Anirudh Suri

Full Transcript

Switzerland 2027: Is a Global AI Treaty Possible? Amb. Thomas Schneider

TRANSCRIPT

Anirudh Suri: What does AI for good mean?

Thomas Schneider: You need to have a societal contract, nationally and internationally, about what is good and what is bad. But it’s also in the consumer’s hands to buy products that are safe and sustainable, and not just go for the cheapest solution. So we cannot just blame the ones in power, because through buying, through voting, each of us has a responsibility.

Anirudh Suri: Today I’m delighted to have with us Ambassador Thomas Schneider from Switzerland, chair of the Council of Europe’s Committee on Artificial Intelligence. He ended up negotiating the first binding international treaty on AI, and he’s the person responsible for organizing the AI for Good summit planned for Geneva in 2027. How is Switzerland positioning itself and thinking about its national AI strategy?

Thomas Schneider: We are an open society with an agile and innovative economy and research ecosystem. If the Swiss label is on something, that stands for a certain quality, a certain reliability. It’s important that countries have a choice among different suppliers, that they don’t depend on just one or two. It’s also good that a country like India, which has made enormous progress in recent years with the Indian digital stack, can lead by example, so that other countries are inspired.

Anirudh Suri: Do you expect global governance of AI to be easy, difficult, or impossible?

Thomas Schneider: It may be difficult to agree worldwide on any one thing. We will not have a single global treaty covering everything about AI in the coming years, that’s probably not realistic. But even when states are competing, they have an interest in guaranteeing each other that the technology they have doesn’t fall into hands that are completely uncontrollable.

Anirudh Suri: In the industrial revolution, you picked fossil fuels as the primary driver of your industries, and as a result the earth and the climate are suffering. How do you think we might find a balance this time, and not repeat the same mistakes?

Thomas Schneider: Until it’s possible to go to Mars or live on another planet, we just have this one planet, and I hope that people will strike this balance better than we may have done with the industrial revolution.

Introduction

Anirudh Suri: Ambassador Schneider, welcome.

Thomas Schneider: Thank you very much, it’s a pleasure to be with you.

Anirudh Suri: I want to start with a question specifically about Switzerland before we move to the broader set of questions around AI and global governance, which I know you’ve spent many years working on. Switzerland has of course been at the center of global governance in many fields, not just AI, home to many global governance organizations, but also a country that has positioned itself as a large industrial power, with major corporations based there in the traditional industrial world. So my first question is: how is Switzerland positioning itself and thinking about its approach to AI, its national strategy? What are some of the key pillars of that approach?

Thomas Schneider: Thank you for the question. Switzerland is a rather small country in the middle of Europe, about 9 million people. We’re an open society with an agile and innovative economy and research ecosystem. For us, it’s very important to have a rules-based national and global order, where everyone can freely develop, and as long as you respect some basic rules and basic values, you can use new technologies to try to do some good, to create innovative business models. This is at the core of our approach internationally, and it’s also why, nationally, we try to build a societal consensus about the limits, the framework within which this freedom to act and innovate can be exercised. Our principle is normally: as much freedom, and as little regulation as necessary, but as much regulation as necessary, so that no one is left behind and negative impacts are minimized. This is the way our participatory political system functions nationally, and it’s shaped our approach internationally too, we try to talk and listen to everyone, and find solutions acceptable to all. Within that framework, people and industry can then compete with each other.

Our national AI strategy follows the same lines. The basic framework, the bottom line, is the convention I mentioned from the Council of Europe, which tries to ensure existing protections for human rights, the rule of law, and democracy apply to AI systems the same way they apply to any other technology. Within this, countries are free to develop their own approach to governance, to foster innovation, to allow progress. There’s no need to create a whole new set of rules, we want the existing rules, the existing levels of protection, to apply to AI too. That means whether a decision is made by human beings working with each other, or by a machine, if something is wrong, any person should have the right to challenge that decision, whether it’s about their social security, their healthcare, or anything else. People have the right to know what the machine is doing, and if they don’t agree, they can take that decision to a court. So there are certain necessities around documentation, so that people can live an autonomous life while using these technologies.

Anirudh Suri: So I think I’ve understood you correctly, on that spectrum between innovation and regulation, you’re leaning closer to innovation, with only the necessary regulation, so that existing individual rights are protected and societal goals are met, without hindering innovation. Every country today is grappling with that balance, trying to find its own sweet spot, some closer to the innovation extreme, some higher on regulation. But from a competitiveness standpoint, one of the things we’re noticing around the world is that two countries, the US and China, have taken the lead in AI. Switzerland ranks very highly on many metrics of R&D and innovation. How are industries in Switzerland, companies, corporates, positioning themselves to make sure they’re not left behind? I wrote a book called The Great Tech Game, and one of the key lessons I drew from history is that as the games around competitiveness change, and new technological eras come around, countries or industries that keep playing the previous game and don’t reposition themselves often get relegated to the annals of history. So how are Swiss industry and government repositioning themselves to protect Switzerland’s competitiveness in a world where two countries seem to have taken the lead, spending the most R&D dollars, attracting the best research talent, and allocating far more capital and critical resources?

