← The AI Policy Podcast
The AI Policy Podcast
Live from New Delhi: Our Takeaways from the India AI Impact Summit
00:50:50
This special episode was recorded in India on the last day of the India AI Impact Summit.
Loading summary
Transcript69 lines
- [00:00]
A
Foreign.
- [00:11]
B
Welcome Back to the AI Policy Podcast. I'm Sadie McCullough. I'm here with Greg Allen, as always, and we're speaking to you live from day five of the India AI Impact Summit in New Delhi. So, Greg, let's just jump right into it. Looking back at the past week, we've been quite busy. So what have been the standout moments and highlights for you in this experience?
- [00:33]
A
Well, overall, I think it's been a really important summit. A lot of big things have happened here. Conversations we like got really important updates in where we stand in the global AI Policy conversation. I can't start there, though. I have to start with my petty complaints. And if you've been looking on social media about just how bad traffic has been here, oh, boy, it really, really has. And so let me just, you know, we love quotes here on the AI Policy Podcast. And let me start with a quote from the former Minister of Electronics and Information Technology who was speaking at a side event. And so this is a former senior Indian government official who was in charge of AI in the very recent past. And here was his quote. Everything broke down. It was chaos on the first day. So, you know, you didn't hear from me. It's the official Indian position is that it was chaos. I mean, I think we spent like multiple times we were in traffic for more than an hour on the way out. On the way there, you specifically had
- [01:38]
B
some pretty brutal experiences.
- [01:40]
A
But that is the price of having 100 plus thousand people come in, including a lot of world leaders. So we're here at the ITC Moria Hotel. It's the same hotel that French President Emmanuel Macron is at.
- [01:57]
B
I ran into him in the lobby.
- [01:58]
A
Yep, yep. You had, you had your brush with glory. Well, my, my experience was when he went on a jog and he closed the entire road in front of the hotel so he could go for a jog. And, like, that was like, another reason why Travigo don't. So one thing I have to say is just like my friendly advice to the government of India. You know, it is expensive to own and operate a helicopter for your head of state and give them the ability to just fly over traffic. Sure, helicopters aren't free. You got to pay for maintenance. But I feel very confident in saying that the GDP of India would increase by more than the cost of this
- [02:39]
B
helicopter if people could get to where they wanted to go.
- [02:42]
A
MODI needs a helicopter, all right, so that we can just have him fly over the traffic as opposed to closing every road. And at one point, the traffic jam was 5km long, which is crazy long. Yeah, so that was it. Inside the summit itself, they. They basically evicted everybody from this university. The university campus was where the summit took place. And that was pretty cool. There were a bunch of sessions, I think they said, over the five days, more than 400 sessions. So, like, you know, the. If you go back to the original Bletchley Park AI summit, that was a very exclusive invitation. Right. If you got to go to that summit, it was not super crowded for France. Their summit was much bigger in terms of the number of people who attended. But they had so many side events, including, like, formal side events that were, like, put on by other parts of the French government, that the actual core summit, once you got into the Grand Palais, it was not very crowded. And you could just, like, as I did, walk up to tech CEOs who were mostly unprotected because, like, they were only allowed, you know, one person. This was crowded everywhere the whole time.
- [03:55]
B
So for those of us who weren't there, like, continue painting us a picture of, like, what the experience was like when all of these, you know, when all these sessions were taking place at the same time. Like, what was a typical day? Like.
- [04:08]
A
Yeah. Well, I think just one final thing on this, this, like, philosophy is India clearly had a. Like. Whereas the Brits had a theory of, like, less is more, India clearly had a hypothesis that more is more. Like, every session, you know, there was, like, more than 400, which was great, you know, because our submission to organize a panel was approved. And so you. I was moderating a panel today that was a lot of fun, but it was pretty hard to, like, know what was going on when. And, you know, get to the stuff that you wanted to go to and avoid the stuff that wasn't, you know, part of your interest portfolio. But, yeah, so you had keynote speakers sometimes. That was head of state, you know, Prime Minister Modi gave the inaugural remarks, both the first day of the summit and also the second inauguration and, like, the fourth day of the summit, because they expanded the days. A lot of big names here, including multiple heads of state, as we mentioned, Emmanuel Macron of France, Lula of Brazil was here, and so on. Also, Michael Kratzios, the White House Director of Science and Technology Policy, he gave a keynote address in which he actually provided some pretty important updates on what the U.S. government is thinking about international AI cooperation. And then every night, you know, there's dinners, there's receptions, there's parties, and it's an amazing opportunity to meet with the AI community from around the world, whether that's government, industry, civil society. You know, we at CSIS put on a dinner on the 16th which had representatives from close to a dozen different government, country, countries and governments, as well as industry, civil society. So it's just one grand big party and sometimes that's chaotic and awesome and sometimes that's chaotic and frustrating, but overall a genuinely big moment for the global AI community. Sure.
