Uncategorized

We are going to start with a very special guest. Minister Bawu is going to join us for a keynote. Very excited to hear what you have to say, coming from Hyderabad, one of the centers of technology in India and in the world. So, Minister, thank you so much for joining us. And if I could ask you to come to the podium. Thank you so much, Minister.

Very good afternoon to all. In fact, we welcome you to our city of Delhi, a beautiful city, a capital of India. And many people are from India, too. And we welcome the distinguished panelists. eminent panelists who are sitting here to sit and discuss the quotes for Better Tomorrow. And I welcome the leaders of the industry and the delegates over here. And especially coming to the subject, AI agents for Better Tomorrow. You know, I wish to see that, you know, where we stand today. And where we would end up tomorrow. The point of discussion over here. We stand today at a fundamental inflection point in the history of governance. As a policymaker, I would like to mention a few points.

Because all the technocrats are sitting on all the eminent, you know, scientists maybe from physics or the maths may be sitting on the other side. To develop AI into next level. You know, for decades, the digital revolution in the government was defined by transition from paper to portals and from physics cues to digital clicks. But today, we are witnessing the birth of the new paradigm. We are moving beyond generative AI that simply answers. We are moving from them to agentic AI that acts now. What I’ve been discussing with Mr. Srinivas just now. And for 30 years, our relationship with technology was a series of commands. We used to give commands and used to get the answers.

We typed, we clicked, we prompted. We were the masters for the such bar. We used to, you know, we were the masters. Nobody can say that. But I stand here. I stand here today. I can see and everybody can see the search bar is dying. In its place, something more profound. Just now Mrs. Sweeney was just telling about agency. It’s just evolving. In the first era of our national building was defined by land. The second by the industry. And third has been defined. More illusory, the intelligence of the system. And the nation that leads this century are those that learn to treat intelligence not as a product but as a form of a public infrastructure.

The idea is no philosophical for our state of Telangana. It is the story of our everyday governance because it’s IT driven state as we are known for. And often say that artificial intelligence has three lives in the country. the first life is in the research labs the second we take into in the policy papers but the third ultimately both of this combined together how we are trying to affect the life that truly matters for each and everybody you know how do we see it is that when ai meets the real challenges of of our lives when artificial intelligence meets the dust where we face ai meets the drought when it meets the monsoons when it meets the markets of the living society and this is where its legitimacy is earned when it really counters this dust doubt monsoons and markets in telangana we see agents not as a tool here we would like to take them as a team mates.

You know as the way the pilots relay and the co -pilots. Tomorrow as our government here in Telangana also see that we relay on AI as co -governors system that can predict a flood before the first cloud gathers over the Moosey. Moosey is our river in the midst of our city. You know allocate resources before the crisis and deliver services before citizen ever need to ask. For example if you take the agriculture a small farmer I hail from a very remote area and that to a rural place. A farmer in my place or in some other place from the rural area when the climate is not environmental concept for them it is right now a daily negotiation with uncertainty.

So when we built our AI advisors we did something unconventional. Right now we are trying to do on the pilot stage. We asked farmers to train the system with us. You know the dialects the soil wisdom the lived patterns become the pattern of the model. This is where the governance comes into picture. To use the best of the technologies where you invent produce or do sitting in R &D use best of your grey matter to come up with some products until and unless we use and induce into our governance there will be no end result. That is what we believe in. That is why our Telugu first AI can record. Land records interpret satellite indicators and compress the time between the climate event and an incident settlement.

So this saved lots of time, you know, for our, you know, government agencies as well as to the end user as a farmer. Our satellite -driven heat analysis no longer stops at mapping temperatures. They now shape zoning, green bells and urban cooling strategies for Hyderabad. Which we are planning to take up to the core by 2035. And across 33 districts in our state, our solar power edge computer nodes ensure that the government service and the climate remains operational when the grid fails. And this is also one of the novel things what the Telangana is the first state where we have implemented. Yet I don’t claim that these are examples for climate. This is just a fact of a story.

This is just a beginning. This is the first preface we can say that because the real breakthrough is not from each project. It is from the architecture that binds together. Our future projects like our coming up the state of art infrastructure in the upcoming AI city, absolutely dedicated AI city and the Bharat future city which shall be the net zero city. Are designed not as a smart districts either for technology or for other aspects, but as a self -learning cities, territories that define sustainabilities, territories which can provide themselves for the compute and make them policy advisors. Our country’s first sovereign AI nerve center. ICOM you know this is our first ever initiative by any state in India that we have come up with the first sovereign AI nerve center that is supposed to be the AI innovation hub that is named as ICOM that the aim and objective is you know this intelligence should be shall go deep beyond just for incubation but also render into R &D and shall be the prime focus of creating AI ready talent for tomorrow’s world and I would like to mention here that Hyderabad and Telangana is the first state to come up with a platform that is This Telangana data exchange platform, the sovereign data open pipeline ensures that the intelligence is grounded in integrity.

So the platform is on the open. And this is the first state we have put all the data on this platform. You know, if we go through it by this open data pipeline, you know, 1 ,084 data sets, they have moved from administrative exhaust to ecological signal. We have created something rare in the global south, that state that generates its own intelligence at a scale. And we have seen the results too. And the results shown. The healthcare doesn’t wait for symptoms. It now anticipates risk. Because the data exchange we have done with our co -partners, even in the healthcare. or with the doctors or with the public health institutions. They are just not waiting to deliver the medication, but predicting the risk and try to put it into action.

And we are not waiting for the heat waves to come. We are trying to analyze through the data and how we should place ourselves and we are preparing corridors for the shed. And farmers also, we believe, using this AI technology, we don’t want farmers to wait for the loss. You know, they have to receive assurance before despair. And we are also planning that infrastructure doesn’t wait to break. You know, it has to whisper when it will fail. You know, when all this, the cutting -edge technologies, especially the AI, deployed with purpose and AI agents offer government something rare in public life. The ability to act before harm, to prepare before shock, to protect before loss and how resilient our infrastructure emerges, how safe the climate resilient cities take shape and how our public services become anticipatory humane and trusted.

And this is the future what we are imagining and we are trying to put all our actions into stream and it is this operating system we dreamt and we started running and I believe the next chapter of the statecraft will not be written in the boardrooms of traditional power centers but in the living laboratories of the global south. In the cities of Hyderabad, and the world can already see a preview of what an intelligent century of governance looks like. Let us leave Bharat Mandapam today. Here, while this great convention is taking place with a shared conviction that the tomorrow we are building is not just the smarter, it is braver. However, and you know, the great caption goes, AI for everyone, AI for human welfare should be the theme.

And also, we should, I as a policymaker, you as a technology expert sitting over there, should aim and anticipate for it. I thank the organizers for giving me, you know, a length of year to air. And I want to hear my pitch on behalf of our state of Telangana. I would like to thank the Salesforce team. especially the team management who are invited me over here for gracing this and having to see you know all the best brains sitting over here and the grey matter who would be doing much more for our welfare of our human being. Thank you very much.

Minister, thank you so much for joining us. We very much appreciate it. It was very exciting to hear what’s happening in Hatsheba and in Teliana. Let’s kick our panel off. Alright, so I am going to start with an icebreaker. Everyone gets 30 seconds to respond. This panel is by AI agents, so what would you say, I’m going to start there and then go towards me, what would you say is the single biggest difference that you see between AI last year, we’ve been sitting here and We’ve been sitting here last year in the AI agents that we are seeing today. Saibul, can you kick us off?

So I think in my mind the conversation has moved decisively towards agentic AI. We are no longer talking about, as Honorable Minister also said, about solving discrete problems or discrete searches. We are now looking at end -to -end AI -led execution of business processes or government processes. I think that’s the single biggest change in thinking that has come up.

Professor Lee Tiedrich?

To put this in context, I was involved in the International AI Safety Report, and we just had our panel on that a little while ago. And Professor Bengio was saying the biggest change from 25 to 26 is the emergence of agentic AI. And my perspective, is its ability not only to do the end -to -end, but to also act on behalf of the… of people is really the big change.

Mike?

So I’m probably going to jump on the train here. You know, what we were seeing last year was narrow agents able to solve specific problems. What we see now are agents that are able to abstract the problem, chain of thought reasoning, being able to take that and turn it into sequenced action and turn the multiagent sort of systems level thinking. So it’s the move from task specific to systems level is the big shift that I’m seeing.

And Srini? Srini Srinivasan

Yeah, so I think for me the big shift has been from co -pilot human in the loop to agents which can act and really provide value, business value. And that’s been the big shift.

So let’s talk about that value. Let’s talk about AI agents as a forceful multiplier. I’m going to start here this time Srini, you lead engineering for one of the biggest platforms in the world. There’s a lot of discussion about AI agents. Can you demystify this? What does that mean?

Yeah. So I think what does that mean? An agent, just like a human, first of all, an agent has to act. It says agency and it acts. That’s the first big difference. And like any agent, it has to have a couple of things. It has to know a role. Just like a human, it needs to know what it’s supposed to do, what are the jobs to be done. It needs knowledge. Just like a human, if I have in my mind, an agent has to have knowledge, some memory, so both short -term and long -term memory. And then it should also be able to act. You know, it should be able to, in a digital world, should be able to act on an API or something.

And then it should be able to act wherever the surface is. Maybe it’s in WhatsApp channel, wherever the user is interacting with it, in a WhatsApp channel or web channel or a digital channel or a SMS text. More importantly, most important in all of this is we should have guardrails on what it’s not supposed to do. that’s the most important and then all of it has to be covered to make it useful with what we call a trust layer because these things can hallucinate it can have bias, it can have toxicity, avoid all of that and they are unpredictable ultimately so it should have governance then it’s auditability, so you can do all of this this and all is to do all of this is what an agent does so this is also the why even though there is a lot of hype in reality it hasn’t diffused enough, this is the business value which we are trying to bridge as the vendors

Thank you. Saibal I’m going to go to you next, so let’s talk about governance, we sit here in Delhi, the capital of one of the greatest nations of the world, the public sector, are they ready for this, how do we think about that?