Thomas Schneider: Thank you for the question. Every country, every economy, every society is challenged by new developments, but that’s the way life goes. We already had the first industrial revolution, where Switzerland was way behind our neighboring countries when it came to building railway lines, they were already discussing building lines around us because we were so far behind. Then, in the middle of the 19th century, some people in Switzerland took action. They didn’t just build railway lines, they created the famous ETH university, which is now among the top ten globally, and they also created a banking system, because there was no bank big enough to give loans to those building the railways. A whole ecosystem was created fairly quickly, we caught up, and now we’re known as “the railway country,” one of the most dense and efficient railway systems in the world.

There are some similarities now. There’s a lot of dynamism, a lot of decisions to be taken, and you’re not always sure which is the right one. I think we’re trying to remain agile and dynamic, open to innovation, open to take some risks. On regulation, we’re a country that’s never been ruled by a king, always a bottom-up, participatory democracy, and people want the right to experiment a little and see how it goes, what the real dangers are, but as soon as risks turn into negative impacts, they’re also quick to ask for regulation. So I think our hope is that we’ll manage, this time too, to be innovative, to experiment ideally on safe ground, and produce good solutions that work, while keeping the trust in the quality of our services, the trust in the quality of our political system and administration. Nobody’s perfect, we also make mistakes, but I think one of Switzerland’s assets is that we’re a small country, we can’t do everything better than everybody else, but we can do a few things especially well, and concentrate on those. If the Swiss label is on it, that stands for a certain quality, a certain reliability, there may be a price to it, but it’s worth the price.

That’s the logic we’re trying to pursue in this new environment too. You referred to the fact that there’s a lot of front-end research in Switzerland, a little less when it comes to scaling up big investments, Swiss inventions are often bought by bigger companies or go abroad. That’s partly just part of life if you’re a small country. But agility, and the courage to take calculated risks, while always giving people the right to say “stop, we don’t want to go further,” that’s the core of our spirit, and we’re trying to live that same spirit with this new technology too.

Finding Switzerland’s Niche in AI

Anirudh Suri: I like the historical context you’ve given, of Switzerland being a bit behind at one point in the industrial revolution and then finding its own niche to catch up, whether through the railways, which by the way have featured very extensively in Indian movies and Bollywood, or through banking, and on the industrial side, pharma and healthcare, and of course watches and chocolates, industries where Switzerland built that quality guarantee you mentioned. What could that specialization or niche look like in the AI world? Is there already emerging thinking in Swiss policy circles, whether in the private or public sector, about what that equivalent might be?

Thomas Schneider: Yes, thank you. One interesting detail is that Switzerland is actually one of the few countries that doesn’t have a specific AI strategy as a political strategy. We have lots of digital strategies underlying it, and a very distributed system, not run by a central entity that controls the whole country. Again, we try, with the economy, the research and academic ecosystem, and the political ecosystem, to let intelligence not sit in the center but be distributed at the edges, and whatever is good gets picked up by others.

So, for instance, in finance, people are looking at new applications in fintech, insurers are used to dealing with risk, they’ve used algorithms for a long time, they’re used to using data, they know it’s an asset, and they need to stay at the forefront. The pharma industry, another big factor, knows that to produce new medicine, to support healthcare well with AI, you need data, a certain mass of data, but definitely a certain quality of data, because in the end this is about people’s lives. If you have bad or biased data, say data that’s only about men when you’re trying to help women, that makes no sense, you won’t get the right results. So there too, data quality, access to data while protecting people’s privacy and rights, but also making sure people share their data in a way that lets them benefit from it themselves.

We have a number of threads we’re pursuing to build this kind of ecosystem. It’s not just about having algorithms, it’s about having data, rules about how to use data, energy, infrastructure, fiber to the home, enough bandwidth, and the skills of the workforce and the administration. It’s not just one element, you may buy algorithms, but if you don’t have data, if you don’t have storage capacity, none of it works. So we try to see the big picture, and advance everything to the extent necessary, so we can build on it step by step, with good, serious risk and impact assessment mechanisms, which is something else the Council of Europe developed alongside the treaty, a legal instrument plus a risk and impact assessment mechanism for human rights, democracy, and the rule of law, because in the end people care about a functioning political system, freedom of expression, and the impact on things like social security. So we try to advance on all fronts, so the ecosystem works out, and if something goes wrong, we can realize it and correct it, together with the people, together with industry, to arrive at a reasonable way of using these technologies while minimizing harm to anyone.

Anirudh Suri: That makes sense, focusing on the foundational pieces and layers, and then hoping the magic happens on top.