- [06:07]
B
And as you already started mentioning, we've been doing a few events related to the summit. So can you tell us more about what the Wadhwani center has been doing and what events we were involved with?
- [06:18]
A
Yeah. So we here at the Wadwani AI center, you know, really have, I think it's fair to say, like an important role that we play in Washington D.C. with respect to international AI policy, which is to the extent that the Washington D.C. aI policy community has an opportunity to hear from and hear about and converse with foreign governments. In the lead up to these summits, we've done events before, the UK AI Safety Summit, before the Korean AI Impact Summit, sorry, the Seoul AI Summit, I should say, and then the Paris AI Action Summit. In the most recent instance, we had S. Krishnan, who's a very high level official in meti, which is the Indian Ministry of Electronics and Technology. So he got to go sort of the official Preview to the D.C. policy think tank community as well as, you know, D.C. government officials. We also had a private roundtable with Abhishek Singh, who's the CEO of India AI, their government roundtable, where he could solicit input from stakeholders in Washington D.C. and then most recently we had a big conference which I should give credit to you for doing most of the organizing legwork, which was terrific. And that we had the Ambassador of India to the United States in a fireside chat with the Ambassador France to the United States, as well as the principal scientific advisor. So we've just been doing a lot to help the Washington D.C. aI policy community engage with India and to help India engage with this community to sort of give it, have a good sense of what they should try to accomplish during this incredibly jam packed week here at the summit.
- [08:03]
B
So just about a year ago you were coming back from the Paris AI Action Summit. So how do you see these two summits comparing to each other?
- [08:12]
A
Well, both of them were big expansions of what had come before. You know, as I mentioned, both the Bletchley Summit and the Seoul Summit were comparatively intimate and small. And the Paris Summit was a real expansion of just about everything. And one of the ways in which it was an expansion was that there was aspects of it that kind of felt like a trade show. The Bletchley Summit in the UK really was focused on a policy conversation around AI safety. And industry was there, but industry was there to participate in that policy conversation. By contrast, the Paris AI Action Summit, there was a lot of industry that was there to be celebrated and to announce big investments, which France did manage to attract some big investments to showcase the strength of French industry and position, France and to a certain extent, Europe, as a leader in the global AI conversation. And that's one thing that they have in common. The India Summit unambiguously had big trade show elements. I mean, the Expo center for industry had everybody hawking everything, talking about all
- [09:21]
B
day, just in there.
- [09:22]
A
You really could. And all of these, you know, different pavilions. There was a little bit of, like, a Davos element to it, because there was one Expo center devoted to industry where, you know, companies would organize meetings, give talks. Anybody could just walk up and, you know, talk to that company in a certain sense, just like at a trade show. And then there was also a country Expo center where, like, there was the Japan Pavilion or the Italy Pavilion or the Switzerland Pavilion, or the Gates center put up, like, an Africa pavilion. Sorry, not the Gates Center. The Gates foundation put up an Africa pavilion. So there was just a lot of interesting stuff going on that really did have, like, this interesting mixture of, like, a policy conference, a trade show, a little bit like Davos thrown in there. And again, more is more. I'm not sure that there was any aspect of the global AI agenda that India did not want represented here. Like, everything was here.
- [10:17]
B
So on that note, let's talk more about, you know, what India's goals were for the summit going into it. So what do you think India's goals were going into it, and do you think that the summit lived up to these goals?
