Katherine (Kate) Warren (3:05)

So I've heard This described as a big challenge for AI right now. And I wonder if you agree that we're missing a single operating system for AI governance. Is that the goal?

Amandeep Singh Gill (3:18)

No, I don't think so. I mean, this was examined by the high level advisory body on AI and its conclusion was that we cannot have a single operating system today. But that said, from a UN perspective, we have the norms that could inform multiple operating systems which are connected together or eventually, if you need a single operating system for some parts of the AI governance challenge and these norms, the human rights norms, solidarity in terms of humanitarian action, the sustainable development agenda that we have, the agreements around that, I think those will be important regardless of the open source, the operating system we

Katherine (Kate) Warren (4:03)

end up, that's really a helpful framework to hear you say we need multiple operating systems. And yet I'd love to dig into a little bit more. What is the UN's role when it comes to those multiple operating systems? Who has the convening power to get the kind of alignment that's needed, even if we have multiple systems? And where does the UN fall into that puzzle?

Amandeep Singh Gill (4:26)

Well, the UN is the most inclusive body on earth in terms of bringing member states together and with an increasingly sophisticated multi stakeholder component to those interactions. I think AI has helped advance multi stakeholder participation in UN's discussions, which have been heavily intergovernmental so far. So I think it is the indispensable place where AI conversations should be happening on governance, on science, on capacity building and applications for development. But that said, you have these other places you mentioned, the India Summit. So this is, from our perspective, very complementary. The work that OECD has done, the Global Partnership on AI, some of the other frameworks that have been explored, and especially regional frameworks, the ASEAN Blueprint, for instance, or what the African Union is doing with the Continent Wide Free Trade Agreement and its Digital Trade Protocol. So we need to ensure that there is more connectivity and complementarity across these different initiatives, while preserving the central role of the un, mainly because of that connection to our shared norms that I mentioned.

Katherine (Kate) Warren (5:48)

Fantastic. Just one follow up there. I was interested in hearing more about when you said AI has helped to advance multi stakeholder participation in the un. Intuitively I get why that is, but can you expand on that?

Amandeep Singh Gill (5:58)

So we've had the Internet Governance Forum, for example, for 20 plus years, which has been a successful experiment in getting private sector civil society, the tech community, because that's a separate set of stakeholders into this. Now with AI, you see that a lot of the research, a lot of the applications are being done in the private sector and you have the role that like akin to the technical community in the Internet context that independent scientists, you know, and the Turing Award winners that have been very active on this, have played, the open source community have played on this and of course the very strong role of civil society in highlighting risks, including risks of exclusion, discrimination, bias, etc. So, so without these important stakeholders, the intergovernmental conversation would be dry, not substantive enough and not impactful enough. So we need the perspectives of these communities, but also we cannot implement what is discussed in international forums without their full participation. So take the high level advisory body on AI, so 39 experts and many of them connected to the private sector and now the international independent scientific panel, 40 experts, some of them with connections to private sector. We found a way to manage the conflicts of interest, manage different competing interests and bring diverse multi stakeholder voices into these conversations. And frankly that's what has helped member states develop the confidence that the UN can actually play a role on this. You know, without this won't wouldn't have been possible.

Katherine (Kate) Warren (7:40)

I actually want to ask you about one of these actors. When you talk about the importance of multi stakeholder collaboration. You mentioned the tech community specifically I do want to talk about AI labs. So you know, there are these companies like Anthropic OpenAI. What I'm seeing and what I'd love to follow @Devex is they are really becoming de facto development actors, really new players in global development. You're seeing partnerships formed between of these traditional development entities and these AI labs. And I wonder how you would advise the global development community to think about these companies as new players in global development, because I know you've also engaged them in your work.