Thomas Schneider: And one thing I should add, as I said, we’re a population of 9 million people, growing every year, but we won’t become as big as India for the next million years, I’d guess. So we need to keep partnering with others. We have our European neighbors, who are very close to us, but also a lot of economic and other cooperation with India and other emerging economies. It’s necessary to partner with each other in a peaceful, respectful way, to create win-win situations, not just for us but for others too. I think that’s important in a world as complex as this one, with so many challenges, we need to respectfully partner with others, otherwise no one gets anywhere in the end, because we’re sharing this planet with many, many other people.

Global Governance of AI

Anirudh Suri: I think that’s absolutely right, and this idea of AI for good, AI for all, an inclusive, sustainable AI framework that manages risk while enabling innovation, is ultimately the baseline many countries around the world, including India and many others in the Global South, want to embark on. But in an increasingly geopoliticized world, one characterized by fragmentation, by fraying alliances and friendships, whether transatlantic or elsewhere, where the rules-based order the world has lived through for decades seems to be under attack and in a state of churn, and where AI and emerging technologies have come to the core of that competition, as I argue in my book, do you expect global governance of AI to be easy, difficult, or impossible?

Thomas Schneider: Things are impossible as long as you don’t try to make them happen. I negotiated this treaty, but I’m not a lawyer, I’m an economist and a historian, and we’ve had more or less troubling times in the past too. There are reasons why Switzerland is politically neutral, that comes from the history of societies fighting around us in Europe. When I say neutral, I don’t mean neutral in values, but neutral in armed conflict, an important distinction. I think we have to remember what happened when people believed the strongest one should win, because many people who thought they were the strongest turned out not to be, and if people go to war with each other, in the end no one wins, everyone loses.

So for us it’s important to remember that if we work together respectfully, with rules we all determine together, that no one imposes on others, we’re all better off than if we work against each other. If we define rules within which we can compete, then of course those working harder, those better organized, will get a bit more, but we’re convinced that, difficult as this period is, with more instability and insecurity, that shouldn’t derail us from continuing to talk to each other. How do we want to treat each other? How do we want to live together? How do we want to compete? Sometimes these voices are heard more, sometimes less, and my historical recollection is that when people forget that working together is the better solution, that normally leads to catastrophes, situations where, at the latest, people realize they should start working together, and end up better off. That’s a bit of a cycle with human beings, we tend to forget the past. But we’re strongly convinced that talking to each other, listening, respectfully competing, and making sure no one is left behind, will in the long run be better for all of us. Some people may see it differently in the short term, but we hope this vision will be shared more seriously again, globally.

Anirudh Suri: That’s the right lesson from history, that working together has led to more progress, even if differentiated, whereas conflict, one-upmanship, or trying to concentrate too much power or wealth in one nation or group, often leads to much greater backlash and social and political upheaval, going back a hundred years, World War I, World War II, the Cold War, all proofs from history that non-cooperation doesn’t end well. So let me ask you this: this idea of global governance of AI, much like cyber and other emerging fields, while in the long run you hope for cooperation and better sense to prevail, in the short term, actors try to act in their own national interest as much as possible, avoiding constraints, especially the countries at the leading edge of these technologies. In a world like that, do you see two scenarios: one where you negotiate a global treaty bringing the majority of countries together, but the treaty itself becomes the lowest common denominator of what everyone can agree to, or a scenario where a smaller grouping, regional or like-minded, agrees to much more, but without everyone at the table. Which do you think we’re more likely to see in the coming years?

Thomas Schneider: As you say, it may be difficult to agree worldwide on something. A global treaty in itself is just a piece of paper, if you don’t have mechanisms to enforce it, then it’s difficult, and this is what’s happened with some treaties of some organizations, if there’s no willingness to abide by the rules you’ve set yourself, this becomes a constant discussion, recalling the reasons why you engaged in the treaty in the first place. So we will not have a global treaty covering everything about AI in the coming years, that’s probably not realistic. The instrument the Council of Europe produced is a treaty that isn’t built on the European legal system specifically, but is designed for all countries that value democracy, human rights, and the rule of law, at least to some degree, so they can work together agreeing on principles they can implement within their own culture and legal system, in the way they want. It gives a lot of freedom in the “how,” as long as you agree on the “what,” and the “what” is applying the same level of protection you already have, which differs by country, but at least applying that same level to AI.

We already had 55 countries part of the negotiations, including 10 from Latin America, some from Asia, and now Brazil has joined the implementation process too, along with many countries from Africa. We shouldn’t forget, even when states are competing and maybe don’t trust each other, we saw this in the Cold War too, they have an interest in guaranteeing each other some security, that the technology they have doesn’t fall into hands that are completely uncontrollable, or willing to sacrifice themselves for, say, religious grounds. There are shared security interests even among so-called rival poles, and there’s economic cooperation between these countries too. So there’s always some interest, it’s a combination, never black or white, always a mix of competition, mistrust, and shared interest.