- [10:31]
A
Yeah, so I think there's, like, the official goals, and folks who listen to this podcast have heard a version of that official narrative, both from the Krishnan interview that I did, as well as the fireside chat with Ambassador Quatra. But I think there's also the unofficial goals. And one of the unofficial goals was very much to showcase India as a rising AI leader with an important role to play in the global AI value chain and ecosystem. And I think they accomplished that. I mean, I think every company, you know, the big frontier AI model developers, the big cloud, hyperscalers, the other parts of the value chain, India was pretty good about sort of like holding their hand pretty tightly and saying, what are you going to announce at our summit? So Anthropic, for instance, announced a big partnership with Infosys. In the months leading up to the summit. Microsoft, Google, Amazon, all announced infrastructure investments of more than $10 billion in AI infrastructure in India, which we're going to
- [11:37]
B
talk about in a few minutes for sure.
- [11:39]
A
And so, you know, converting this spectacle into an opportunity to announce deals while you have the world's attention, you want to convert that into benefits for your domestic country. And I have to say I think India played their hand pretty well here. A second goal in terms of showcasing Indian leadership was not just on the industry side, right? Attracting investment, but India is kind of in a moment where it's interested in going out into the world. Right? Like more and more Indian companies have global ambitions, not just Indian ambitions. And we've seen a version of that with India's IT industry, which has been a major exporter of IT services to American companies for a long time. But I think what's, what's new in recent years is more and more Indian companies are trying to sell to Africa. More and more Indian companies are trying to sell to Indonesia. So there's this whole like Global south opportunity. And India really stands out as a Global south country that has a legitimate venture capital ecosystem, that has a legitimate tech startup ecosystem. And as that ecosystem increasingly tries to go global and including going global in the Global south, this was an opportunity for India to showcase its leadership and like what opportunity looks like when you partner with India. And that's true both on the private sector perspective, but also on the public sector perspective. One of the interesting outcomes of this was sharing of public sector AI use cases where India, you know, they said, governments of the world, come talk to us about your most successful use cases for AI in the delivery of public services. And we're going to do the same. And the honest truth is that India does have like impressive achievements in the delivery of AI enabled services. And that includes stuff like use. You know, India has a serious problem with tuberculosis and one of the things that they've managed to do is, and sorry, one aspect of that problem is a shortage of professionals who are trained to interpret tuberculosis tests. And a very popular means of screening for tuberculosis in India is interpreting chest X rays. So if you see, you know, certain kinds of lesions on your lungs that you can infer that there's tuberculosis, there's not a huge population of trained medical professionals who can reliably interpret these chest X rays. And so AI services interpreting these chest X rays makes those diagnostic tools available to a much larger population. You know, they also have voice focused AI enabled chatbots. Some of this is done in, in in concert with philanthropic organizations. Whether that's the, the Wadwani AI Institute is, you know, separate from the Wadwani AI center, but has the, the same namesake or the Gates foundation which is also interested in delivering health enabled AI services. So India does have like interesting stuff on its resume. And this was a cool opportunity for India to say to the rest of the world, we have a playbook where we think important things work and we're happy to tell you how we did it. Like we're not ramming anything down your throats and you know, demanding that you adopt, you know, our approach as like some kind of new del effect. But I think it was an opportunity to be proud and to share successes and I think India did that well.
- [15:09]
B
I think so too. I liked how they talked a lot about how AI is the solution to a lot of problems that they're facing as a country.
- [15:17]
A
Yeah, I mean it was just such an optimistic tone.
- [15:20]
B
Very much.
- [15:20]
A
There was a lot of frankly, you know, poo, pooing AI doomerism at this summit. Even though, as I said, the Indian approach was more is more. So the AI safety community was here. They were talking about the latest edition of the AI safety report. Some of the tech CEOs made some notable remarks on the risks of AI, which we'll talk about. So it's not like India said no to the AI safety community participating fully in this conference. But there was aspects of that were almost drowned out just because the overall tone in so many of the sessions really focused on optimism.
- [16:00]
B
I agree. And now let's get into some of those investments that you started talking about earlier. So as you mentioned, the summit attracted major investments in India's AI ecosystem and served as a venue for deal making. So what were some of the most notable investments that were announced throughout the week?
