of this. So all this to say, long story short, I talk a lot, sorry. I have four boys and I have to make sure they understand everything I say at home. I’m a minority there.

So long story short, the human rights approach that we’re trying that we’re thriving for is going to be extremely essential in the technological development of tomorrow. If we choose the ethical approach, we’re going to have big problems. But if we choose the human rights approach in which we all thrive to make sure that the right to life is protected, the right to free expression is protected, the right to privacy and you name them, I think we’re going to be in a better space. But that’s easier said than done.

Any more questions? Anybody here? I have the microphone. First, this is awesome and thank you to all of you.

This is a very engaging and interesting discussion. And this is a question for all of you because you’re all connected to this in some way. And it’s kind of a callback to earlier in the discussion. It’s very entertaining, educational.

But there is a part about the involvement of private companies and being able to sell and all of that. In all of your opinions, I guess according to this regulatory act, when does the seat for an insurance company or a meta or a company that’s wanting to use AI and their data in a very unique way, that seems to not be allowed at this point in time, when do they get a seat at the table to be part of the discussion and moving forward so that they can take part and they can sell? When do you stop maybe potentially impeding potential commerce? Yeah, this one’s for you.

First of all, I don’t see the gentleman asking the question because the city code mandated us to put some polls. It’s regulated by the way. Yes. First of all, I can talk of course on behalf of SAP, but we were highly involved also to the lady lawyer points there when there was a call for feedback, right?

It’s about being active and not waiting to drive an impact. I was trying to be funny finding many geeks like me talking to the government and others because we also have a brand of involvement. And I think it’s a call for action for everyone to wake up and help Gerard and the teams to regulate. It’s good for the industry, right?

And this is where definitely, I mean, if I understand the question correctly, there is always a spot to impact and involve on the regulation. But it’s even go before that. And I think it’s all start from how involved and how big of a championship you have for this kind of topic in the C-level by the CEO and so forth. Because I can talk about SAP coming already with opinion even way before the EU-AI Act came, even before the JDPR came in 2018, right?

We became an opinion and we were involved and it helped us to help shape the agenda. Did we get everything we want? No, but it’s okay. But you can come and be proactive.

And I think the only message I can give here is that the leaders, and we have many leaders in the room, need always to have opinion also about, I mean, the future of the business. Because the next generation is right across the corner, right? We talked a little bit about Quantum and there was the blockchain and others. People need to think about it and encourage the critical thinking, but also how we can help and impact the industry.

So, and I think it also benefits a lot of companies because eventually you create this trust and you create this kind of engagement that helps you to sell more to your point and impact. Yeah, and that’s kind of the short end from my side. And yeah, I’ll stop here because there are many points to say about what was said earlier, but you know, I kept the babysitter today so we need to finish it somewhere. I hope I addressed the question.

I don’t know if anyone else wants to double down on that. I have a question for you all. Thank you so much for being here today. I want to think back on the opening keynote and you mentioned that one of our greatest issues is the need for trust in order for people to use AI and in order for it to really yield the success that everyone in the industry is talking about.

My question is whether or not the passing of this EU AI Act actually increased trust in AI in Europe for the general public. I know it’s hard to generalize for various people in countries of Europe, but after the passing of this policy, have you seen tangible increase in people trusting the AI products on the market and showing interest to start using these products? Great question. If you go first, I can contradict you rather than the other way around.

I think it’s very strategic. I obviously can’t talk about your question, but I will add to it. If you go to the European website, you have all those banners about accepting the use of data. You need to click one, two, three, four times when you need it.

Has that really improved the data usage for your behavior? I’m sure about it. Hugo. That’s the famous AI kind of law, etc.

It’s too old It’s a regulation! It’s too old a chestnut to start unpacking. I mean, if you ask people do you trust AI, more people in the US do not trust AI than in the EU. That’s an interesting fact.