So I think let me not answer that question, I think the public sector needs to be ready so all the way from managing public finances public procurement managing their workflows and processes better, there is no way that public sector can avoid this. However, as Shrini, you pointed out, the stakes here are very, very high. So imagine an agent crafting an RFP, a multi -million or a billion dollar RFP on behalf of the government. How do we and you know in public procurement, we often sacrifice speed for procedural tightness. So how do we actually, what guardrails do we put around an agent or more? So can it really be end -to -end? Can it really be fully autonomous?

Or do I still need that last human layer to make sure that the T’s are crossed, the I’s are dotted because the stakes are really high and a mistake can really, you know, lead to a lot of negative impact. So I think the public sector has to be ready. but I think some of these guardrails has to be thought through and in the context of public sector are agents fully autonomous or do they still automate or do they still operate with a little bit of that human layer I think that has to be thought through.

That’s great thank you. I love that you said RFPs because that’s a concrete example so let’s talk a little bit about use cases and Mike I’m going to go to you let’s talk about resilient infrastructure one of the examples I hear a lot for AI agents they can help you make reservations and I love to eat I think making restaurant reservations actually pretty valuable to me but could an AI agent do something like design a bridge could it design an energy grid like where do we stand between reality and science fiction?

Yes I think we’re tracking pretty quickly to agents being able to do just those kinds of things In the past, what’s been difficult is using computational methods in AI, which has been around for a reasonable time for these things, has been very difficult. Because if you’re using some form of computational method or AI to design a bridge, you have to specify that bridge perfectly. You have to give it perfect inputs. Now, it turns out that when a designer is designing something, they don’t have perfect inputs. That’s the process of design is actually figuring out what your inputs are, right? So this has always been a little bit of a barrier for people to use these advanced methods.

With AI, and specifically AI agents, you’ve now got a much easier way of interacting. It’s more forgiving towards fuzzy requirements and earlier stages of thinking. It’s able to give you things that inspire you. So one of the things I talk a lot about publicly is that the notion of agents and creatives working in a loop together, that it’s breaking the cycle where the engineer has to come up with every idea from scratch, from a beginning. Rather describe what you’re doing, let the agents explore. So I’ll give you one example specifically in infrastructure because you wanted to get concrete. I mean, something that we work with is water systems, for example. So we’ve built AI agents that can analyze floodplains.

They can analyze how you might want to think of water drainage and these kind of things. So every time you’re making a decision early on in your design, you can let this thing run through, and it’s going to optimize your design in order to ensure that drainage is going to be successful on that. Now, drainage seems like a small little side thing, but it’s a pretty massive part of infrastructure. And having an agent handle that for you, it’s a pretty big deal.

Mike, I have very close family ties to Louisiana, so drainage and flood zones, that is not a small thing. That is a very, very big thing. And actually, that’s a perfect segue to the question I wanted to ask Srini. So one of the most complex things that a government might have to deal with is disaster response. Is that a place where AI agents could be helpful?

I really like the theme, welfare for all. And I think while we can think of very big things of where the AI is doing, AI can add value right now. And disaster response is one good example. Another small example. Another example which I wanted to give was like, you know, the key is to give back time to the people. That’s very valuable. Giving back time is a very noble goal in my opinion to everybody. So we have this very interesting use case where there is a city in New Thames in UK where they created an agent called Bobby. It’s like Bobby is a UK term for policeman and the citizens are asking a lot of questions which are not emergency and Bobby is answering them.

More than 90 % of them they get a lot of value. What was interesting for me was we have another city in Tasmania which is using a product agent force to roll out agents to their police people more than thousand police people because lot of times when they are in the field the policemen new or more experienced they have lot of questions and they are asking and they call this agent Terry and lot of policemen say Terry is their best partner. you know they have been so I think while we can think about futuristic ways here and now there are a lot of things we can provide right now with the technology guardrails in the public sector in private sector obviously where with if you have the right platform where you have trust governance as a foundational value with all the right guardrails we can still add a lot of value and we are seeing thousands of examples across public and private sector where you have the crawl walk run mode you know you start something basic you can still add value you still have the most esoteric cases with multi agent orchestrations I feel like but you can start with basic today and still get a lot of value that’s what we are seeing

that’s great so professor Tiedrich we’ve talked a little bit about how agents can help governments serve their publics are there are there risks there Are there risks of over -reliance?

Yeah, I mean, there are definitely risks, and I think I share the view of my co -panelists that I think there’s a lot of benefits to using AI in government and improving government services worldwide. But like everything else, we have to do it cautiously and smartly, and I think some of it kind of comes back to the human factor, like pick your use cases wisely. One of the themes in the safety report is that AI is emerging very jaggedly. We have some use cases like computer programming that are really good. There are others that may not be quite ready for prime time. So I think when we think about over -reliance is thinking about where AI is excelling, focusing on those use cases, and maybe doing sandboxes around some of the others to give them a little bit more time to mature.

I think also the over -reliance, picking up on some of the great points, is the guardrails. You know, one of the things in the safety report is good news. We’ve made a lot of progress on guardrails and risk management, but still as the technology moves quickly, a lot more work to be done. And so not relying too much that we overlook guardrails and thinking about where humans should be in the loop. And then the third thing I’ll just mention is, you know, the interoperability of different agents. And as agents start to call upon third -party agents, it’s just thinking through, you know, what guardrails, how do you choose that? How do you allocate liability? How do you test the agents that you’re going to bring into your system?

So guardrails have come up. Srini mentioned it. You just mentioned it. Let’s talk about guardrails a little bit. So, Srini, we hear about chatbots. We hear about hallucinations. Those can be annoying. When you’re talking about a government deploying an AI system, AI agents, the consequences can be extremely significant. They can be a hallucinating of an agent can be quite dangerous. So let’s talk about guardrails. How do you engineer trust into a system so that a minister, a secretary, a secretary can be able to say, feel confident that that’s a tool that they can use to serve their people?

drift, they can hallucinate. So you need a command center where you can say all of it is, this is the difference between a pilot or a demo, which you can find thousands of demos in YouTube versus real life, where these things become So we had to build all of these things for both the customers or governments to build confidence, they can audit, they can test, not even if themselves an independent party also can test, all of this infrastructure is what is required to make this a reality, but once you do that, there’s a huge value you can immediately provide to either the customers or the citizens.

Can I just add to that quickly? Because I think you hit a really interesting point at the end there. When people talk about guardrails, they think of guardrails as this perfect thing, that at some point the guardrails are going to get strong enough that every result is perfect, it’s completely predictable, and we’re good. And I think we need to talk about the honesty of that. We’re talking about systems that are inherently probabilistic systems. You’re never going to make a probabilistic system 100 % deterministic. It’s an acronymism. Right. So what we’ve discovered is that, I mean, you do do all the guardrails work that we’re all talking about, but where you were going at the end there about making systems that can look at the accuracy of what’s produced, give you some feedback on how accurate the solution or how well it’s going to perform, and then, and this is very important, what we’ve discovered is giving control to the human being, giving control, in our case, to an engineer, right, who is able to say, oh, I get it.

This is kind of, the result is a little off. I’m going to give it some more feedback. I’m going to reassess the results. I’m going to run it again. Or I might even go in myself and kind of tweak that information. And what we’ve discovered, when I’m talking to an engineer and explaining how this stuff works, if I don’t give them that level of control, they don’t trust the system. The minute they know they can actually control it, so it’s not, trust doesn’t depend on a perfect answer. Trust actually depends on transparency and understanding and then the ability to come in and control something.

But I think that’s also because the engineers understand this. This is the tool. It’s a tool for them to use. to help them. It’s not something that is going to take control. Is there anything specifically with respect to infrastructure that you think government should be mindful of?

Yeah. Well, look, so, I mean, infrastructure is not known as the easiest and quickest thing to build, right, in countries. And I think, you know, one of the really boring things but absolutely necessary things with infrastructure is to make sure your digital ecosystem around that infrastructure is set. And I see a lot of places in the world getting into building infrastructure trying to do this quickly without getting all that digital infrastructure in place. So building information, modelling, ensuring that every part of your infrastructure is correctly modelled, it’s represented at the right level. AI is not going to just magically come in and solve a bunch of problems unless you’ve got a lot of that digital stuff in place already.

So it’s kind of a little bit of the boring work, but getting that stuff in place early is one of the biggest things. I mean, I’ve had a number of conversations here this week about the 2047 initiative in India, the Man. of infrastructure that needs to be built in this country and the importance of using something like building information modeling, getting standard data, getting that in place now. If you get that in place now, all this AI goodness is way easier to deploy against it.

Yeah, please.

Yeah, so I think I heard a lot of discussion around sovereignty, and I think the way we should think of sovereignty is two levels. There’s strategic sovereignty and technical sovereignty. So by strategic sovereignty, I mean is like you get control on data, your governance policies, you know, and your operational policies. That I think you can implement it right now and get value. And I think and then on the technical one where people want to control their entire supply chain from the chips and all, I would like to for governments to and public officials and policy officials to think is as two tracks. One. One takes longer in a lot of capital investment. Don’t let the second track.

stop getting the benefit of the first track the first track is easy you can ensure the data doesn’t leave your country your policy guardrails have control human in the loop you still get a lot of benefits while you still want to continue on the second track that would be my request to all the government

can I just make a quick build on what Mike said because I do a lot of my work in the public sector with governments I think one of the biggest guardrails beyond policies is actually the skilling the upskilling like Mike you said it’s a probabilistic system inherently right so you cannot expect it to give correct results all the time there’s nothing called a correct result so the person who’s actually using this who’s using the tool at the district level, at the state level to make real government decisions that person is not an AI engineer that person needs to be upskilled and needs to be told what can be trusted and what requires that additional layer of check.