I think that’s one of the strengths of Switzerland and the Geneva ecosystem, we try to provide an environment where anyone can talk to anyone, in the hope that this creates mutual trust and fosters cooperation, and we’re willing to continue offering that space, with all the experience we have as a small player, to foster dialogue. It’s never black or white, and we should always bring up the arguments for cooperation, for giving ourselves rules that we define jointly, ideally not just governments but also industry, academia, and the people, because in the end people have needs too. People want a certain level of protection, and depending on the democratic system, it’s easier or harder to request that, but a government can’t just do whatever it wants, at some point people will say “we want this” or “we don’t want this,” and there are different ways that plays out. In the end, we’re sitting on the same planet, and we need to cooperate with each other. Sometimes it’s easier, sometimes harder, and there’s no one-size-fits-all solution. We may end up with hundreds of treaties, hundreds of legal and technical standards on specific AI applications, like we have on cars, airplanes, or standard software, because there’s a need for security and reliability at a technical level too. We cannot not cooperate, but it won’t be black or white.

Anirudh Suri: I think that’s really the space for diplomacy, where things seem difficult or impossible, but you still figure out ways to cooperate and ensure security, prosperity, and progress for the countries coming together. You mentioned earlier that you chaired the Council of Europe’s Committee on Artificial Intelligence, where you negotiated the first binding international treaty on AI. Can you reflect on some of the real practical challenges you faced getting countries to sign up to that binding treaty, and what practical lessons you’d offer from that experience?

Thomas Schneider: The Council of Europe is basically the institution created in Europe after the Second World War, to try to get European countries to find solutions to problems together instead of going to war, and fairness of solutions is important, sitting at the table and talking to each other. It’s like the UN for Europe. There was a big discussion at the start: should we do a treaty just for ourselves, given we have high standards, a court where, as a citizen, I can sue my own country if I think my rights aren’t respected? That’s unique, other regions don’t have that instrument, though they may have others. I was criticized, as chair, for a decision that wasn’t actually mine, it was the Committee of Ministers’ decision, that they didn’t want an overly strong solution just for Europe that would then be ignored by the rest of the world, they wanted something with a chance of becoming a global instrument. That meant we couldn’t just depart from the European institutional logic, we had to be open, find solutions and implementation approaches that work for the American system, which is different in many ways, for Latin American countries, for Asian countries. That doesn’t mean going for the lowest common denominator, but going more on the basis of shared values, rather than relying on specific instruments you’re proud of but that don’t exist in other countries.

That was one of the big challenges: how general do we go, risking being too general to have real impact, or how specific do we go, risking excluding the rest of the world from adhering to the instrument? That was a very big discussion, within Europe and with other countries. In the end I think we found a good solution, that says: no matter what your rights are in your country, the treaty makes sure those rights, which may differ from country to country, apply to AI too. That’s less than some activists were hoping for, but it’s something that essentially prolongs the line from the pre-AI era into the AI era.

It was a very hard negotiation, particularly between Washington and Brussels. Each side had its own national narrative, people in Brussels under pressure to get the US to adapt more closely to the European system, while the US was under pressure not to become “European.” It was a very interesting, very intense experience. The key thing I learned is the word interoperability. We’re not the same, we’re human beings, but our life circumstances, our histories, are different, our trust in governments and power differs depending on history. Some countries were ruled by kings and overthrew them, others, like mine, never had a king, we’ve always been a bottom-up society, wanting a government with as little power as possible, and people wanting as much say as possible. We have different instruments. We’re also not a member of the European Union, unlike all our neighboring countries, so we have slightly different ways of achieving the same goals. As long as we achieve the same goals, fostering innovation, experimenting with new things, within some limits on protecting rights and the planet, and as long as we’re interoperable in how we do things, they don’t have to be identical, that’s an illusion, but in a way that lets us work together, share experience, buy each other’s products, use each other’s services. Interoperability has become a bit of a buzzword, but it really makes sense, because harmonization across the world simply isn’t possible.

Anirudh Suri: I think that’s key, interoperability lets you keep your options wide and broad, not locked into one system, one set of technologies, or one set of companies. For any country looking at the world today, interoperability and strategic autonomy are key objectives, both for the country and for consumers, so they’re not at the receiving end of a lack of choice, exploitation, or a lack of innovation tomorrow.

AI for Good

Anirudh Suri: The other piece I want to shift to now is the idea of AI for good. We’ve discussed AI for all quite a bit, but I know you’re also involved in the AI for Good summit coming up in July 2026. Talk to us about what that means. Help us unpack what AI for good means, from Switzerland’s perspective and the world’s, especially at a time when AI is getting geopoliticized and weaponized, and the agenda for how it evolves is being determined in only a few places, by only a few people.

Thomas Schneider: Yes, thank you. The AI for Good summit is an event the ITU, the International Telecommunication Union, based in Geneva, started in 2017, in partnership with more than 30 other UN and non-UN institutions, universities, and others. They started with 200 people in 2017, last year they had around 11,000 participants, it’s more than doubling every year, we don’t yet know how many there will be this year. Switzerland is a co-host, but it’s about 40 institutions working together.