Amandeep Singh Gill (8:24)

That's a very interesting framing, Katherine. So in our dialogue with these labs and at the Office for Digital and Emerging Technologies, we talk to them regularly. I see a lot of interest in our work not only in terms of the governance side, which may have regulatory implications. So they all, you know, want to be part of that conversation, but also in terms of how the world is going to apply and absorb AI. So interest, for example, in more diverse data sets, linguistic data sets in particular being developed. So this is where, you know, our interests meet. For the tech sector it's about, you know, their products being more locally relevant and for us it's about the diversity of the world. So I see these pragmatic connections being explored, talent development for example. But I think where we have to be careful and the development community has to Be careful is, you know, obviously there's an interest in moving fast. You know, there's a arms race going on among these labs also. They want to beat the other to the finish line. So there's money involved, huge amounts of money. So sometimes the fundamental understanding about development is not there. You know, to paraphrase Amrita Sen, it's about human empowerment at the end of the day, people being able to have more choices and agencies over how their life unfolds. That's the deepest meaning of development. So often, you know, in the tech companies thinking because they make products that address specific pain points, as they call them in markets now, they think of this as solutions, solutions for solving specific development problems. So we need to, as development actors outside of the private sector ecosystem, we need to have robust conversations with them on this point, equally on the governance side because as we know, technology empowers, but it also disempowers, it includes, it also excludes and there are second order, third order consequences, externalities as we've called them in the past. We've seen that with the fossil fuel industry, with other sectors. So we need to again continue to have those conversations. The energy footprint, water resources, you know, the hate speech in terms of, you know, women online, some of the objectification of women and you know, the risks for children and their well being. So I think those, that's the kind of, I mean we could go into

Katherine (Kate) Warren (11:04)

this more but you know, it's a fascinating topic. I did want to ask in terms of what accountability mechanisms might exist because you know, as you mentioned, these are, these are companies, there's money involved, you know, they have their interests. I think about one very interesting structure I've seen which is, you know, anthropic, as you know, has this long term benefit trust. So that's this governance structure designed to ensure the company prioritizes, as they put it, humanity's long term safety and benefit over pure profit. So that's one example. What other potential accountability mechanisms do you see?

Amandeep Singh Gill (11:40)

I think we're talking about accountability mechanisms inside. So for now it is at a very preliminary stage. You know, we've seen oversight boards in the past, they've been post hoc, you know, problems have surfaced and then you put them in place. I haven't seen that much proactive action so far. It's encouraging what you've described and some others and the industry associations, the partnership on AI and others that are happening. But I think what we need to do, and this is one of our goals for this year, is to work with the private sector to understand how they are applying AI governance on a day to day basis, how they are creating facts on the ground, like in customary law type stacking of, you know, actions. You know, sometimes the national legislation is missing, the executive orders are missing. So they, through their daily decision, they are stacking up something. What's that something? And how can that be better connected to best experiences, best experiences around, best practices around the world. So I think that's a key point in terms of, you know, how we engage with these labs on accountability and governance. The other thing, and I think that's a broader point also in terms of the development implications. I think we need more diverse participation in innovation. Innovation, particularly with a development perspective, cannot be driven from a few geographies because at the end of the day it's a human artifact. Technology and humans have their own, know, context, biases, frames of mind. And we could go into a fascinating rabbit hole on how sci fi and, you know, pop art has influenced the thinking of, you know, the tech labs in the Silicon Valley. But that's not the case in Kenya or in India or in, you know, Guatemala. So how do we kind of diversify? I used to speak of, you know, scale. I mean, the tech industry is very interested in scale. All things scale, including the COVID virus, you know, it's scaled very rapidly, but in nature, if you look at scale, forests scale, but monoculture doesn't scale. So what's the kind of, you know, sweet spot between, you know, these highly verticalized, concentrated centers of innovation and a more diverse ecosystem that responds to the messy reality of development across the world?