So I think if agent TKI has to take off in public sector at scale, then that upskilling at various levels of the government on what can be trusted and what cannot be trusted is also a very, very big component.

Yes, I totally agree. Professor, I wanted to ask you, so it feels so trite to say technology is moving really quickly, but in the last few years, I mean, AI is moving very, very quickly. We’ve talked a lot about guardrails. How should governments think about this? I mean, how are governments going to be able to keep up in terms of setting government expectation, setting potentially regulation for a technology that is moving so quickly?

It’s a hard one. I think, you know, AI has evolved into a global multi – disciplinary field. And I think, you know, we need to bring the global community together. I think we need policymakers, lawyers, talking with engineers, talking with sector specialists to really inform the policy in real time. I mean, I’m a big fan. I spent a year working at NIST, the U .S. National Institute of Standards and Technology. And, you know, we need to figure out how to do some of the guardrails, you know, starting with the science. And then the science can inform, you know, how to develop the standards, how to develop the evals. And then it becomes, you know, a question.

I mean, different countries have different views on whether we should regulate or not regulate. You know, the U .S. has a very deregulatory approach. Europe is the opposite. But if we can kind of agree on what those common standards are for evaluation and testing, then governments can be free to decide, do we mandate this or not mandate that? And I think one important thing is that we need to be able to do that. And I think that’s one of the things that we need to be able to do. And I think that’s one of the important nuance to add to the mix. And this has been a theme of the conference. is, you know, we have to, well, we want some standardization on these evaluation mechanisms.

We have to recognize that we speak different languages. We have different cultural norms. So when we want to have standardization, we’ve got to be able to localize what the evaluation looks like is what might be appropriate in one country isn’t going to be appropriate in another country. So it’s hard, but I think, you know, starting with the science, the scientific report, I would point people to, you know, building on that, working through the AC network, working through standards organizations and all these other initiatives to develop the evaluation, build that evaluation ecosystem, and then regulations can kind of overlay on top of that as policymakers think appropriate for their jurisdictions.

But if I could ask a follow -up question to you or any of the panelists, I mean, I think one of the challenges there for companies is that it’s really helpful for companies. I also speak for enterprise software companies that I represent. It’s helpful to know what those government expectations are. Like industry is looking for clarity and predictability. And I think that’s one of the things that I think is really important. And I think that’s one of the things that I think is really important. And I think that’s one of the things that I think is really important. And I think that’s one of the things that I think is really important. And

Should I take a shot at it? Yeah, I see. As a software provider, you know, at Autodesk, we definitely deal with that, Victoria. You know, we’ve had a couple approaches. One, I mean, we’re obviously going to stay on top of this all the time, working with governments, making this part of a conversation. I spend a good part of my year traveling around the world, talking to governments and trying to sort of help them understand what needs to happen, but also help us understand, like you said, what they’re wanting. But the main problem is just the sheer variance. I mean, even within the United States, we have things between different state efforts, right? And then you get around the world, it just gets even more complicated.

What we’ve tried to do is we’ve – We’ve tried to run as far ahead of this as we can. So if there is a way that we can build in good controls right from the beginning, we actually build those controls to the maximum extent that we can within reason, right? So what we’ve done is we’ve found now that we’ve run, I’ll give you an example. In every AI feature we have in our software, we have something called a transparency card, which looks like a nutrition label on food. But that nutrition label tells you what kind of model is behind it, what data was used to train it, what kind of level of control you have, what accuracy it has, any bias that we know about in the model, that kind of stuff.

And it’s a standard thing. So we rolled that out about a year ago really to try and stay ahead of things. So if governments started asking for these things, well, we’ve got a transparency card. What’s actually happened now is that there’s a bunch of interest in that becoming part of a standard. So I mean I’m not saying that really just to tout us because I think other companies are doing great things in this space as well. You guys are doing a bunch of good stuff in this space too. I think this is an opportunity for us. For us in industry to run ahead to try and help define some of these things because it is moving so fast.

And I hate to, maybe I shouldn’t say this publicly, but the government doesn’t always have the best answers, right? So, I mean, we can work with government to help them develop those answers and come up with good things, which helps us then, you know, resist some of the complexity that’s coming down the line.

Yeah, so one of the challenges in this is you can project it too much. It’s an exponential curve. It’s very hard to project. So I think sometimes it’s learned by doing. I think the biggest thing the government can, all governments can do is the policy framework on how to update these standards. Today, usually it takes a long time, and so everybody’s afraid, and then it takes even harder to change a standard. So then everybody, so then they try to solve everything, and things are changing. And so I think the main thing policymakers could do is, just like the feedback loop, if there’s a way to improve. If there’s a way to improve the policy framework, because then you don’t need to be afraid of getting everything right.

You know, you understand that, hey, you told me some basics, and as new data comes in, you can update it. And then I think that, I think that in engineering and product, we call this the product feedback loop and agile development. If we have something equivalent on that, then I think then everybody is clear because we all want the right thing. I think there’s no disconnect in the foundational thing. We want AI to help, correct, in a positive way to our net positive way to our entire community. And the regulatory framework in a changing technology, if the regulatory framework is able to change, if we can change that, then we are not afraid to say that we need to get everything right on day one.

And we can learn by doing it. So agile regulation.

I have loved this panel. Unfortunately, we’re coming to a close. So I’m going to ask each of you one final question and Saibala, I’m going to start with you and then head this way. But if we were so fortunate to meet again in three years, so fortunate to meet in Delhi again in three years, looking back what would you say is the one thing that you think would be the best way to determine whether or not we have succeeded in addressing some of these challenges i know it’s a big question sorry but uh thank you

so since we’re in uh delhi i’ll give the answer in the indian context i think as um the primary theme or one of the primary themes of this particular conference is inclusivity so for me the success of ai the true success of ai will be if a farmer could talk to a small language model powered tool in his or her own vernacular language and get practical advice on how to manage the crop how to manage the cattle and if that could be scaled up uh across the board i think that would be a very good idea across the length and breadth of india Then I think that, for me, is the real win for AI.

That’s a

big win. I mean, that’s a significant impact. Thank you. Great. Professor Tietrich? Yeah, so

I’m kind of coming back to the evaluation ecosystem. We’ve made a lot of progress over the last couple of years, but more work needs to be done. You know, more countries, including the Global South, are launching ACs, you know, AI safety or security institutes, which is not hard regulation, binding regulation, but it’s governments weighing in. And I think real progress three years from now, we have an active AC institute that’s sharing information, making real progress on evaluation techniques, and one of the commitments that came out of some of the companies yesterday is, you know, also localizing that so everybody can benefit from that Global North, Global South. Thank you.

Mike? So earlier

on, I spoke about infrastructure and, you know, physical infrastructure that is in countries. What I would hope to see is, in a couple of years, in a couple of years’ time, we’re actually seeing infrastructure genuinely get developed faster than it’s ever been developed, which is a really, really tough problem, making that happen in the physical world. So as a measure of AI truly doing this, that’s an incredible measure. But on top of that, it needs to be doing that without compromising safety, without compromising. It’s not a big black box that nobody understands, right? So what I would love to see is not only is that infrastructure being developed faster, but the public is engaged with it.

The engineers and people that are doing it feel comfortable with it. They feel secure. They feel fine signing off on that because they feel that this is reliable. Thank you.

Srinu? If AI

is so revolutionary as we all assume, I would hope in three years, the bottom 50 % income percentile, their per capita income has been measurable. That for me is the real impact of this technology. That’s

fantastic. I want to say thank you to all of our panelists I want to say a special thank you to Srini and to Salesforce for bringing us all together here today thank you to our audience for joining us big round of applause for our panelists thank you you Thank you.

Key area should we think about it as a governance data governance problem, problem in model design or should it be more on a verification or compliance angle.

Thanks thank you very much Shyam for having me and good morning to everyone and welcome to this session. So I think okay let me maybe just start with saying that I’m not an AI or AI safety expert so whatever I say take it with a pinch of salt. My work is in biosecurity and that’s the angle I’ll come from. I think all of those things whether it’s a model evaluation and other things those are there and those are very very important factors and that those are the things that we need to keep in mind. But on top of that there is also a very important deep structural change that is happening. For example in the field of life sciences historically whatever risk and risk governance things that we had were very much linked to the physical infrastructure and lab facilities and facility inspection and material transfer control and things like that.

But that seems to have changed and seems to be changing very rapidly now with the kind of AI biodesign tools as well as LLMs that are emerging. So I think Rand also did a study on this, but there are more than probably 1 ,500 biodesign tools that are out there, and those are totally transforming how life sciences, but in general, science is done. Now, what kind of change that we are seeing is with these capabilities, now it’s much easier to engineer proteins, optimize DNA sequences to do things that we want, have better pathogen host modeling, interaction modeling, and things like that. Now, these capabilities are… because of AI becoming partly decoupled from the physical containment measures which were usually used in the life sciences.

So we have a lot of this risk landscape shifting a little bit more upstream to the design side when it comes to at least biological side of things. So yes, data governance, things matter. Model evaluation and red teamings are essential and we should be doing that. But also it is very important that especially for a country like India where we have a very vibrant scientific ecosystem but that is also very uneven. How we can use this AI -enabled science which is rapidly evolving into the existing mechanisms to some extent but also at the same time develop those capabilities, have more people with the core capabilities and more people with the core capabilities and more people with the core capabilities chemical security, AI nuclear security, and things like that.

So we need to train more people on those things. So integrating, again, going back to the life sciences, so integrating AI evaluation into biosafety system, strengthening the institutional readiness. Some places there are information, some labs and some institutions have information security labs or information security offices. How we can get them better prepared for these new emerging risks that are coming due to AI. Some places they have biosafety officers or biosecurity officers. How we can enable them better to address the AI risk is what the direction that we need to move towards. And have more adaptive oversight mechanism that is not only based on the, limited to this once in a while inspection that happens, but that goes more with the rapidly evolving things that we are seeing with the AI models coming up.