AI for good means that, with any technology, you can do something good or bad with it, like a hammer, you can use it to build a house, or to hurt someone, which is very bad. So you need an idea of what you should do with it. We know there’s huge potential to use data and AI to make our healthcare systems better, to have more targeted education for people with special needs, to optimize traffic systems, to optimize our use of scarce resources like energy. There’s huge potential everywhere with AI, but it can also be used for the opposite, it can be abused. And I think what you said earlier is crucial: it’s important that countries and people have a choice, that they can use services from different suppliers, that they’re not dependent on just one or two. It’s also good that a country like India, which has made enormous progress in the past years with the Indian digital stack, becomes a player offering services and good practices that can lead by example, so other actors, other countries, are inspired, and can also choose, “this service is good for me,” or “this fits our society, we’ll take this one and not that one.”

Of course, the question is: who defines what is good and bad, and good and bad for whom? The ITU and the UN system build AI for good on the seventeen Sustainable Development Goals, which range from protecting the environment to societal goals like education, skills, and human rights, and the slogan is “leave no one behind.” The idea is that if you use AI for good, every single person in the world can benefit, and no one is left behind. The AI for Good summit is basically a gathering to share these best practices from around the world, from industry, academia, government, civil society, so we can learn from each other how to use AI for good. It’s not necessarily a political event, but of course sharing good ideas has political power, because more people start using them.

But, again, like a car, you can use it to drive someone to the hospital and save a life, or you can drive it into a crowd, as has happened around the world several times in recent years. You can’t put a fence around every gathering of people, and normally it’s not the case that a car drives into people every day. So you may have laws, but you also need a societal contract, nationally and internationally, about what is good and what is bad. And even without a legal instrument for everything, people need to feel responsible for themselves as part of society, and behave, by their own will, in a way that says: treat your neighbor as you want to be treated by your neighbor. AI for good only works if you have shared values, and a discussion about what’s good and what’s bad. The background we still stick to is the internationally agreed Sustainable Development Goals, which are a very important guidance for us, and that’s why we support the ITU, UNESCO, and all the partners working together on the AI for Good summit, alongside similar events that also try to help people use AI for good.

AI and the Environment

Anirudh Suri: The other piece I want to touch on, which you mentioned briefly as part of the SDGs, is the environmental piece, the ecological impact of AI. The current approach followed by countries at the leading edge is building bigger and bigger data centers, bigger and bigger models, which of course comes with very energy-intensive requirements. Going back to history again, during the earlier industrial revolution, the ecological impact of energy choices wasn’t taken very seriously, 150 to 200 years ago, fossil fuels were picked as the primary driver of industry, growth was pursued as the objective, without much balance with the environment, and as a result the earth and the climate are suffering today. How do you, and Switzerland as a whole, think about this challenge, between the drive for AI growth and innovation on one hand, and the ecological impact of these technological choices on the other, impacts that sometimes take decades for humanity to fully realize? How do you think we might strike a balance this time, and not repeat the mistakes of the past?

Thomas Schneider: There’s a big discussion about to what extent humankind keeps making the same mistakes, but the situation is always slightly different each time. With the industrial revolution, it took time for people to realize some of the negative effects. In the early twentieth century, people were running behind cars because they liked the smell of the exhaust, they didn’t know it caused cancer, we’ve known that now for about seventy years, and we still haven’t fixed all the issues. It’s also easier to fix smaller, local issues, like car safety, we introduced seat belts and other measures, or local environmental issues, like when I was a kid in Switzerland we didn’t have water cleaning everywhere, there was too much phosphate, and there were lakes where swimming was forbidden fifty years ago. Now the water quality is so good you can drink water straight out of a Swiss lake, but that was a local problem, solved locally, because the people living at the lake agreed they wanted to be able to swim in it again, which is easier than solving a global problem like global warming.

I guess, to some extent, we first need enough research and some confident level of certainty about what the real problems and challenges are. We may still have people who deny this, we’ve just seen some recent decisions about climate change where you might question whether they’re really based on evidence-based policymaking, but that’s life. I think we need a coalition of the willing, to start with, to say, let’s go in with the best intentions, find the right balance between sustainability and thinking about the future of our children and grandchildren, while accepting we’ll have competing interests. But until it’s possible to go to Mars or live on another planet, we just have this one planet, and I hope people will strike that balance better than we may have done with the industrial revolution.

We also need to talk about technological solutions. Sometimes they can help, but normally technology creates new challenges while solving old ones. In the 1960s, people thought nuclear energy was the solution to all the problems caused by coal and oil, and then we found out it has its own dangers. But then again, with many things, if you have a safety net, if you have to pay for the risk you run, if a nuclear plant has to carry insurance for the potential damage it might cause, maybe fewer plants get built the way they were. Liability, I think, is a core issue. If there’s an agreement that whatever you do, you’re actually liable, if you cause damage to others you have to pay for it, that creates incentives that shape how industries and politicians behave, because if you’re liable, you think twice about the risk you’re running. But if you’re not liable for anything, you might think, “who cares, it’s not going to be me who pays.”