- [16:16]
A
Yeah, so we already talked about like the, the biggest dollar figure ones were the ones that actually happened in the weeks running up to the summit. But in terms of like, I think what happened here this week, which I thought was interesting was OpenAI announc partnership with Tata and this is Tata Consulting Services, which among other things is like an IT solutions integrator and a powerful one for the delivery of IT services. And so the fact that, you know, OpenAI, which is the provider of ChatGPT, obviously is now working with Tata to develop AI infrastructure for India and to train more Tata Consulting employees to apply the tools in their work. That's really interesting. Basically this is a flagship company of India. It's the de facto national champion in a bunch of industries. And it's interesting that, you know, OpenAI is their partner. And then OpenAI's rival, Anthropic, announcing a different partnership with a different Indian national championship, Infosys. That's really interesting. A couple other things which came out of Google, google.org, google's philanthropic arm, which I thought was really interesting. So there was a quote, $30 million, our google.org global AI for government innovation Impacts Challenge to support partnerships that transform public services using AI from everyday essential services to complex societal challenges, as well as a $30 million google.org AI AI for science impact challenge to support researchers globally who are using AI to drive scientific breakthroughs. So two big announcements to be aligned with with as this summit theme is AI for Impact. You know, what are the good things that you're going to do in terms of delivering government services? And I think, yeah, orienting the conversation around that was, I think, like a genuine con contribution to the field, I guess. Let me say one more thing, actually. You know, the big outcome of the Bletchley Park AI Summit was the signed declaration, right, where so many different countries got together and said, we're going to work together on AI safety. We acknowledge it as a big, big problem. Then on the sole AI Summit, I think one of the big outcomes was these commitments where companies said, we're going to publish this information. We're going to work with governments in the following ways, such and such. So getting those commitments was one of the big outcomes, the Paris AI Action Summit. I would actually say that the biggest outcome was probably just shifting the global AI narrative. I mean, there was a real cutback in the resonance of what the White House would call the AI doomerism message. The Paris AI Action Summit really did orient the global conversation back towards optimism, investment, economic growth, potential productivity, et cetera. And to give India a little bit of credit, I do think that they have certainly increased the volume in the global conversation about the delivery of impactful public services with AI, which I think is to their credit. And that, that, you know, Paris really shifted the global narrative for a year. I don't know if this one is going to, you know, replicate that achievement in terms of this being something that we're, we're talking about and thinking about for the next year, but I hope it is. It's a wonderful, wonderful Goal.
- [19:40]
B
Sure. So in addition to all of those investments and deals that were made, we also saw tons of CEOs from leading tech companies. So let's talk about some of the most interesting speeches that we saw.
- [19:53]
A
You know what we have to talk about first, though, right?
- [19:54]
B
Oh, that they wouldn't hold their hands. Yes, yes, of course.
- [19:57]
A
Okay. So, you know, Prime Minister Modi got all these, you know, these tech CEOs, including, you know, all the leading cloud provider leadership and all the leading AI model provider leadership. And whether it was by coincidence or by government demands, OpenAI CEO Sam Altman and Anthropic CEO Daria Ahmadi are standing next to each other. And at some point, Prime Minister Modi, you know, grabs the two people next to his hand and raises the arm and makes it clear that everybody else on the stage is supposed to do the exact same thing.
- [20:31]
B
Everyone's looking left and right to see if everyone's doing it.
- [20:34]
A
Yes. And everybody out there. It's difficult for me to believe that you haven't seen this clip yet, but if you haven't seen this clip yet, go watch this clip.
- [20:42]
B
At least five different people have sent it to me already.
- [20:44]
A
Yes. It is amazing how Sam Altman and Daria Amidi just, like, very clearly refused to hold hands.
- [20:52]
B
It's like they would do anything.
- [20:53]
A
And they're trying to decide, you know, what should they do, given that they're not going to hold hands. And at one point, like, they have one arm up and the other is down. And then Sam, I think it was Sam, to his credit, cleverly realizes how odd that would look. So he just raises the other hand as a fist so that it's less obvious that he's clearly not holding Daario's hand.
- [21:10]
B
But the video is great.