And I think, I think, So, just in terms of paradigm change that we are seeing and that you mentioned, is that there need to be more decentralized checks and balances and oversight mechanisms. If there is one authority sitting somewhere in Delhi and trying to do everything, that’s not going to work. So that is one of the things that we have to collectively think about. How do we decentralize these kind of oversight systems to some extent? For example, as I was saying, how we can empower the information security or biosecurity offices and create what in the field of disarmament where I have worked on called way of prevention. One measure is not enough. It’s not sufficient.

You need to have a number of measures in place which collectively can help prevent something bad from happening. Thank you.

Thank you. That’s very insightful. And I think we’ve already touched on some areas that, you know, that would be follow -up questions. P .T., focusing a bit more on… open science where high risk domains, especially in biological data and AI capabilities, as Surya was mentioning. How do we preserve the benefits of open science while preventing the destabilizing diffusion of capabilities that we were just discussing about?

Thank you. Thank you for having me today. So I guess like I would love to be able to give like a binary yes or no answer. Right. I think we all want to have that. But unfortunately, that’s not quite the case. So we need to find a way to balance the openness and also the restrictions as well. So I guess my answer here would be sort of like a tiered access and contextual norms. I think those are really important. And I think RAN Europe has done a really great job at establishing the global risk index on AI enabled biological tools. And also just generally looking into AI safety in general, where they do this thing where they call the pre -deployment assessment.

with structured rubrics. And I’m a huge fan of that because I think that when you release very frontier models and frontier tools, the danger is already out there once released. It’s really hard to withdraw the danger. But however, prevention, right? There’s this window before you release where you can do a pre -deployment assessment. So I think I’m a really huge fan of that and also the same way that I’m a big fan of KYC, know your customers. And I guess this principle also pretty much applies whereas in the case of biosecurity, where we differentially allow the development of medical countermeasures and also the defensive measures that is necessary for the research, but also don’t limit the researchers from actually innovating either.

And I guess my point here is that we’re not going to be able to do that. The non -safeguarded access, like private access to credential researchers where necessary for like defensive research is absolutely necessary. And then, you know, like open source tools, it’s necessary. Like we can’t turn away from being open source. Like any governance structure that conflates open source with danger makes a huge mistake because that also is a very critical development point, especially for lower resource settings. So we cannot afford to conflate that altogether. So I guess a very long way to answer this and then to summarize my answer is that differentiated governance at capability level is always better than blanket restriction at access level.

Yeah.

I think that’s a very structured answer and I think, you know, there’s a start of a very valid framework level conversation that’s already happening there. Geetha, turning to you, thinking more about institutional gaps in enabling some of the solutions that we are discussing, potential solutions, what are the most immediate gaps that you see in evaluating systems, technical capability, regulatory and coordination, largely from the policy angle that you work in?

Thank you, Shyam. Good morning, everyone. So on the technical capabilities, right, the most fundamental thing I see is the AI readiness aspect of deployment. So in general, when we see India stands or ranks third globally, and when we see the Southeast Asian countries, I think Indonesia is around 49, and so there we see the gap, right? So whatever we do from the Western context or in the Indian context can never be catered to the AI readiness aspect of deployment. So I think it’s important to the needs, the unique needs of the Southeast. Asian countries and moreover what there is the end user perception where we see that we have to build lot of capacity for creating awareness among the end users who are actually going to use the products and from the policy perspective I would like to give you certain aspects where we think about the socio -cultural aspects that is relevant to the deployment environments.

So in general the large language models are usually trained on the western data and the very recent research work maybe I will cover a bit of both tech and policy here. So there is a Southeast Asia related benchmark, safety benchmark which says that all these leading large language models have failed when they evaluated for more than 20 to 30 percent of the risk. So in the biological settings so which means that we did not have enough safeguards which will protect people from encountering all these risks. And moreover, so this lets us know that we have to build in more sociocultural evaluations and assessments which will cater to the harms that is more particular to that particular deployment environment rather than just having a high level evaluation strategies.

And this cannot come just from the policy side, right? So we need to bring in all the participatory approach which will bring in the end users, the different stakeholders involved in using all these AI systems, be it model, right from the requirements definition, right? So when we assess whether we need an AI system or not, generally now there is a perception saying that for whatever we are going to build or the problem that we are going to solve, by default we assume. We assume that we need a large language model which will not care. which is not even possible to have it deployed in a low resource setting, right? So we need to think about small language models which will enable edge deployments at the low resource settings and also consider all the multicultural and socio -economic diversity that exist in these regions so that your model doesn’t hallucinate, is still fair and also establish some governance and accountability frameworks which will make the developers more accountable and also because having the developers more accountable will enable them considering more safeguards, right?

And also create more awareness about the main fundamental thing is that they will be expected to document whatever testing that has been gone through. And on the policy side, there is one more aspect which is the Indian government also endorses, right? The self -regulation. voluntary commitments on managing and mitigating risk that comes out of all these AI models. So I think we have to have a unified framework which can still be adaptable to different deployment settings.

I think we are already getting a diversity of perspectives here and it is very useful to hear. Moving ahead and thinking about institutionalizing these kind of capabilities in scientific AI context, PT turning to you. Should independent evaluation and red teaming of AI systems from a technical kind of solution perspective for this problem that generate biological outputs, especially thinking biosecurity, given your perspective on this, should it become a norm and part of the global scientific specialist infrastructure? And if so, how would we go about that?

I think we have to have a clear understanding of the role of the AI system and how it can And I think that is a key point. And I think that is a key point. And I think that is a key point. And I think that is a key point. And I think that is a key point. And I think that is a key point. And I think that is a key point. And I think that is a key point. So I guess a good example to use here is probably we’re thinking of nuclear weapons, right? Which falls under this organization called the International Atomic Energy Agency, the IAEA. Now, from my perspective, I think fissile materials, correct me if I’m wrong, they’re very scarce.

And they are, to a certain degree, technically trackable. And they are also, more than anything else, highly regulated. Whereas biology, on the other hand, is everything but that. It’s diffused, it’s dual -use by nature, and it’s also nearly impossible to trace. And also, most importantly, commercially available, right? And so in the recent study, actually, this was done by this organization called SECURE. Bio, where they actually tested frontier large language models against expert virologists. And it turns out that ChachiPT -03 actually outperformed expert virologists by 94 % at troubleshooting wet lab protocols. So that’s a very shocking number, right? And then, I mean, obviously you mentioned earlier that there’s a very concentrated effort that is happening between the US, UK, and China, like the global superpowers, basically.

And I guess there’s, we, in the recommendation from the RAND Europe that I was, you know, helping out with is that we recommended that governments and also independent researchers do this six -monthly ritual of monitoring and also assessment of risk on a continuous basis. And we also suggested, obviously, like using AI as an automation tool to increase the efficiency of this risk monitoring system. But I think, to your point, I think stuff like this, stuff like that is non -interactive methodology that doesn’t require, you know, researchers to actually query directly with the danger systems is actually already in and of itself a very meaningful, you know, safeguard. But that is not enough. You know, we need something that is much larger than that.

That is the integration into, like, you know, institutionalizing it. And I would argue that, like, a six -monthly, you know, ritual, that refresh cadence, for it to be delivered, it’s going to require a very significant investment from the government at multilateral level, right? And so we can’t go without any investment at all. So my suggestion would be to actually implement this AI safety or security institute model that we’ve been applying where largely… It is technically credentialed. It’s independent, but also has a very… formal relationship with the government. And something that I would caveat from the bio side is that for the institution to have some kind of anchoring around biological weapons convention or the WHO.

Because right now that relationship is not quite there yet. And I think, you know, back to my point of like pre -deployment assessment, I think that is definitely needed and then the result has to be shared then across the credential network with tiered confidentiality that rather than being kept, you know, as a proprietary to the different state. I think it’s kind of a

That’s an interesting position, PT. Suryesh, thinking more about safety measures at large, how can we make sure that they remain rigorous and feasible within research ecosystems that you’re quite familiar with, you know, from a biosecurity angle, if you will, but largely also in the larger scientific ecosystem.

Thanks Shyam. I think first, yeah first thing that we need to understand is how that ecosystem is and then see if certain measures will work there or not, right. One of the hallmarks of let’s say Indian scientific ecosystem is there is a lot of heterogeneity. There are some places which are really extremely well performing and there are other places who are not well resourced or have other all kind of challenges. So, understanding how the ecosystem is, what kind of regulation within the institutes that are there, what kind of administrative measures that are there, what kind of safety teams these kind of institutes might have, all of those things are extremely important, right. The governance capacity, compliance culture and technical expertise varies widely in Indian institutions.

And I believe this is true for many other countries in the global south as well. So it’s not something very unique. Particularly to India, we have challenges related to different kind of resources. And even when the resources are there, sometimes it’s also problematic to use them efficiently enough. Now, given that context, if we just import safety frameworks that are developed in a well -resourced place in a Western country or any developed countries, I don’t know if those would be a very good fit for the kind of system that we have here. So those might become more performative than functional to some extent. Another challenge that also P .T. mentioned to some extent is that the speed and scale of AI is huge, right?

And we need these traditional review mechanisms that institutes have for safety audits and all of those things are not going to work. We need something which is far more adaptive and quick. And also what we had traditionally is this periodic paper -based facility -centric kind of measures. And those are very much outdated in the era of AI that we live in. Now, so what… Now the question becomes, how do we design proportionate capability -aware safeguards that would be better matched for the challenges that we have? One of the major challenges, as I think a lot of us realize, is that there is limited awareness about AI safety when it comes to scientific issues, even among the scientists.

So a huge number, a large majority of scientists just don’t know what they are putting, let’s say in chat GPT might be harmful or what they are getting out of biodesign tools could be harmful to some extent. So there is some understanding about the privacy -related issues, but safety and security is still a big gap in understanding of even the scientific experts that are there. Now also regarding AI, I think there needs to be a tiered risk classification. So not everything is highly risky. There are certain biodesign tools, for example, that are trained in… in virus data. Those we’ll put in a higher risk category compared to something which is just working, let’s say, on certain animals which are not dangerous.