The second is, if you ask people do you think AI should be regulated, a large majority in this country says yes, AI should be regulated. They don’t trust it. Regulation in itself is not enough, but we think it’s necessary. Again, the question is would you step into a plane if it wasn’t regulated, if there was no certification.

It’s almost we take this for granted. We have internalized that things are safe, but they’re not safe because everybody wants them to be safe. They’re safe because there are rules in place and these rules are enforced. What is absolutely necessary as well, and we talk a lot about AI, but we haven’t seen much of the benefits of AI yet.

It is still very abstract, very fuzzy in a way. This is where hopefully the AI factory will help. If people see that there are applications that actually give, like in healthcare, a greater degree of accuracy for breast cancer and treatments and everything else, people will want to use AI because they see it’s giving them a better quality of life. Maybe it’s saving their lives in some ways.

Here, people coming from Europe, we don’t have the self-driving cars in Europe yet, so we have the Waymo. One of the things we do is when we get these visitors, we stuff them into a Waymo. We had the high representative, the Secretary Blinken of the European Union, who came here a few months ago and said, no, you get the Waymo right. I’m not going to sit into a self-driving car.

I drive myself. I don’t trust that self-driving car. I don’t know how we persuaded him, but he found himself into a Waymo car and then 20 minutes later he came out and said, I’m still alive. It works.

And now he goes around in Europe and says, oh, these self-driving cars really look like San Francisco, all of these steep uphill roads and dense traffic. And it worked. So people will need to see that it works and it brings benefits to them. And then you will see the tide turning.

Then people see the 50% will become 30%, 20% until it’s a negligible part of the population. So don’t look at rules as a competitive advantage. As an industry, don’t always look at rules as something that’s going to hinder you, that’s going to put burdens on you. How can you use it to your advantage?

Because what you’re selling, SAP and many other companies, is trust. And secondly, kind of show the applications, make sure that people see the benefits of AI in their daily lives. And I think that combination will really get us onto AI in the world since OpenAI came on the scene. AI exists since 1936.

And to say we have not seen benefits of AI is gross misrepresentation. Like we do AI in healthcare, we do it in media, Facebook is basically a lot of AI. We do AI in image processing. If you enter the US, you get scanned by AI.

I mean, AI is by now everywhere, not only in a car. But it’s everywhere. And it’s not just in the car. Just saying like, OK, chat GPT, then we can chat to AI.

For the rest, it’s actually down low. Not to mention 50% of the code on GitHub right now is artificially generated. So a lot of growth there. Yes, please.

Thank you. My name is Ana Ruiz Villarreal. I’m from Costa Rica. I studied almost 13 countries in Latin America.

And all of our countries follow like a copy-paste of the European Union docket. And it’s like a copy-paste. Like that. We don’t have kind of agreement in between countries to create the ethical values or principles.

It’s like copy-paste. But now, my main concern about it is how to do it in a way in which we respect our culture, our content. And for you, Mr. Gerard, I’m sorry, that’s a really Spanish way to pronounce your name.

That’s actually it. What not to copy. What not to do. And if you can speak about the process and the content of the law.

What not to copy and paste in our countries. Thinking about the Latin American context, we don’t have this kind of infrastructure. We don’t have the kind of many, many professionals who knows about it. And we don’t have human infrastructure for that kind of knowledge.

Right now. We are building something. But right now, we don’t have that. Thank you.

We’re going to give a two minute answer and then we’re going to take 30 seconds for each speaker to quote on the website. If you’re a Spanish speaker, my name is… You pronounce it absolutely accurately. Because you all have this same Dutch language.

On this point, I don’t think culture is necessary. We’re not regulating culture here. I don’t think the cultural dimension is so important. But it is the capacity that is really important.

These laws are not easy laws. More than 300 pages. You need to have administrative capacity. Don’t underestimate also the challenges for the regulators to implement.

We’re building up the AI office at the moment. We did it the same for the DSA and the DMA. I think what countries need to… I mean, they should not bite off more than they can chew.