Kate Warren (14:28)

Hi, I'm Kate Warren, Executive Vice President and Executive Editor at devex. At devex, we don't just cover the biggest moments in global development. We create space to understand who and, and what are driving the headlines. Alongside gatherings like the World bank and IMF Spring and annual meetings, the World Health assembly, the UN General assembly and beyond, we host Devex Impact House, where our journalism comes off the page and onto the stage. We bring together a curated group of leaders for live interviews, intimate roundtables, hands on workshops and candid conversations you won't hear in the official, official meetings. That's where tough questions get asked, the spin gets stripped away and meaningful connections happen. If you'd like to join us or stay in the loop on all of our events online and in person, please visit devex.com events I totally agree about

Katherine (Kate) Warren (15:21)

the need for more diverse participation in innovation broadly, including AI, of course. And I do think that the India AI impact Summit presents an opportunity here. So I'd love to hear you talk a little bit more about that. I think you know, of course, we both know, but perhaps you can expand on it. India has been a real champion of digital public infrastructure, a real pioneer of digital public infrastructure. A big conversation right now in AI is in terms of the role of low and middle income countries being not just recipients of AI, but really being in the driver's seat. And I think India has demonstrated what that can look like. So would love to hear from your perspective, how do you see the India AI Impact Summit as an opportunity to showcase lessons from India as well as other low and middle income countries? And can this really be a pivotal moment in terms of what you were just urging, the need for more diverse participation in innovation?

Amandeep Singh Gill (16:22)

I think the India's AI Impact Summit is a great crucial bridge to the global mechanisms that are kicking in. The science panel, the regular policy dialogue. It's also a crucial bridge to the Global South's aspirations. On how we address the AI divide, the desire, the intense feeling that we don't want to be left out of this technology leap, that we missed the first industrial revolution, the second one, the third one, don't want to miss this one if we want to take that kind of simplified framing. So I see India with its successful experience of digital public infrastructure as a story on building public trust in this technology. In the 1980s, when the railway reservation system was digitized in India, public trust started to grow. Less corruption, more transparency, just ease of doing things. And the DPI wave, the India stack has kind of taken that even further. So there is anticipation about what AI can do for social and economic good. So as we were talking about pain points earlier, which are important for the private sector, but also in the development space, you know, the sheer variety of use cases that you see from health, you know, detecting cervical cancer early or you know, addressing diagnostics inequity in terms of, you know, providing AI based diagnosis of diabetic retinopathy. The sheer number of these applications is huge. So it's a huge lab for experimenting at scale with AI for social good. So I mean, that's exciting, but it's also frightening because then you have to do it carefully. So I see the summit as an opportunity for that conversation and for reinforcing some of the discussion that have been happening at the UN on addressing the AI divide, on building capacity, a minimum irreducible capacity across all nations in terms of at least having local language data sets, minimum amount of compute to work with those data Sets a minimum amount of talent. Take an island in the Pacific Ocean, for example. You know, you have climate change resilience challenges, natural disasters and shrinking linguistic diversity. So, I mean, we can't just plug them into a chatbot somewhere and hope to solve some of those problems. How do we land this? What's the kind of capacity every country needs? That's something that we've been debating at the UN post the Secretary General's report on innovative voluntary financing options for AI capacity building. And I see the India Summit as an important turning point to move that conversation forward.

Katherine (Kate) Warren (19:36)

I'd be curious to hear your thoughts on this, because I do think there's a lot about India's own journey that other countries might try and replicate. But of course every context is different, as you were mentioning, and I think many countries will take a very different path. In some recent reporting that I did, for example, one of the experts I spoke with talked about in terms of lessons learned from India of what perhaps to do differently, say on DPI and AI infrastructure on the African continent. There was. He basically urged more involvement from the private sector, bringing the private sector to the table. You mentioned trust earlier. Trust is such an important pillar of all of this and that looks very, very different country by country. So we often hear that India is a pioneer, and it is, and there are many lessons that can be replicated. But what do you think is likely to look different for many other countries? I suppose I'd be particularly interested in hearing about this on the African continent.