- [21:12]
A
It's so good. If you were wondering whether or not there's still any bad blood between Sam Altman and Dario, maybe Amadea, I should say. Sorry, I often mispronounce his name for reasons I don't understand. But, yeah,
- [21:27]
B
there just might be.
- [21:28]
A
There's still a little bad blood. Yeah.
- [21:30]
B
Yeah. So as you mentioned, like, there are CEOs from Google, Google, DeepMind, Anthropic, OpenAI, and they all gave speeches on Thursday, so yesterday. So what themes did you notice coming out of these speeches and what other things really, you know, sparked your interest?
- [21:46]
A
A bunch of big themes. I think one of them was we saw an updated timeline from technology CEOs about their predicted arrival of artificial general intelligence and artificial superintelligence. And, you know, we've seen this oscillation since the launch of ChatGPT about whether or not the gains from AI, the increased performance potential of AI was likely to keep going or whether or not it was likely to plateau. And one of the moments that we really influential that we've talked about in this podcast was, you know, the rise of inference scaling, defeating the ARC AGI benchmarks. And that that was really impactful in telling people like actually no, there's a lot more Runway to continue making exponential performance increases. Well, just in the past few months with the launch of Claude Code, there's really been an extreme uptake in optimism about AI agents and saying, yeah, we have a lot of Runway with AI agents. These are deliver real meaningful boosts to productivity. And so that data point of just how powerful AI agents have become really just in the past three months. Well, that matters for what CEOs think and what companies think as to how long we're going to continue making progress and when that progress growth curve might like cross the line of basically as smart as a typical human for all the work that you can do on a computer. So I'm just going to read some of the quotes of what the major CEO EO said on this topic because I think it's, it's a really important sort of new, yeah, you can almost call it like the new conventional wisdom right among the leadership. So here's from Sam Altman, quote, we may be only a couple of years away from early versions of true super intelligence. By the end of 2028, more of the world's intellectual capacity could reside inside of data centers than outside of them. I mean, just, just, just think how much the world has changed since the launch of ChatGPT. Yeah, what he's saying is that the roller coaster ride that we're going to go on over the next two years makes what we've gone over the past three years look like nothing. Right. That's just how much change, you know, he's predicting when he says true superintelligence. More of the world's intelligence is going to be in computers than in brains. And there will be a computer, at least one, that's smarter than the smartest human in every category of intelligence. I mean, that's a real revolution. So now here's Dennis Hassabis, who's the CEO of Google DeepMind. And now in 2026, we're at another threshold moment where AGI, artificial general intelligence is on the horizon. Maybe within the next five years, something like 10 times the impact of the Industrial Revolution. But happening at 10 times the spe. Now that's no, he's saying five years. Right. But that's notably a big speed up because not that long ago he had been talking about it being 10 years. Right. So he's like cutting his own prediction in half based on what he's seen in the AI agents. And then Dario Amodei saying, quote, we're increasingly close to what I've called a country of geniuses in a data center. And I think the progress that AI agents have made, and especially the increasingly widespread adoption of agents for meaningful work getting done, has just made the country of geniuses narrative a lot more visceral. Like going from chatbot who tells you things to agent who does things for you. Like, you kind of can understand at a more visceral level what a country of geniuses inside a data center might look like, because now we know what these agents look like. So that was probably the one big thing, is just updating everybody's sort of data points on what superintelligence and artificial general intelligence might look like over the next few years.
- [25:56]
B
Okay, well, let's shift gears then to economic disruption of labor. What did you see happen around this topic at the summit?
- [26:04]
A
Yeah, I think, you know, just what the CEOs are saying is that labor disruption is going to happen right there. I don't know of any major tech CEO who's not predicting radical labor disruption. So if you think of like Dario, Amadeus was kind of an outlier a year ago, two years ago, in pretty consistently predicting radical change in the labor market because of AI at this stage. I would just say that's the conventional wisdom amongst tech leadership and they're willing to say it openly. So from Sam Altman, quote, it's going to change the economics of a lot of things. Current jobs are going to get disrupted. We're not so worried about the long term. Technology always disrupts jobs. We always find new and better things to do. And then here's Dario, quote, AI will greatly grow the economic pie, but because it's happening so fast, it may lead to a time of disruption. And then finally, Sundar Pichai, the CEO of Google AI will undeniably reshape the workforce, automating some roles, evolving others, and creating entirely new careers. So not everybody sort of saying what Dario is saying. And like all of white collar work might be automated in the very near future, but all, all predicting very profound disruption in the not too distant future, which they kind of have to predict that if their, their timelines look like this for when they think AGI and especially when super intelligence might, you know, show up.