The European Union is probably one of the few jurisdictions, if not the only jurisdiction in the world, that can actually implement and enforce an AI act. Or the Digital Services Act. Or the Digital Markets Act. For many other countries, this is still a bridge too far.

And there maybe you should start. Maybe you do a bit of like starting with transparency and then you move a bit up to like more substantive obligations. But to take the EU AI Act when clearly you don’t have the administrative capacity to implement it is probably not a good idea. So we always…

the zombie community wants to be valuable to the guests and so generous with your time. So let’s take 30-40 seconds each to maybe plug a website or how folks can get in touch with you. Jan, please, please start. Yeah, first of all…

First of all, just a word. Thank you for this discussion. I mean, I learned a lot and it was very engaging. I checked all the people here.

I don’t think I saw anyone checking their phone, which is also a good… I mean, I think one thing I can say, SAP released the AI ethics white paper recently. I know there are many AI ethics white paper out there. It was released last month.

Really, I’m not saying this because I work for SAP, a big leadership. It takes everything we heard here. The EU AI Act, the AI pact, the readiness for this. Different perspective and I think this is the best optimization of what we heard today.

In my opinion, industry boys want to sell, but a lot of the European value regulation, they want to sell. So, I think we can start with what I said earlier, where is the high risk, where is the low risk, and such and such. So, this is my Google for AI ethics policies and we can continue from there. So, I started by mentioning this report that we just released a few weeks ago.

So, you can find this on the Stanford Cyber Policy Center website. It’s quite long, but I hope it’s comprehensive and it provides a lot of key of the AI Act among others. Thank you. Thank you, Nick, for an amazing moderation.

Please check out oversightboard.com for our latest report about content moderation and AI, where we encourage meta to be transparent and to put the human back in the loop anyway. And also check out Internet Without Borders, where we defend for expression, privacy, and we also help civil society use AI, take advantage of AI in a manner that does not harm human rights. I have a microphone. I think this was fun.

I mean, if you come out in the evening, there’s so many other nice things you can do in San Francisco. You come here and you listen. I mean, I think panels are much more interesting when people disagree than we will all agree. So, we had our disagreements, we’ll drink a glass afterwards, but I enjoyed it.

I mean, one thing is to continue to engage in these issues, because this is a very important issue. And I think this affects all of us. So, don’t defer these, like, you have interests, you have views, come forward. The European Commission, people often think it’s like faceless bureaucracy, it’s very difficult to interact with the European Commission.

Not at all. It’s one of the most open administrations in the world. And that, so, I would recommend to kind of continue to, because you can shape this. We can collectively shape this.

You don’t want it to be shaped just by the industry or just by governments or just by kind of other stakeholders. We need to shape this collectively. Part of it, but you need to continue those efforts. I totally agree with you.

For the first time, it’s hurt! For the wine! I’m going to have a note. So, I told you guys that I’m a startup founder, so I recently founded my own company.

It’s called R2Decide. We believe that all those chatbots, they are way too spammy, they talk too much, they don’t help you really. We integrate, essentially, a better search engine. And we have a lot of people that are very, very specific about their research.

So, if you search for something very, very specific, something like cushioned running shoes, normally those e-commerce sites, they don’t give you those real context-aware search. We changed this, we built something, we didn’t apply for e-radiation for this. We got some people in the garage, we’re working hard, we can’t do that. But, it’s working, it’s amazing.

If you know e-commerce companies, there’s a lot of people out there, and would like to test it out, and if you can get them to send it to traffic, I will make your search in sexy. If you are somebody who wants to support us, join us, in terms of an angel, you know how to find him, and then you find me. Or, I’m a senior singer, there are not so many other singers in the internet, so, all the rest. What’s the two-round model?

The one thing is that I feel like something that I’ve always directed staff so check out stilltonguehobbies.com. For more, if you press S brains, and if you go to X signs, take care of yourselves, and take care of each other. Thank you. Bye.