Amandeep Singh Gill (20:39)

That's a great question, Katherine, and something we've been thinking about for the past one year plus in the context of DPIs. So through experimentation, India landed on its own vision of DPIs, which was informed by the practice in Estonia, you know, exroad digital interoperability, data exchange and some other experimentation. I think Africa would find its own vision of digital public infrastructure. And my guess is, and it's a hunch, and we've been exploring this hunch with some African leaders, is that this will have more to do with intra regional trade. In India it was more about inclusion in terms of participation in the financial sector, etc. But in Africa, it might be more about how do you build scale for the digital entrepreneurs across the continent so that, you know, the. The intra regional trade grows, regional integration grows, jobs can be retained on the continent. So you are at an early stage on DPIs in Africa, digital ID, the payment side, largely private sector driven in West Africa, for example. Now, how do you create interoperability there so that an Entrepreneur sitting in Ghana can access the market in South Africa and thereby the continent grows the private sector engagement part. I think that's a big lesson again from India and the experience in Estonia and other places that you need to have a startup ecosystem around these applications driven by digital public infrastructure. So government services, for example, and this is not a Global south issue only. I've seen it in Germany with the GovTech initiative. Officials are reluctant to talk about their problems, et cetera, but you create the right space and they start sitting with startups and talking about, oh, I'm having this problem with permits for power companies, et cetera. And these startups are more agile and they say, okay, let's have a look at it, et cetera. So I think in the African context as well, if we can engage the African startups, smaller companies, SMEs on how to solve some of these problems, then you build DPIs while growing the digital economy with a private sector led approach. So those are kind of, you know, a couple of thoughts on how this could happen differently on the continent.

Katherine (Kate) Warren (23:23)

So, Ambassador Gill, in terms of private sector engagement, you mentioned earlier the speed at which some of these tech companies move, and you often hear in Silicon Valley the phrase move fast and break things. So there is this tension between speed and safety, which of course is front and center in the AI conversation and requires consultation and consent and many of the efforts that you've been involved in. So as you try and bring private sector leaders to the table, how do you design governance that protects the public interests without stalling innovation, which I think is sometimes their fear.

Amandeep Singh Gill (24:05)

Yes. So I think it's a challenge for national governments, partly and mostly, but it's also a challenge for the global governance piece. There is, of course, you know, there's a lot of understanding in the private sector, which is often lacking in the public sector, including in diplomatic communities. So we need to kind of address this information asymmetry by building up our own understanding, our own knowledge, not only of what works with this technology, but more importantly, what does not work, separating the hype from the reality. So that's why the importance of the International Independent Science Panel and also within national jurisdictions. You know, I've heard of leaders who spend time themselves with AI developers to understand what does it mean, et cetera. I think equally important would be what some call development of public technology, but in this case it's open source. So open source has its own risks, I'll be the first to admit that, and we need to address those. I mean, look at what's happened with Malt book the Maltese hype and drama to some extent, but we still need public tech so that all tech development does not end up only in the private sector. Academia also is stretched in terms of its access to compute, in terms of resources and the fact that it's leaking talent to the private sector. So I would say, you know, the information asymmetries the development of public tech. The third important issue is how we design governance. I mean it could be easy to just sit down with the private sector, tell us how should we do governance and you know, the common denominator there is going to be that less is better. You know, don't bother, we are very responsible, we'll take care of things, et cetera, self regulation, et cetera. But I think in the design of these things at the UN I mentioned this multi stakeholder perspective coming in this bidirectionality, but also some more concrete discussion in terms of how are you landing governance so that we don't do things post hoc, we anticipate, et cetera. So equally we need to do that at the national level, you know, find ways Agile governance, public private sector, participating in that, informing it, but still with the public interest in mind. And that's the responsibility of leaders who are elected in democracies, you know, with the public interest as the driving force.

Katherine (Kate) Warren (27:05)

So I do want to ask you specifically about this new international scientific panel on AI. We've, we've heard from you about it a few times and you know, There are these 40 experts that the UN Secretary General has recommended from across different sectors and geographies, as you mentioned earlier, the importance of that and this panel is meant to build shared understandings and guardrails at this moment, when as the Secretary General put it, AI is moving at the speed of light. But what I wonder. I do think there's some skepticism from the DEVEX readership and the broader community at, you know, various panels and reports and convenings and kind of back to the question I started with. It's hard to kind of tell how, how they all connect and what they will really produce. So how do you think that this panel's recommendations might actually shape real world decisions and kind of avoid this scenario where it's an expert group that produces reports that sit on a shelf.