- [27:33]
B
So what are they saying about India's large role in AI at the summit?
- [27:36]
A
Well, nobody's being rude, right? It's the host country. Everybody's saying very nice things. And I just think in general, all three of them say India is doing great stuff. They're going to be doing more great stuff. And they really do have a shot at being a leader in global AI. Which, you know, just from some of my conversation around the summit, not everybody really sees it that way. Some people are quite pessimistic about, you know, we've seen a recent big drop off in stock prices of software technology companies here in the United States and kind of the flavor of software as a service. India's got a lot of companies who are big at that. So, you know, one open question is actually, is the structure of India's economy uniquely vulnerable to the way in which AI might disrupt labor, in which AI might disrupt the economy in the near term? Right. Like, are they kind of the low hanging fruit to get picked and get disrupted? That's one argument at least.
- [28:36]
B
Okay, so as you mentioned before, there were quite a few heads of state at the summit this week. So Prime Minister Modi and President Macron also gave keynote speeches on Thursday in addition to all of these tech CEOs. So what were their main messages?
- [28:50]
A
Yeah, I think let's start with Modi because he's obviously the host now. You know, I had a ticket to his address, but his address was in Hindi. So the, the transcript ended up being more influential than my, my ticket.
- [29:06]
B
More than the sign language robot.
- [29:08]
A
There was a sign language robot. An AI powered sign language robot. Okay, so, so I think the first thing is he did talk about impact. The real question is, what is AI actually doing? What is AI actually accomplishing? As I said, it was an optimistic message. At one point he said, we harbor no fear. Again, the sort of rebuttal to the doomerism. But what's interesting is that he is talking about the need for safeguards. So even as he's sort of rejecting fear, he is promoting certain things. So here's a quote where he talks about standards and safeguards. Quote, there is a paramount need today to establish global standards. There must be established levels of authenticity for content within the digital world. People must know what is authentic and what has been generated by AI. As AI increasingly generates text, images and video, the need for robust standards is rapidly increasing. Therefore, it is imperative that trust technology is cultivated from the very outset this must be an inherent part of the system. So it's sort of of like, you know, he's obviously in this part of the section talking about content authentication, but he's not saying we don't need some kind of interventions. He's just saying he doesn't want that to translate into pessimism. He thinks the. The correct answer here is optimism. Another one of his big messages for the entire summit was inclusion. And, you know, that actually ended up being a challenge in the United States because diversity, equity and inclusion DEI this administration is very staunchly that. And so, you know, I think India was saying, look, when we say inclusion, we don't mean that like we're talking about something else. But it was nevertheless, it looks sort of a political sticking point between the two countries and the United States. Ultimately, I think. I don't know if the actual definitive answer has come out as of this recording, but I don't think the United States is going to sign the overall declaration. But part of the inclusion was what they called democratization, that AI must be available to everyone. They don't want a world where just the United States States has a monopoly or just the United States and China have a duo poly. And so they're talking about we need more AI capabilities available to more countries more broadly. So that was a big part, I think, Modi's message.
- [31:31]
B
So what about Kratzios? Michael Kratzios, Director of Office of Science and Technology Policy at the White House. What did he talk about?