Now, also the safety measures, as I was mentioning earlier, as the risk has moved a bit upstream, it has come more on the design side, we should also have more safety measures moving upstream. And as Piti was mentioning that, you know, certain kind of evaluation that are before launching AI tools are necessary, but also integrating AI evaluation modules into grant review processes, creating cross -trained AI biosafety review panels, so panels specifically for AI biosafety at, from the bottom -up side, instead of having them from the top -down approach. Investing more in domestic evaluation capacity, having more AI safety institutes like Geeta’s home institute at IIT Madras. So we need a lot more of that. And lastly, I think what we have in the US and UK are these, a lot of AI safety work is being done there, right?

And as I was mentioning, importing that directly might not work. And we in the global south are largely the users and importer of this technology. So we have to see from the bottom up side, where do we put those safety measures? Do we, like when it comes to import, what kind of, when the data is being transferred, is there certain places where we can put those kind of safeguards? Also, how we can use some tech sovereignty measures in this context, right? That tech sovereignty measures are used for a number of things, but AI security is something, AI safety and security is something where those could also be used to some extent. So, yeah, I would stop here and then we can discuss.

Thank you.

Thank you. And I think a lot of useful thoughts here for us to explore a bit more. I think we’ve… just crossed the mid mark and I’m going to use Geeta to kind of like bridge between the next two topics by combining two of your questions sorry for that so just as Surya just mentioned will the emerging scientific powers you know global south middle powers would they be able to shape governance in this context especially you know enable science or will they continue to inherit the frameworks and if they were to show leadership what would that look like in scientific AI and research ecosystems and you know you’ve already been working on some of this so I’m looking forward to kind of hearing concrete measures that you know are happening

Sure. So in general what I think is definitely the emerging powers right they are putting on all efforts to bring in all the tools and frameworks that are required for governing these AI systems and for example, so India’s strategy towards all these emerging techs is that they are trying to create sandboxes which are highly essential for deploying or evaluating safety aspects for the models, right? So they do it for healthcare systems, they do it for ideology systems and whatever, right? So these type of tools and frameworks come from Indian settings will actually help the other underdeveloped countries to learn from the strategies that we use and then build something of their own or something which cannot go cross border can still happen through learning and collaboration, right?

So for example, we are going to launch a global south network for trustworthy AI and we are going to launch a global south network for trustworthy AI and we are going to launch a global south network for trustworthy AI which will enable all these mechanisms to happen, enable people to… develop and deploy AI systems which will be deployed in the low resource settings. And the other initiative which is going to give a very big leap in evaluating AI safety is coming up with an AI safety commons for the global south. That is part of the safe and trusted AI pillar that is one of the pillars in this impact summit and I think in another one or two years we will have safety commons which will help us evaluate and assess how these AI data models and systems work for different deployment settings.

Another important thing is that as Suresh mentioned about the audit frameworks. So when we come with, when we focus on the kind of risk and audit mechanisms that we have here, we still have it from an organization perspective and not from the end user perspective. So at CRI, we have come up with an incident reporting mechanism and a framework that caters to the Indian settings. So it tells you how to operationalize incident, AI incident reporting in the Indian settings, which is completely different from the Western settings. And here we have to get the harms that the people experience in the marginalized communities, which will never be recorded everywhere, right? So how do we enable all these things?

So since it is all about all these CERN -based systems, right, even those things will have certain impacts to the marginalized communities, which may be an indirect impact. But how do they are knowing about such things are happening to them, right? So those kinds of gaps we should mitigate by building more awareness, creating more AI literacy. And we should also be able to provide more privacy to all these people. The final thoughts about combining all these things is that we have to bring in some kind of collaborative work between the different stakeholders who are involved in developing and deploying these systems. And the governments have already given certain prompt knowledge about how to enable all these things through the techno -legal framework and guidelines that was recently published and the AI governance guidelines.

Which was recently published by METI. So the Southeast Asian countries can learn from the developing countries like India and then have curated a more tailored approach towards their unique needs. So that is what I think. So whoever has an opportunity or a willingness to have more things that will actually help them use or leverage these technologies can learn from whatever. Learn from. the mistakes as well as the experience that the other countries have, which is now openly available through all these summits.

That’s very useful and I’m looking forward to following up on IIT Madras’s work in this front as well. Going to Suresh for kind of the last question in this series really, should, you know, safety measures, evaluations, primarily focus, where should the focus be at the model level? And you talked about upstream quite a bit. Should there be more broader socio -technical readiness measures, misuse considerations? Where do you think it should be?

And also, very importantly, how we have to also see it from the context of, you know, people doing their own thing, DIY kind of science that happens. And also, small -scale commercial activities which are not fully under the oversight mechanism of the government, right? So, considering all of these points, right, the policy evaluation must expand from model -centric assessment to socio -technical assessment. And this would include, you know, evaluating things like how much capability uplift relative to the government capacity that is there. So, government has certain capacity to manage or do oversight, but these AI tools, how are they changing that? Incentive structures, very, very important, that shape the model deployment. Also, the diffusion of risk across borders.

All of these things don’t respect national borders, right? So, how it’s going to spread. If people using VPN or other things, a number of other things that are there. So integration, lastly, the integration with existing biosafety and resource security systems as I had already mentioned. So briefly, like performance evaluation is necessary, but governance -relevant evaluation must be systemic. And otherwise, we risk auditing algorithms while ignoring the institutions that operationalize them. And that is very, very important, how we focus on that institutional level mechanisms. Thank you. Thank you.

Piti, kind of the last structured question before we move into a bit more of an open conversation. AI becomes embedded not just in new capacities, but also existing programs like biosurveillance, public health systems. And so there’s a mix between emerging kind of scientific knowledge with more legacy, let’s call engineering knowledge as well. So. So how do we make sure that safety, evaluation, interoperability, all of that exists in this divide without fragmentation happening across the ecosystem? Because, you know, you can easily imagine everyone’s doing their own AI, you know, safety evaluation and not necessarily talking to each other.

Thank you, Shyam. I think this is a very important question. And it’s also a topic that I’m really passionate about as well, which is biosurveillance. To your point, I think, you know, countries are already deploying AI -enabled biosurveillance systems that are, you know, either syndromic surveillance or it could be, you know, genomic sequencing pipelines or outbreak modeling. The countries are already doing that, but they are not building on… the unified data standards. So they’re basically building on very incompatible data standards with very different legal regimes across the borders. We’ve seen that in Southeast Asia. We’ve seen that even countries like, for example, Singapore to Malaysia, you see different legal regiments on how they monitor the data and also the biosurveillance.

And so the fragmentation risk is actually not a technical risk, I would argue, because it’s not just a technical risk, because we’ve seen COVID. I feel like if anyone is anybody saying, I think we all were a little bit traumatized by COVID. We’ve seen how data hoarding and incompatible reporting actually cost lives. And I saw that especially happening across the region in the lower resource settings. Like countries like Cambodia, for example. AI systems that are trained on non -representative data obviously are going to perform much worse. And guess what happens? When they perform worse, the region that is most affected is the region that needs the help the most. And because of that, and also that region is also the same region with the least data infrastructure.

And so I guess to sort of like answer your question and what I think we need to do, I think there are three things to be addressed here. The first one is obviously the data standards harmonization. Currently, we don’t have that. I think we would need not like a global overhead standard that enforces on every country, but more of a federated interpretability that applies frameworks that applies to different countries. So I can think of like HL7FHIR, which is the federated… healthcare interpretability resources that are attempting to address these very specific issues on clinical data, but this one would be adapted for public health surveillance. And the second point is the legal safe harbors for basically just kind of cross -border sharing of data for public health emergencies that are negotiated beforehand because, and this is important, beforehand, because if you negotiate during an outbreak, people are going to be freaking out.

People are going to be like, I’m not going to share my data to you. What are you going to do with that data? So this needs to be done beforehand. And the last point, and also the most politically challenging point, is actually to have some kind of shared evaluation criteria across the board between different countries that are embedded into the national surveillance systems. And, for example, like Singapore data infrastructure environment might not apply to countries with like different climate data or like different demographic data. So this needs to be applied into, you know, the national surveillance systems. And what I noticed, I guess like the last message is that what I noticed the AI governance framework often thinks of biosurveillance as like an edge, like a niche edge case.

And then people in biosecurity frameworks, like doing biosecurity frameworks, thinks that AI governance is like a tool. And these people don’t talk to each other. And that gap, that gap right there is where the risk happens. So, yeah, we just need to talk to each other more. That’s easier said. Yeah.

So I think I’m just about to close with maybe five minutes or just under that for audience questions. Thank you, Justin. 10 second final thoughts on each of you from the panel. Suresh.

Just wanted to very quickly, we need to also keep in mind that how AI could help solve some of these AI safety challenges. How agentic AI could be used, let’s say, when people are trying to develop vaccines. CEPI has developed this platform where agentic AI is being used to check if there is someone who is trying to jailbreak or someone who is trying to misuse the tool that is there. Second very quick point, also, with all that what I said, there is still a gap to transfer things from digital to physical, what is called digital to physical barrier. So, even if you have everything, you still can’t just develop, modify viruses without having a proper physical infrastructure and there are still some ways to control that.

Thank you.

I think we should move on transforming from issues to intelligence like learning from the risk that happens and feeding it back to the model training and other assessment activities to mitigate the risk in real time so that is where we need to move towards bringing in more people into evaluations and then making it safer for people to use

I guess I’ll make it quick the point that I want to make here is that Shurya should echo his point I think you’re right that we should not shoot ourselves in the foot especially for developing countries, I think it’s really important and so my message for the last message here is just kind of like while we are forging ahead in innovation and while we are innovating ahead in whatever domains of scientific domains that we’re doing we need to be conscious of the impact that we have and I think in the AI Impact Summit is one of the really good places to jumpstart that kind of conversations and break the silos. Thank you.