Amandeep Singh Gill (28:03)

I think that's a great question. And this was a design consideration in a sense, you know, for the member states negotiating the global Digital Compact or when the high level advisory body on AI gave its seven recommendations, the panel was one of those. Now placing the Panel alongside the Global Dialogue is perhaps the answer to your question. So the first report of the panel in July is going to be through this dialogue. So that the dialogue on AI governance is informed by science and evidence. And I think the other reason why it's different and why it's unique and a first is that this is the first global effort based on clear criteria. There is the modalities that have been decided. The nominations went out 2,700 plus people from around the world. No strict criteria were applied and you have selected 40, gender balance balanced across regions, roughly Global South, Global north, even though Global south is about 19 and Global north is about 21, given the preponderance of the research in the Global North. But this is the most diverse, independent, duly selected body that we have today with a link to this Global Dialogue. So if you look at its remit, it's going to think about, you know, not only risk, but also opportunities. So often we look at, you know, these reports as a kind of, you know, speed breaker. I think governance is not only a break on risk, but it can be, if it's proactive. Science led can be an accelerator of opportunity. So for, for instance, to identify where AI can do the most good in the fastest possible time, or anticipating some of the things that will make development challenging, job displacement, for example, what's coming around the corner. So so far we've only had ad hoc work on this. We've not had a systematic annual reporting, and the panel will also have the opportunity to do thematic reports, you know, let's say a couple of times a year, in addition to its annual report. So. So I think it's a good start in terms of building a common understanding, shared understanding of the opportunities, risks and implications, and having more science informed policy approaches, including in the sense of pursuing where the opportunities are the most exciting or most quickly available.

Katherine (Kate) Warren (30:40)

So another question for you is just about the incredibly difficult context you're operating in. Now. You mentioned at the beginning of our conversation that it's an uphill task to bring the world together to govern this technology for everyone's benefit. And I think that's especially true when the us, which is of course a dominant AI superpower, you have an administration that is wary of being constrained by international rules. So how are you navigating that challenging environment and trying to continue to move the needle on AI governance?

Amandeep Singh Gill (31:18)

I think the Secretary General has been very clear and, you know, we've been following his lead on this. You know, we stay humble. So I think it's important for The UN not to overplay its hand on this issue because this is not a 1, 2 year, 3 year, you know, SG's tenure, my tenure, kind of a problem. You know, it's a long term institutional challenge, not just for the un, for all of us. You know, you mentioned the development, cooperation, ecosystem. You know, this is going to shift how we work. So we are being kind of like, you know, long term in our thinking. We start where we are. So I think that's part of the answer. The second is, you know, working with all perspectives. The U.S. is a very, very important player. And within the U.S. you know, you have academia, civil society, private sector. And I'm very, very happy with the engagement we've had with the US private sector and I dare say also with the administration in terms of the opportunities, the diffusion side of it, and also the interoperability side of the equation. So everyone realizes that if we just, let's say even the hardware around this, if there's only a national market for it, there's a break in terms of what US companies can sell, etc. So I think there is something to work with. But we also need to kind of, you know, be more, let's say, impactful in terms of the outcomes so that we address the skepticism more broadly about multilateral institutions today. Do things with not only less resources but also innovatively and in an agile manner so that, you know, this old trope like, you know, a UN moves too slowly, etc. You know, this technology is moving too fast, so don't even bother. So, you know, I think there are ways in which we can handle these difficult times. And my hope is that as we've kind of now moved into a more of a member state owned approach to AI governance, my hope is that leadership broadens on this issue and many more countries step forward and say we will lead on some aspect of this, etc. We've seen that earlier with the development agenda. I'm sure similar trajectory might be one

Katherine (Kate) Warren (33:56)

thing that strikes me about your background that I see as very relevant to the AI conversation. And I'd love to hear how you connect this and I'm sure it comes up often in your conversation. So you helped build international consensus on regulating AI in lethal autonomous weapon systems. And you also served as India's ambassador to the Conference on Disarmament. So there are calls for red lines to prevent unacceptable AI risk. And we've seen this in weapons governance. And so I wonder if you can just connect those dots and tell us what you see as the Non negotiable in development context. And who gets to decide when it comes to AI and AI safety.