It’s about how we work with each other, how we set the rules. We have a chance to do better. We’re still human beings, capable of doing good things or bad things, usually somewhere in between. We think dialogue, honesty, respect, and evidence-based policymaking are the solution. I hope we can convince as many powerful people as possible, whether in politics or industry, to follow this road, but people will have to insist too, in whatever way they can, telling their governments, telling industry, it’s in the consumer’s hands to buy products that are safe and sustainable, and not just go for the cheapest solution. So we can’t just blame the ones in power, because through buying, through voting, we give power to the people, and each of us has a responsibility in the end.

Anirudh Suri: I couldn’t have put it better. Ultimately consumers, much like voters, have a way of shaping the choices of corporations and decision-makers who sometimes make decisions for the short term, taking the benefits for themselves while distributing the liabilities much more broadly.

The India AI Impact Summit

Anirudh Suri: Let me move now to the upcoming India AI Impact Summit. Ambassador Schneider, what are your expectations from this summit, especially as someone who may be leading the organization of the next one?

Thomas Schneider: First, we’re very thankful to our Indian friends for hosting this. It’s the first time this summit series, still young, is taking place in a G77 country. That’s very good, because the UK started this with a focus on safety, that was in 2023, then Korea, then France, each country brought its own particular focus that they thought important. It’s good to have a country like India presenting a focus from an Indian point of view, and we do have a lot in common, we may be a so-called “first world” country, but we face many of the same challenges. We need to make sure everyone is educated, that people know how to use AI, what the risks are, but also what the opportunities are, we don’t want people to deprive themselves of opportunities by not using AI. You need access to the technology.

We have four languages in our country, which isn’t a lot, but one of them is spoken by only about 30,000 people and doesn’t exist anywhere else in the world, it’s called Romansh, a kind of mountain-Latin language that developed across five countries, about 2,000 years old, still spoken by 30,000 people, which would be a very small town in India. We try to use AI to help this language survive. The Swiss universities have developed an LLM called Apertus, Latin for “open,” not meant to compete with ChatGPT or the big LLMs, but built as a blueprint that can be used by communities without billions of texts or huge computing power. It’s fully open source, with a curated dataset, not just anything found on the internet, we tried to go for quality rather than quantity. It was published last September or October, and has drawn a lot of interest, particularly from smaller communities, not the ten biggest languages in the world. We hope this creates real value, helping people cultivate their culture and language while using the tool in their own language.

That’s just one thing. I also co-chair one of the seven “chakras” for the Indian summit, together with colleagues from India and Nigeria, on empowerment for social inclusion, one of the seven key themes for India, alongside the others which are also important. We’re very happy to see that the idea that everyone should be able to benefit from AI, not suffer under it, is very strong in the Indian summit. I also love the three “sutras,” people, progress, planet, it’s simple, but it makes a lot of sense. I think it’s a very nice framing that our Indian friends have put into this summit. We hope to be the next host country, and we’ll build on this, saying, there are bigger cultures, bigger industries, and smaller ones, AI should be usable by every culture, every industry, every company. Let’s create an ecosystem that isn’t just for a few big players while the rest are ignored, there’s something in it for everyone. We’re very happy to see this focus of the Indian summit, and we fully support India in this.

Anirudh Suri: I love the example of the language apps, because in India too, Indian companies, startups, and researchers have thought hard about where in the broader AI ecosystem to focus their efforts, and one pillar has been the multilingual challenge, since the global models have focused on English and the more widely spoken languages, because that’s the data they’ve been trained on, that’s the data more available digitally. Many startups, and I’d point to Sarvam as one example, have taken the multilingual diversity of India and asked, how can we build for groups that might not be represented in the broader models? I think that’s a great area of common interest between countries like Switzerland and India, what does our research suggest, what does India’s research suggest, and where can we learn from each other to build AI for all and AI for good?

Rapid Fire

Anirudh Suri: This has been great. I want to enter our rapid fire segment now, Ambassador Schneider. We’ve had long, insightful references to history and democratic politics and how that plays into AI, in Switzerland’s case, ground-up democracy. Let’s now move into a short, brief, rapid-fire section. I’ll ask some questions that might not be as complicated, but let’s keep the answers brief.

Let’s start with some questions around students, a key audience for us on the podcast too. If you were talking to a fifteen-year-old high school student in Switzerland about AI, how would you describe the importance, significance, or challenges of AI to them?

Thomas Schneider: My kids are now twenty and twenty-two, so I’d say: use AI, but think about which skills you can delegate to it, and which skills you want to keep for yourself. If you want to be the master of your own life and future, there are some things you need to at least know how to do yourself, and not just let AI do everything without any idea of what it’s doing or what comes out. You need to be able to judge what’s coming out, and decide how to use it.

Anirudh Suri: If your children asked a follow-up: “Dad, which skills should I delegate to AI, and which should I focus on developing myself?”