Thank you everyone. I’m just going to take probably one minute to kind of summarize key points Evaluation, I see largely a systemic question, safety measure systemic question. I especially like the point on incident response not being already there. And a couple of points on the cross -border solutions and problems, we already have that. Discussion on open signs, we talked about how managed access, safeguards, and comparing government capacity to manage that versus letting it out for more DIY -oriented signs, which is a good term, I really like that. That’s a key area. And for emerging scientific powers, of course, collaboration is key. Tailored approach, that’s something that I’m again waiting to see from IT Madras as well, their contribution on this.

And some cross -border work on legal safe harbors, data standard harmonization, PT that you mentioned, really land well from this panel. I’m going to… I’m going to stop my… summary right now and you know more of this would be kind of put together in a blog at some point in the uh nearby future uh perhaps uh we can go for questions uh first uh yes please i think i can give you mine

Thank you so much for your wonderful insights i really enjoyed this session as a researcher in safety of ai at the university of york so i focus on psychological harms of ai and so what i want to ask particularly gita is um when it comes to the definition of harms and traditional safety engineering they’re catering more to physical harms and now we see the whole spectrum of harms expanding beyond that so i would love to know the work being done by karai and you in this area and and in fact enrich my research with it

Yeah, sure uh so when we actually assess harms and impacts right we do we have to do it from the different two different perspectives one is on the functional side where we assess all these algorithmic risk and other stuff. From the human centric perspective, like you said, we can keep doing everything from the psychology perspective and other ethics and other stuff. So, here at CIRI, we do work on assessing bias, determining whether the model is stereotypical or not and how do we generate explanations for the high level scientific models and all. So, from the perspective of the psychological things, there is this cognitive science or cognitive capabilities of AI models which will actually enhance or degrade the capabilities of humans.

So, those things are we are trying to do some assessments from the incident perspective. So, if you go to read the incident reporting framework that we have, we have a taxonomy of risk and harms and also the impacts. So, from the kinds of harms that we have defined, we have categorized it as physical, psychological, cyber incident based harm set. And moreover, we have all the generic kinds of harms like algorithmic harms, socio -economic harms, the environmental harms and all. So, we are trying to come up with a taxonomy that will cater to the different hierarchies that will be applied to these kind of harms and impacts which will again be model specific, use case specific and the domain specific.

So, that is where we are trying to work on. And we also have a healthcare based tool, a toolkit which will enable people to actually assess the perceptions of how they treat these models, how they see whether these AI applications are helpful for them or not and then come up with some capacity building programs for different roles in which they are working on. And this has been done with CMC Wellure Hospital and we have been assessing the perceptions of healthcare workers. and then come up with a training module which will enable them to use AI models or tools more confidently rather than, say, being resistant or not relying on them for so much.

Last, probably last quick question. Maybe keep it short on the responses as well, please. Sorry.

Hi. So my question is about, like, we are discussing all the geographical barriers, right? The modality is geography. When we change the geography, the models tend to perform poorly. Are we concerned about the temporal modality as well? When we go forward in time, the data is going to change eventually, and that is going to affect modeling. And how do we plan on, like, you know, mitigating such a problem if it arises?

Yeah. So this comes under the model monitoring, the system monitoring approach, where we consider the data drifts out of distribution. So we consider the distribution aspects of the data and models. So definitely this is one of the criteria where you assess safety and evaluate the impacts of it.

Yes, I think last question

Thank you so much for the insightful discussion, really appreciated the expertise that you’re bringing to the topic and thanks PT for bringing up COVID because my question is about that. As we learn from COVID biosecurity risk can quickly become a cross border existential threat. So what would a successful web of prevention and incident response framework look like and who are you looking up to in this space? Like who’s doing it well in this space?

I can start maybe PT can add. So I think as I was mentioning, it will have to be more decentralized but at the same time integrated to the leadership. So I think there needs to be more empowerment of people who are like biosafety officers in the lab or who are institutional biosafety committee members, who are people who are working on the ethics and research security side at the institutes. So those are the people who need to be empowered. So there needs to be more capacity building of those people and at the same time there needs to be a mechanism established so they can report those incidents to the very top and there is top leadership sitting in the capitals.

They can in some way get an overview or monitor the situation as it is going on at different institutes level.

Thanks. I can add a little bit to that. So in Singapore we actually have different agencies responsible for this. So we have the National Environmental Agency and then we have the MOH, obviously the Ministry of Health and then we also have different smaller agencies like Communicable Disease Agency and also like Prepare Agency where they are responsible for different tasks. But I want you to envision this as almost like the way that Singapore is trying to establish itself. I think it’s trying to establish itself almost as a firefighter. So when there’s an incident where there’s a crisis, who is actually doing what is very clear but it’s always not always clear across like different countries. For example, in Laos, Vietnam, might be looking very different, but I think having a very coordinated response across the different agencies on who is doing what.

Like, for example, National Environmental Agency is responsible for wastewater surveillance. So monitoring how the sickness is increasing or spiking or not, those are the people, yeah, that you would look up to. And I think that’s the last word, right? It all comes down to prevention and preparedness, even in this much like anything else with biocontext.

Thank you, everyone, for the question, and thank you to my brilliant panelists, Suryesh, Geeta, and P .T. This was a very insightful discussion. On the screen is the work from RAND Europe with CLTR, some of what was referred to by P .T. and other panelists as well, some aspects of what we were discussing about risk typification. You’ll probably get some ideas there as well. And with that, I close. I’m surprised. I’m supposed to hand over these mementos to apparently including me, so let us do that now. Thank you.

Institute. Then we have Raphael Leuner, Data Scientist at the German Federal Foreign Office. We have Dr. Shahani Yaktiyami, Senior Officer, Technology Program at the German Marshall Fund. And we have Norman Schulz, Consulate at the Coordination Staff, AI and Digital Technologies at the German Foreign Office. And to kick off the conversation today, so we will cover both AI as a topic and as a tool, I would like to first start with a tool. So going to Raphael, who is a Data Scientist, how do you use AI in the Foreign Office? And I also know that you have data labs, data and AI labs in the Foreign Office. So could you maybe share a little bit of your day -to -day work?

And yeah, actually, how could AI be used in diplomacy?

Yeah, thanks so much. Yeah, maybe to get to take a step back and answer the question, how like someone like me as a as a data scientist by training ends up in a foreign ministry. I think that’s something that at least when we talk to colleagues around the world is still rather rare. We had kind of the lucky coincidence that I think in 2021, the German government decided to start data labs in all of its federal ministries. And so in the coming years since then, 2022 or until 2022, kind of 16 data labs have been founded in the German federal government. And I was lucky enough to be part of the one in the German federal foreign office. Yeah.

And I was working on AI ever since we started more on traditional data science, I would say. So tearing down data silos between governments or government institutions, in Germany and, of course… And ever since JetGPT and the AI revolution, we have been working mostly on AI tools. And I think the big advantage that we see is that we are in the ministry itself and have very, very short contacts and short paths to our colleagues who are working in Berlin and, of course, all around the world. And I believe in a field that is as fast moving as AI, that is so important because it doesn’t really work to develop these tools in sort of a traditional IT way of doing things, right?

We used to have IT development projects that take two years, have huge teams, cost a lot of money, but that are just not fast enough to deliver on an AI solution that our colleagues, our colleagues are experiencing themselves in their private lives, right? And some of them even… some official aspects. So what we think is the big advantage that we have and what we kind of from our experience would always advertise for is kind of this fast co -creation from within an organization. And I think that is for a topic like diplomacy that is the best way of leveraging AI. And I’m happy to go into more detail about that.

Thank you. And I will later ask you more on concrete use cases. But first, I would like to switch to the geopolitical dimension. So Sharini, switching over to you. Taking a step back, AI is now in the political landscape more or less present everywhere. From the Arctis, but also here at the Summit. Can you give us a broader picture? How is AI shaping diplomacy or foreign policy in general? What is the debate and where are we at the moment?

Thank you. Thank you for the question and also the invitation to be here, which is actually also me being in my home country. So you’ve invited me to my home country, which is an interesting space to be in. But at the broader sort of geopolitical level, AI is shaping not only sort of how we use technology in our strategic communication as countries as well, but as a tool of technology diplomacy. And I don’t necessarily think this is particularly new. Throughout the history of international relations and foreign policy, technology has always shaped our foreign policy. So this is the AI revolution. But if we take it back to the Industrial Revolution, if we take it back to the nuclear revolution, if we take it back to the space race, technology has always informed diplomacy.

And today it is artificial intelligence. So the technology is not new. Yes. But the tactics aren’t. And today we are here at the AI Summit, and this is also India’s way of communicating that it is being a part of a particular technological revolution, which in its previous histories, because of colonial encounters and things, we’ve been excluded. So in this space, this is a way in which countries from our parts of the world are also trying to kind of claim a space in global technology diplomacy. And this is through AI. And what I would also kind of want to just qualify is what we’re seeing in this particular sort of AI race is narratives of competition.

So if you look at sort of policy documents coming out of the United States, coming out of China, there’s a clear connection between kind of winning an AI race or securing leadership in artificial intelligence. And if you are a country of that size and you are the country that has, invented the frontier technology and you’ve been sort of the first movers in that. if a kind of geopolitical leverage which countries like Germany and India perhaps don’t have because we aren’t at that frontier capability but that being said we’re not powerless we just have a different form of power expressing power and that is when the entire middle power conversation comes into play both India and Germany can see themselves are in fact arguably middle powers and they have different ways of using their specific leverage on an AI value chain as geopolitical leverage so for Germany historically this has been through rules and through regulation and regulatory power for India now it is making a case for applications so India and we’ve seen the fact that the summit has changed from the AI action summit which was the French presidency now is seeing India framing it as the impact summit the slogans of the summit are very very much to do with aspirations to deployment or aspirations to impact.