- [31:39]
A
Yeah, so Michael Kratzios was the senior United States official leading the delegation here, here in New Delhi. Other notable officials who were in attendance were Sriram Krishnan, who's the senior AI Policy advisor at the White House, and then also Under Secretary of State Jacob Heidelberg. So not a head of state. You know, President Trump is not here, not even a vice president who the United States has in the past sent. You know, Kamala Harris attended the Bletchley Park Summit. Vice President J.D. vance attended the Paris AI Action Summit. But even though, you know, the United States did not send that, that level of official, I do encourage folks to actually read Gratios remarks because it does provide a lot of new information in terms of where the Trump administration is in AI policy and also how it's reacting to the changes not just in the technology over the past few months, but also the changes in the political landscape. So see you sort of seeing a really important update to where the administration is at. So one area where I think that's really important is on AI regulation. So he's critical of many regulatory efforts, but he's not critical of all new potential AI regulation. So if you look at, you know, the executive order that was designed to curb state level AI regulation, it's really noteworthy. But here is, here's what Kratzio said about regulation and this is a long quote, but I do think all of this is actually important to understand where the White House is coming from. So, quote one is trust. Regulatory and non regulatory policy frameworks that safeguard the public interest are necessary to earn the public's trust in AI. To give the American people confidence in these tools. The Trump administration seeks to support legislators as they construct a national policy framework that protects children, prevents censorship, respects intellectual, intellectual property, and safeguards our workers, families and communities. So, you know, rewind the clock. What is he commenting on? He's commenting on the Congress's effort to pass a moratorium on state level AI regulation. Certainly large categories, not everything of state level AI regulation. And the topics that he's talking about here were a lot of the sticking points in that debate when that moratorium ultimately failed. And there was sort of a debate between those who wanted the moratorium first, but theoretically also wanted some kind of national AI regulatory framework. Congressman Jay Olbernolti, who we had on this podcast, notably made that argument. So that was one side of the debate is, you know, let's have the moratorium first and then let's establish the framework. A lot of other folks basically said, yeah, we're not sure we really trust you. Like, even if we support the idea that this should be regulated at the national level, not at the state level level, we're a little bit worried you're going to kind of bait and switch us here and that you're going to have this moratorium and then you're going to give up on it. So it's clear that to get the moratorium, you have to have some kind of national level framework. And here is the Trump administration officially saying, we want to work with Congress. We know these need to be in the national framework. Let's figure out a national framework. You know where that goes? I don't know, but that's really, really interesting. Second thing that Kratio said on regulation, quote, the other primary limiting factor to US AI adoption is regulatory certainty and clarity. It is our position that with smart updates to existing frameworks to reflect new technological realities, use case and sector specific regulation best allows adoption. This gives infinite industry confidence that tomorrow's rules will be common sense deployments of today. Sorry, developments of today's allowing them to focus on creative deployment. Deployment. So that's really interesting because when you say what you need is regulatory certainty, that actually is not an argument for very low regulation or an argument for very high regulation. That's an argument for stable regulation. That's an argument for picking something and sticking with it. And so it's actually kind of an argument like let's have this national AI regulatory framework, let's figure it out fast. Like let's get something on the books and then, and let's make sure that companies understand this is where we're going to be at. So I thought that was, that was really interesting.
- [36:14]
B
That is interesting.
- [36:15]
A
So Kratio said more interesting things. One is actually an extension of many of the issues that he was talking about when he came to CSIS and I got to interview him about the release of the AI action plan. And it's intersecting with this debate this term that was all over the AI side summit, which is AI sovereignty. Yes. And I gotta tell you, Sadie, I'm not sure I've ever heard two people define AI sovereignty the same way. You know, a lot of like the AI sovereignty debate is like, I know I want AI sovereignty, but what actually is it? But what actually is it just sounds good.
- [36:53]
B
It's a big buzzword of this summit for sure.
- [36:55]
A
So Kratzios now has his version, which I think is like important. And he's really, really, this is kind of interesting. He's kind of connecting the idea of AI sovereignty with buying American technology, not renting American technology, not buying the technology as a service. So what I'm saying is going a little bit farther than what he said. So let's, let's read what he said. Apologies, this is a long quote, but I think it's really important. So here we go. Quote. Real AI sovereignty means owning and using best in class technology for the benefit of your people and charting your national destiny in the midst of global transformations. It does not mean waiting to participate in an AI enabled global market until you have tried and failed to build self sufficiency. Complete technological self containment is unrealistic for any country because the AI stack is incredibly complex. But strategic autonomy alongside rapid AI adoption is achievable. And it is a necessity for independent nations. America wants to help. We believe that independent partners are critical to unlocking the prosperity AI adoption can open to all of us. That is why the President launched the American AI export program. That is why I am here with you today. America is the only AI superpower willing and able to truly empower partner nations in your pursuit, pursuit of meaningful AI sovereignty. American companies can build large independent AI infrastructure with secure and robust supply chains that minimize backdoor risk. They build it, it's yours. Now isn't that interesting, right?