Thomas Schneider: Trying to put it simply: keep the creative process to the human being. Having an LLM summarize 500 pages of scientific research into two pages, the machine is probably more efficient than you at that. But then considering whether something is missing, whether the machine got the balance right, whether you agree with the value judgment it’s making, and how to apply that to a new context, that’s where the creative element, the repetitive work can be delegated to the machine, like we do with production machines and cars, but a car still can’t jump a fence, a horse or a human is still better at that. Whatever is special, whatever needs intuition, a strength we have as human beings, a machine can only calculate probabilities and repeat what’s worked most often in the past. Maybe we’ll get there eventually, some people say we will, but we’re not there yet. Keep the creative, flesh-and-blood element for yourself, and keep asking whether something really makes sense, even if it’s written on paper or on a screen, does this make sense in my life, on this planet?

Anirudh Suri: So, staying critical, not accepting everything at face value. Sticking with your kids’ voice: will AI create more jobs than it displaces?

Thomas Schneider: That’s a very good question. As an economist and historian, so far, every new technology has created more new jobs than it destroyed, though it was challenging to help people whose jobs were disappearing learn new ones. But in the end, economically, it created more work. The question is, that’s the past. I wouldn’t predict anything, but if life goes on as it has, we could assume the same will happen with AI, I hope that’s the case, but nobody knows. What we do need to discuss, and some institutions like the Windfall Trust and others are looking at this, is what happens if, in less than five years, ten or twenty percent of jobs in offices, lawyers, interpreters, judges, secretaries, assistants, get replaced by AI. I’m not saying it will happen, but there’s a risk, and I don’t know how it plays out. So it’s important that we discuss this, develop some kind of alert signal, discuss it with industries, get people on track to know something’s moving here. I’m not making a prediction, it could be more or less smooth, but it will change things. Technology changes the world, and we need to prepare, no matter how good or bad it turns out to be.

Anirudh Suri: If your children were a bit younger, entering college, and asked, “should I study the humanities, or technology and engineering?”

Thomas Schneider: I personally was always interested in everything, and my kids are similar, but of course you can’t study everything. Whatever you do, try to keep the other side as a hobby. If you go into humanities, try to understand a bit about how technology works, and vice versa, if you go into business or research, think about what it means to be a human being. Try to have a broad horizon, and sometimes, almost by accident, you get into something you like and stay there. I don’t care what they study, as long as they stay human beings, open-minded, and hold onto their values.

Anirudh Suri: Let me move now from students to policymakers and countries. Are we looking at a potential second “great divergence” in the world?

Thomas Schneider: Everything is possible. I don’t believe the future is planned by somebody, whether a god or the nature of physical laws. I think we have the future in our hands, as a collective, we take decisions that have consequences, sometimes things happen and our reactions to them have consequences too. It’s in our hands whether we follow people who say “me first,” or whether we decide we can care about each other and try to make sure no one is left behind, normally it’s somewhere in the middle. I don’t believe in a deterministic future, people have choices, sometimes you need to fight for them, but you have them, they’re up to us, in which direction we go.

Anirudh Suri: If you’re a policymaker today, in the Global South, or even in a developed country that isn’t the US or China, would you prioritize AI innovation, or AI adoption and diffusion?

Thomas Schneider: I think normally it’s not either-or. To keep your autonomy or sovereignty, you also need to know how these things work, so you should be part of innovation. But not everyone can do everything, and especially if you’re a smaller country, you need to partner, it’s always good if you’re good at something and partner with someone good at something else, and together you’re good at everything. So it’s not either-or, we need to try to be good at as many things as we can. But with this race for faster, bigger, more, better, I think, coming back to us as human beings, we’re not necessarily happier the faster our lives are. There’s a saying, I read somewhere, from Africa: “Europeans have watches, but we have time.” Quality of life isn’t necessarily bigger, faster, more, it’s about how you use things, how you spend your time, how you treat each other. Of course you need enough to eat and a roof over your head, I’m not saying starving is a good perspective, and the race for bigger, higher, more, faster, louder is natural to some extent if you live in a competitive, capitalistic economy. But we shouldn’t forget aspects of happiness and satisfaction, and we should also tell our machines that they shouldn’t only optimize for bigger, faster, louder, more, but also discuss with them what quality of life means, and see whether they have new answers.

Anirudh Suri: Deep questions for machines to consider. Let’s play a quick word association game. I’ll name a country, and you tell me what word comes to mind in the context of AI, could be AGI, innovation, strengths, or capabilities. Switzerland?

Thomas Schneider: Agile.

Anirudh Suri: The United States of America?

Thomas Schneider: Big.

Anirudh Suri: China?

Thomas Schneider: Also big.

Anirudh Suri: Singapore?

Thomas Schneider: Interesting, dynamic.

Anirudh Suri: Nigeria?

Thomas Schneider: I don’t know enough, but I guess it’s interesting too.

Anirudh Suri: Brazil?

Thomas Schneider: In many ways close to European discussions, same, but different from Europe.

Anirudh Suri: The United Arab Emirates?

Thomas Schneider: Courageous.

Anirudh Suri: And lastly, India?