So that is really a way in which a middle power like India is also trying to kind of claim its position on the stack. So what you’re seeing are the great powers who are competing at the frontier level, and then there are middle powers who are claiming their specific power on the value chain in different ways. And I’ll stop there for a second.

Thank you very much. And I would like to pick up this statement that you said. But the tech is new, but the tactics aren’t. So I have here a diplomat sitting next to me. Would you agree with the statement? And how do you govern AI in the Ministry of Foreign Affairs? And would you say, is this still the right approach to AI?

Oh, well, the short answer would be no. But the topic is so broad that obviously I could give you a four -hour talk about it. But as a diplomat, as you said, one has to start by saying that the AI Impact Summit here in Delhi, where we are all gathered, showcases the broad variety of AI and the broad picture that AI is now part of every day’s life, of all strands of life. That it is a tool in communication. It is a tool in agriculture, in industrial entrepreneurship, in finance, and also in diplomacy and foreign policy. So I find that very interesting what you alluded to, that we have these revolutions all the time. like the Industrial Revolution, like the nuclear revolution after the Second World War?

And where do the foreign ministries, where do foreign policy comes in? I mean, the technological revolution created frontrunners like the UK, maybe a little bit like France. But there was a point in time when people saw that only being at the front and adapting the frontier models is not the way to success. But we have to find a way to regulate things because otherwise people will lose their lives. It’s not work safe. It’s polluting the environment. Even back then, there was a problem. Nuclear power, the same thing. There was a race in the 50s. And the Cuba crisis beginning at the 60s showed to the world that the nuclear race could not go on like it was.

But we need international cooperation to mitigate somehow the risks of it. And I think AI is at a similar point. Maybe it needs a couple of more years when the U .S. and China will actively come together and work out what limitations and regulations we have to put on the technology because the risks in the end are outweighing possible and potential benefits. And the other great question is where do the middle powers come in? And this is what India and Germany are talking about. Well, we had the speech of Mark Carney, the Canadian Prime Minister in Davos, where he actively called for the middle power cooperation. And he said, well, we don’t have the power to do that.

I think India is at a one. wonderful place because you are a digital powerhouse and you have all the structures and all the workforce to also become an AI powerhouse. I would also make the case that Germany has also some advantages. We have infrastructure, we have the money to invest into AI, and we also have industrial data to be a frontrunner. Even if we didn’t succeed at the stage of large language models, maybe when it comes to robotics and embodied AI, Germany will still have a role to play. And obviously we at the Foreign Office are there to accompany the development of this and to prepare. Prepare the ground for international cooperation. And I believe it at that because others…

thank you thank you i would like to turn now to the printing perspective um the pranav institute works at the intersection of emerging technology public policy and society from an india first perspective um how do you see um how do you see potential room for cooperation between india and germany like we hear now the middle power those are middle power i hear a lot at the summit that india is leading in a ai adoption um i wouldn’t say so maybe in germany maybe my german colleagues would agree or disagree with me but from your perspective where do you see cooperation like potential cooperation could you also go a step back and um explain to the audience where you see india at the moment maybe also in light of the ai summit

yeah can you hear me i think that’s a very challenging question to answer. Where is India at? India is at a very interesting place, certainly. India is not lagging behind. India is not yet at a place where we can build frontier models. I think the infrastructure capacity for that is very high. I do see some interesting innovation coming out of India. When we saw those 14 models that was released over 14 days and very, very interesting in the sense that this is innovation which is grounded, contextual. It is coming from the grassroots. You are able to find native language use cases. You are able to do inference at scale at much an order of magnitude cheaper costs.

So, you are seeing technical innovation which is more context appropriate coming from India. There is, of course, a large workforce which is talented in technology and there is an upscaling possibility that certainly exists when into AI and that is a very large pipeline. So I think India is a very interesting place. India is adopting, India is innovating, India is building applications and use cases, which is a very useful way to think about the technology in its early stages, right? Because there is a huge possibility of investment booms and busts that can come in when you go in a technologically challenging direction without being adaptive. So I think the focus on saying what can we solve is a very useful way to think.

I think the counselor did allude to industrial AI. That’s a fantastic use case of cooperation where you and India could possibly, Indo -German cooperation would certainly work out in that sense because there is industrial expertise, there is automation expertise in Germany, there is industrial data. India has the capability to build technology, build models. So I think if we were to identify and not worry about the race for frontier models, because transformers are not going to be the same. They’re not going to be the only technology paradigm out there and not play the game that leading powers are, but to really think as middle powers do as Sharon said and say that can we focus on sectoral expertise?

For example, AI in healthcare is a fantastic opportunity for. Indo -German cooperation, there is fantastic data available. India performs 10 times the number of surgeries that other countries do. So there’s very interesting data available. Germany has the capacity to invest. Can we cooperate? Germany has expertise in automation. India has, you know, people who can build AI models. Can we cooperate? So I think there is possibility for bilateral cooperation that, you know, gives an argument that is more than one plus one in the case of some of these. And I don’t think it’s a zero -sum game that U .S. is winning or China is winning and they’re all left behind. I think the focus on applications is really where a differentiator is possible, and that need not come at frontier -level costs.

Thank you. And I would like to focus now on this application side because this is maybe the way to react to big tech or like us as a country being in the middle between these mentioned countries. Rafa, can I hand over to you to share a little bit how you have the foreign office approach? I know that you are working on it. It’s a negotiation tool. And to what extent can open source also be a solution or might be a solution? to the situation where we are at the moment.

Sure. Yeah, so I think it’s exactly as you said, that the focus is on application. We have made a consequential, but I think important decision at the beginning that when we are implementing AI, we are focusing for most of what we do on open source technologies, not just the models themselves, but also a lot of the kind of scaffolding and applications around it. So on the one hand, for example, we are reusing applications that, for example, come from one of our state governments who have done like kind of a general chat and knowledge -based application that we are reusing. But of course, we have specific applications in the foreign office like supporting negotiations. A lot of what diplomats nowadays do is not necessarily sitting in rooms and negotiating face -to -face, but actually digging through huge piles of documents and…

trying to understand the positions of other countries, the impact that NGOs, academia, corporations bring into huge negotiation processes. And, of course, that’s, as we probably all know, is a great chance for artificial intelligence to leverage. I think one important point when we’re talking about AI and open source AI in governments is that we have seen a big trend shift or a shift in the trend last year where we have seen that a lot of the kind of leading open source AI models and actually also the ones that have been adopted in many parts of the world are coming from China nowadays. I think that’s an interesting intersection between my position as a technical observer here where we are looking at the numbers and seeing that really, like, you know, the world is adopting Chinese AI models at the moment.

And, of course, the consequences that that might bring for a country like Germany or a country like Europe. Like India on a global scale, if maybe… some of our partners are implementing Chinese AI models. So that is something that when it comes to open source, I think it’s really important that countries like India, and I think India is at a great position, and I’m super excited to see these new Indian AI models as well, these Indian LLMs, to see if there can be pushes that offer alternatives to these Chinese models.

Thank you. I would like to come back to this impact aspect. Now we heard impact in the public sector, but maybe also reflecting on the summit, AI Impact Summit. What are your thoughts on how will we now continue the conversation regarding impact, regarding really being concrete and not only writing governance formats or governance frameworks, so how can we make this cooperation very concrete and also continue where we are and face this geopolitical challenge

Shoni, yeah I like that all the geopolitical questions then somehow come back to me but I don’t blame you because my background is in international relations so that serves very well this purpose but I want to kind of also connect your point to what you just said about open source and the China connection I think we’re reaching a stage in international relations in which geopolitics and technology can’t be separated when we are integrating artificial intelligence into our daily life and into our government systems we can’t really separate the security risks that come with it And I think every country has a unique security situation. For Germany, obviously, there is the concern with Ukraine. With India, we have border security challenges as well.

We have territorial disputes that are very significant and have very serious national security implications. So the kind of technology we deploy into our systems, and if it’s open source Chinese models or any other form where we perceive or any country would perceive a national security risk, that needs to be factored in. And that is why even in our technology decisions, they have to factor geopolitical risk, which back in the day was not something that, say, companies would have to do. But now every single company that I see has now a position for a geopolitical risk advisor. And that really comes from the fact that we are living in a world in which if we are using technologies so seriously in our lives, lives, we do need to factor in how those technologies can be weaponized in a particular geopolitical situation.

And then that kind of brings me back to also some of the points that were on, you said, you know, where we, you, foreign office would like find it helpful for reports to be kind of processed to AI. As a think tank, I think I’m a little bit hurt, I have to say, because a lot of our work is producing a lot of those reports, but we will force you to read them. We’re very persistent at the German Marshall Fund. We will reach out and invite you and make you read them. But jokes aside, it is really, we’re aware also that our ability to consume information as well is kind of becoming shorter, but the world is getting more complex.

And therefore, we are also kind of preparing, even in the think tanking that we do, even in the way in which we kind of do our daily jobs, to factor in that. There will be an AI in this system, and we kind of need to put that into consideration as well.

and since there will be an AI in the system we have to make sure that we can trust this AI and that it’s also inclusive and that it’s yeah ethical in a sense or trustworthy regarding to standards so how do you and the government react to this could you also share more about the global digital compact and what is this panel about this scientific panel I think it’s called and how do we make sure that from this governance it goes to the system to the AI system like how do we make sure that the systems are aligned with our values

well that’s big question the best way to align the systems with our values is to develop to develop them ourselves right and not just procure them from from outside and I couldn’t agree more with the point that you made about the Chinese models, that even if it is open source, even if it runs on our servers, there are still Chinese models. They still have the Chinese ways and the Chinese ways of thinking, which comes through maybe not all the time. So using AI to do diplomatic work will not be the way because then every report will be the same, right? So I hope that Germany will not go the way to write the diplomatic reports now only using AI or summarizing it.