Amandeep Singh Gill (34:37)

Right. No, I start with a caveat. My work currently is about the non military application of course, of technology. But governance I think overall improves life across all our three pillars, Eastern security, development, human rights and humanitarian action. And this is the quintessential multi purpose technology. So, and, and, and you as you cannot count the civilian application, you cannot also count the military application. There's so many ways in which AI is seeping into weapon systems and military broadly, not just, you know, kinetic kill type of systems. So we, and, and it's, it's going to change the role humans play. And the focus in those discussions on lethal autonomous weapons was on human accountability, human control, human involvement in terms of what's there already in international humanitarian law. Something that has taken us, let's say 150 years to build. You know, technology could just sweep that away in no time. So, you know, important conversations to be had across the board in terms of, let's say, take gender on at the other end, when we've taken centuries to build mutual respect into, you know, gender relationships and a lot of other good things which human civilizations would be proud of. But then if technology kind of, you know, takes us back, then we have a problem. So there are, I see these parallels between the governance discussion on the civilian side and on the military side. I also see positive externalities in reinforcing international cooperation. It's like in the nuclear domain, the cooperation around, you know, fundamental physics, particle physics and on civilian applications in agriculture areas, you know, help to kind of bring the world together, even though there were tensions on the other side. And I think we will also see the role that the vast majority who are more interested in addressing the AI divide, leveraging this technology for development play in terms of a break or a kind of a bending of the arc of investment. All the investments go into the military domain or into, you know, let's say technology that disempowers or is addictive. And you know, I mean, think of porn for example, you know, it's all going that side. So if we can bend that arc of investment to this other area, then you know, again, it's a dampener on the nasty competition that could end up happening there. So perhaps these are kind of, you know, more, let's say, aspirational objectives. But we also have, as the SG has mentioned, clearly clear objectives on the military side. He wants a ban on weapons that can take life and death decisions autonomously and I think as we move into the agentic AI conversation even further, you know, I mentioned Molt book earlier. With these agents apparently interacting on a social media platform, we will need to decide, you know, what are the actions that we allow AI to take on our behalf and what are the actions that we exclude from the consideration and even those actions that are being taken on our behalf. How do we handle liability, accountability, etc. Now that's strikingly parallel to the military space.

Katherine (Kate) Warren (38:21)

Absolutely. So just a final question for you, Ambassador Gill. Things are moving very quickly in this space and there are decisions being made every day that will help to determine the future of AI for better or for worse. And so for our listeners, whether it's someone working in a health ministry, a tech company, an ngo, civil society organization, if you wanted to leave our listeners with a key message on the urgency or the actions that need to be taken right now, what would you say?

Amandeep Singh Gill (38:51)

You know, I'm reminded of what Mahatma Gandhi used to say, that whenever I'm thinking about something to do, I think of the most wretched person that I've mentioned, someone who has nothing, et cetera, and ask myself, does that help that person? So if all those who are taking decision on a day to day basis can just step out for a millisecond out of their current rush to get to things, how does this benefit humanity at large? And in particular, because human beings are emotional people. So just to ask ourselves, you know, what kind of world I want to leave to my children and to my grandchildren. So would they be slaves of technology, you know, just going along with what's happening, what they are immersed in? Would they be living in cognitive spaces architected by, you know, some other people, or would they still have freedom, an agency? I mean, take children for example. I mean, we all, when we were growing up, benefited from free play. And you know, there's plenty of research now that tells you free play is great for, you know, emotional relationships, being productive in society, and just mental health. How much time are we doing free play with kids these days? I don't know. So, you know, those kind of questions that we keep asking ourselves, I think then we can do a better world for future generations where tech will be doing great things, but we'll still, as human beings, we'll be masters and we'll be pursuing our own growth because we have a long way to go in terms of our own evolution as sentient beings.

Katherine (Kate) Warren (40:38)

Ambassador Gill, thank you so much for helping us connect all of these global debates and forums and initiatives to the real stakes of what this means. We really appreciate your time.

Amandeep Singh Gill (40:48)

My pleasure. Thank you for having me, Catherine.