Thomas Schneider: Very exciting.

Anirudh Suri: Great. Now let me move to some personal questions. Which applications of AI are you most excited about?

Thomas Schneider: On one hand, all these large language models are exciting. I also think it’s exciting to use AI for art, to create images, knowing there are lots of discussions about manipulation, but in the end art is manipulation of reality, the question is with what intention and background. So AI and art, though I don’t do it myself, I work for the government. Anything that makes processes more efficient, uses fewer resources, cuts down on how much people or goods have to travel around, I think that’s exciting, using AI for good and for fun. What’s not exciting is using AI to damage or harm other people, whatever form that takes.

Anirudh Suri: Who are the ultimate decision-makers today driving the path of AI, companies, governments, or the people?

Thomas Schneider: All of them. The more capital and resources you have, the more immediate impact you may have, but in the long run, we don’t know who among those considered powerful today will still be around in ten or twenty years.

Anirudh Suri: Very diplomatically answered, Ambassador Schneider.

Thomas Schneider: I’ve been doing this job for more than twenty years, so I hope it’s had an effect. Personally, which application of AI do you use the most?

Actually, in my job, it’s translation software, since I work with different languages, and sometimes it’s just faster than consulting someone. That’s probably the one I use most. But then, of course, everything I use has algorithms in it too, my music streaming tool proposes things based on what I’ve listened to before, which is nice many times but sometimes boring, since it would be good if it proposed things I haven’t heard yet. So I’m using many algorithms in daily life, but interpretation software is core to the texts I write and the dialogues I have.

Anirudh Suri: Will AI make the job of diplomats easier or harder?

Thomas Schneider: It will make it faster. In the end, it’s human beings, and I don’t think it changes the quality. Some things will be easier and faster, other things may be more cumbersome, and technology in particular is a nightmare if it doesn’t work, if it doesn’t deliver. But as long as we’re dealing with human beings, it’s the human beings that make our lives worthwhile in the end, technology is a tool, one we should use to the extent that it helps us.

Anirudh Suri: And finally, to conclude, two quick questions: which commentator, thinker, or leader in the world of AI has influenced your thinking the most?

Thomas Schneider: There are several, but the one who impressed me first, almost twenty years ago, was an Austrian professor, I believe teaching in Oxford or somewhere in the UK, whose name I recall as Viktor Mayer-Schönberger, back when we had new things like the internet of things and what was then called “big data.” One sentence of his stuck with me: while we’ll be able to automate a lot of repetitive things, and everything will be much more efficient, we may have to fight for autonomy, in the sense that because it’s so efficient to do the same things every day, we may have to fight for our freedom to do something differently today than we did the day before. That phrase struck me, we didn’t discuss AI directly, but I think it’s one of the fundamental questions: how do we remain autonomous human beings, autonomous societies, free within the limits of respect for others, to lead our lives the way we want? That’s something I keep in the back of my head.

Anirudh Suri: To add to that analogy, the saying you mentioned, about watches in Europe versus time in Africa, I’ve heard others say that ever since the invention of the clock, lives have become more clockwork, more mechanical, which connects to the point you were making earlier about quality of life and slowness. Great point. And lastly, books you’d recommend our listeners read, from any field, given you’re an economist and historian by training, and second, podcast guests you’d recommend.

Thomas Schneider: To be honest, I don’t have that much time to read, and when I do, I tend to forget the titles. But there’s one book, an economics book by a French author named Thomas Piketty, that impressed me, in German the title translates to something like “the history of equality.” He shows that even within the same country, inequality changes over time, and while you can have capitalistic systems with competition, it’s not a given that the rich always get richer and the poor always get poorer, a society can take decisions that make it more equal or less equal. For instance, the history of Sweden, where inequality was once very strong, and at some point they voted for politicians who brought society closer together. That’s a book I’d recommend, it’s also about how a society should keep discussing how we want to live together, what fair competition looks like, so hard workers can get ahead, but also what the minimum level of equality and solidarity we want is, because I might have an accident tomorrow and need your solidarity, no matter how rich I am. We can shape this together, it’s not the politicians who decide alone, it’s us, who elect politicians, and who go out on the streets if we think they’re not doing the right thing. We are the masses, we have power as consumers and as voters.

Anirudh Suri: Very important to keep in mind. Thank you so much, this has been a great conversation. I think you bring a certain level of practicality, combined with a long-term eye on humanity as the fundamental thing that brings us all together, which is rarer now than it should be. I thank you for bringing that lens to our conversation. Time and again, no matter what question we were discussing, you brought a very practical understanding of the world and its realities, the pragmatism of someone who has clearly negotiated treaties and brought countries together, while focusing on the humanitarian piece and the long term, rather than the short-term gains humanity sometimes gets stuck chasing. Thank you for doing that today, and I look forward to maybe coming to visit you in Switzerland for the next one.

Thomas Schneider: Thank you very much, the pleasure is mine.

Anirudh Suri: Thank you so much, Ambassador Schneider.