But we need our diplomats to insert that innovative thinking. And innovative thinking does not come from AI. Because AI… AI is much rather replicating, summarizing, in my understanding. The new ideas still come from the human side. As far as I make it out. Global digital compact. Thanks for the question. The Foreign Office was the lead in Germany to negotiate the global digital compact. And obviously you can make a point that this is a UN compact and the UN system is under immense pressure at the moment. So what does it achieve? And I would make the point that despite all that, it has at least produced two valuable avenues for future cooperation and discussion, two platforms.

The first is the AI panel. I think it’s called Independent Scientific International Panel on AI, but I could be wrong with the two I’s. It’s rather complicated. But it was just yesterday that the UN Secretary General made the point that the AI panel and the second one, the dialogue I will come to in a second, are the two major things where the UN is coming into the picture. And the panel has the task to put our discussions that we have on a global level about AI on a scientific basis. So those are experts, and I’m happy that there are two experts from Germany on the panel. Only the U .S. and China have also two experts.

I’m terribly sorry. I don’t know how many Indian experts are on the panel. But we’ll find. We’ll find that out. True. So they will produce a first report, a summary of where the AI science is now standing in time for the first global dialogue on AI governance, which will happen in July in Geneva in the margins or back to back with the AI for Good Summit at the International Telecommunications Union. And this dialogue serves the other big purpose of the Global Digital Compact, which is to make the AI discussion inclusive. And so it’s also the UN Secretary General nonetheless that said that AI cannot be a discussion among the few, the ones that are the front runners like the US and China.

They should not be the one to they should not. Not be the only one to set the rules, but it has to be a truly inclusive discussion about the AI. Up until now, more than 100 countries were not part of this discussion because they were not members of the European Union, not members of the Council of Europe, not members of the G7 or the G20. But they are the ones that will use AI, that will adapt AI, and they will also feel the bad results if AI is not doing what it is supposed to do. So it’s good that they have a voice at the table, that all UN member states will in July come together and talk about AI on the scientific base that the panel has provided.

So that is something that the Global Literature Compact is doing. And, of course, we can talk about geopolitics all the time, but I think that’s a way forward. And it’s to make a point. And I stop here and make a point.

Thank you. Thank you. And, Jian, let me turn to you now. And the fact that there is not a zero -sum game in a lot of this. I think the idea that we can work together to bring a larger voice beyond the worries of two countries or three countries which are able to compete at the top. I think that’s something that they shared. And I think the role of middle powers in bringing a more inclusive conversation is really important. And I think Indo -German cooperation is an opportunity for that. Including, for example, industrial AI that the counsellor mentioned or other opportunities where we can practically create tools like what Rafael is also talking about. Where we can practically create tools that are beneficial and maybe open source.

Why should open source models only come from strategically challenging sources? There could be Indo -German open source models, smaller models, not frontier models that could be beneficial.

Yeah, I can maybe react to that directly because I think it’s super critical. I don’t think some people believe or make us believe that the AI race is already over or kind of only being decided between the US and China. I don’t believe that. I think we’re more at the start of what’s going to come. And I think we can feel this at the summit. And Gunda, you asked the question, what comes next? I think next comes building and implementing AI in all these kind of fields that we have. I think we see so many ideas around here and first steps towards that. But we don’t really see widespread AI adoption in every field, in every kind of part of life.

I do think this is going to happen over the next five years. And I don’t believe for a second that this is only going to be done by the U .S. or China. And, yeah, I think that when it comes to middle powers, Germany, India, I think we are going to see much closer collaboration in like smaller groups that don’t try to kind of, you know, build dependence, right, making you dependent on us, making us dependent on you, but rather ensure that every country can bring to the table what they are particularly good at and make the results kind of improve the application of AI for everybody involved. I do think there is a strategy for that.

And I think, yeah, the way forward you have asked is to start with it and to build AI together. I think this is a great, you know, a great rally cry for

Yeah, yeah, please.

Rafal, you led me on to a very interesting trail, so I had to intervene. I think one of the interesting moments, if you think about technology again, in the 1990s was the open source revolution, right? And when you really saw operating systems, consider the frontier technology of that time being built by volunteers at a fraction of the cost. diffuse the race in a certain way or diffuse the dominance in a certain way, but also enabled accessibility across the world. So I think even coming together as middle powers, the power of open source and democratizing and reducing the factor costs of access to AI, it becomes very powerful if you draw from it. And now you led me on to a trail as well.

I just want to kind of contextualize the sovereignty thing as well.

And I do think that when we talk about artificial intelligence, it’s not just one application that we see when we use our phones or interact with a particular model, right? It’s an entire stack. And the question of sovereignty or the concerns vis -a -vis the sovereignty debate is also born out of geopolitics, right? So we don’t want to be as a particular country in which suddenly one day we wake up and our technology is not available to us because, because of something else that happened in another country. corner of the world. So the sovereignty debate is coming out of geopolitics as well. That being said, we don’t need to be beholden to it. I fully agree.

And I really like the point on us understanding what our strengths are. I mean, Germany had a high -tech strategy that came out last year. There’s also an emphasis on Germany being a space for data as a data hub. And India is trying to do that as well. Germany already has that. One of the things China is really good at actually is industrial data, because they have been collecting this data for a very long time because they automated quicker than a lot of us. And that’s something where we can collectively build competitiveness. So I do think we need to reset some of the inequalities in the AI stack and that sovereignty, as much as I kind of understand where that comes from, I don’t always think that that’s the…

best language to talk about where we are at. I do think we need more sophisticated and nuanced ways of kind of talking about a managed interdependence where I have a certain value on an AI stack. That is my strength. And the likelihood of you weaponizing that makes it very limited. So that’s why I have leverage. And I do think leveraging a country’s strength on a specific AI stack is a prominent and powerful middle power strategy.

Thank you. These are all beautiful closing remarks, but I would like to open the floor to the audience. Are there any questions? Yes, everyone around.

Hi, I’m Sreeni. I’m a student at Ashoka University. I have a question for everyone in the panel. Feel free to. answer. The question is, what are some parts of foreign policy research, decision making and implementation which can be automated by AI or that will use a significant uses of AI to sort of do your day -to -day tasks?

Maybe I can quickly answer this question. Well, I certainly don’t think that AI will make any decision in any time soon. So there’s always going to be the human that is making the decision. And it’s not going to be me. It’s not going to be my boss. It’s going to be a collaborative decision by the government and the legislature and all of that. But our job will also not go away. We will use AI to make our job easier to consume data, to consume, I would like to say information, but it’s nothing. Consuming information easier, quicker and which in turn will free diplomats to do the other time, which is connecting, which is connecting the dots, which is thinking out innovative ways of cooperation, which is, it’s basically like drinking coffee and shaking hands.

These start traveling to India and learn a lot about the situation here. So, AI will free us from tedious tasks of skipping through these very valuable documents written by not only NGOs, but also governments, and will make our lives easier, but our work will not go away. Thank you.

Okay, one more question. The lady maybe in the back.

Hi, I’m Sanjeevni, and I work in radio journalism in the UK. And my question was for you, Norman. So, specifically, So, Norman and Sharini, both of you, actually. So, you guys are doing your Masters in Journalism. I was studying about how journalists were framing the Russia -Ukraine war. And we were observing how the narratives were changing based on different outlets. But something when we were talking about how AI is coming into play, do you think AI will help change narratives for the better? And I’m not speaking from journalism point of view. In general, geopolitically, do you think the narratives will be framed in a way that’s unbiased? Or how do you think it will help in that?

I don’t take a stab. I’d be very curious to point a question, actually, to you, because one of the things I’ve been really intrigued in in my line of work is how AI has been being deployed in the media and newsrooms. And I’d be very interested to have a chat after to learn how you’re doing that in terms of methodology. But to your question on AI shaping, I think it’s a great question. narratives, I would not let AI shape narratives. I would hope we shape narratives as human beings, depending on sort of what we think and feel and analyze through empirical evidence about the world. And I would be very worried in a world in which AI, we allow AI the space to shape narratives, especially on geopolitics, because then that would depend upon what that particular AI model that is doing the narrative shaping has been trained on.

But that being said, I would also see how AI can do the harm in terms of amplifying incorrect narratives or geopolitically challenging narratives. And that’s when we know that AI cannot replace society and it cannot do. So I do think that we are in a world in which if AI, we allow AI to shape narratives, that’s not a world we want to live in. Thank you. but at the same time, if it is a world in which that can happen, we need to find the right mitigation strategies to do that. One thing that I know India is doing, Shyam, we talked about it earlier, which is these bias detection technologies that are critical. AI is a technology that, on one hand, we do need strong regulation to make sure that we can prevent the harms from doing what they can, but at the same time, we need also technological tools to deal with some of those harms and push a democratic innovation as well in AI for exactly these harms.

And I’ll stop there.

Just one sentence. If you let AI write the newspapers, they are becoming incredibly dull because it’s going to be repetitive all the time. But I agree with your point about bias. This is something that we all have to challenge and to face. And AI is helping us. That’s a good thing. AI is not only a risk, it’s also the opportunity, helping us detect bias and then contravene it. Thank you.

Just one sentence I want to add because I found your point so important. I don’t see any risk that AI is going to shape the narratives itself somehow, but of course it’s an incredible tool for actors trying to shape narratives. And we have seen this on so many fronts already. We have colleagues in the foreign office who are actually monitoring this and seeing that it’s, for example, used to amplify certain messages across social media, increasingly now across faked websites that, with the help of AI, you can pull up in seconds and suddenly you don’t have one of them or two of them, but you have thousands of them. And that is something that AI has already used quite heavily as a tool of certain actors who are trying to influence geopolitical discussions in exactly that way.

Sorry, we are running out of time, but I’m sure that the speakers stay here a little bit. So thank you for listening to our panel discussion. Maybe a big applause. Thank you.