- [38:40]
B
Like break it down for us.
- [38:43]
A
Yeah, I mean what's so interesting is that this is a, this is about exports and it really sounds like it's about hardware exports. You know, he is not telling countries a message. You could imagine that he would say, right? He might have said, hey, developing Frontier AI models costs $700 billion. So you like shouldn't do that. Like you should rent ours, right? Buy a subscription to Chat gbt, buy a subscription to Google Gemini, buy a subscription to Anthropic. Don't, don't build your own. Instead what he's saying, he doesn't say Nvidia, but he's like, buy American chips, buy American data servers, state data centers, you know, partner with America and you will own it. They build it, it's yours, that you will not rent it. So I don't know, I really am very curious what the reaction to those remarks is going to be among companies like OpenAI, like Anthropic, but also the cloud providers, right? Like think about, about, you know, if you're Microsoft and you rent AI capacity or if you, you know, via Azure cloud services. It's not clear to me like what, what role they see in what this vision is of AI sovereignty. Now why might Kratzios and his team and this White House have reached that conclusion about sort of, you know, not opposing AI sovereignty, you know, not saying, hey countries, what you should really do is get over this dream of AI sovereignty. He is saying, get over this dream of self sufficiency. You're not going to do it alone. But what he's saying is you can have sovereignty. We actually want to help you have sovereignty. Which is a really interesting, you know, sort of rhetorical stance for the United States to take. So why might he be doing that? One possible theory of the case case is that he just thinks there's no fighting it, right? If he just goes around the world and has a message of sovereignty is bad, except permanent subservience to the United States.
- [40:51]
B
That won't go over well.
- [40:53]
A
I would assume not every country is going to be down with that message. And in other parts he criticizes the AI diffusion rule, which, you know, some criticism of that have said it was designed to keep countries around the world dependent upon the the United States AI state stack. This is not an argument of forced dependence. This is an argument of enabled independence, which is, which is really quite interesting. So one reason why he might have made this argument is just because he had to. Because rhetorically, there wasn't a way to make a pitch of being pro dependent on America that would come off nice at a conference like this. Here's another reason, and it actually comes from his colleague at the White House, Sriram Krishnan, who I got to see make remarks at a side event, event organized by the Tony Blair Institute, where he was very positive on open source AI, where he said, free and works Pretty good is a very compelling offer. It was something to that effect. It wasn't precisely that quote, but it was close to that quote. So what if. What if the White House has concluded that open source is going to win? Right? If free and works pretty good is like an utterly compelling offer offer, then what that would imply is that OpenAI as a closed source model provider, Anthropic as a closed source model provider, it's actually not going to dominate the world the way it does today. So if they have this prediction that ultimately open source is going to win, then, well, there's not a lot of money to be made in providing those open source models. Where is there money to be made? It's providing the chips that run those open source models. So I'm speculating a lot here, I'm reading between the lines a lot here, but I don't think the way I connected the dots is that crazy. I mean, it is certainly a plausible interpretation of what Kratzio said combined with what Sriram Krishnan said. If I was the big AI model developers, if I was the big hyperscale cloud providers, I would be calling the White House and being like, yeah, why didn't you talk about how we're a part of this future?
- [43:01]
B
Okay, well, we'll have to wait and see what happens.
- [43:03]
A
We'll have to wait and see what happens. All right. And. And at the risk of filibustering on my own podcast, there's one more thing that Kratio said. Sadie, I appreciate you pairing with me. You're being a great sport. Sorry, I'm filibustering here, but okay, so he did one more thing, which was again, follow on to what happened in the the AI action plan and the executive orders that dropped at the same time. He's fleshing out the American AI exports program and in both developed countries and developing partner nations, and that included creating the Tech Corps as a new partner initiative of the Peace Corps. Sorry, Peace Corps, I should say. I don't know why I'm, you know, ignoring the silent letters. So the, the Tech Corps is going to actually go into developing countries and help them adopt AI related public services and help them sort of build out digital infrastructure associated with AI. AI. Why is that interesting? Why is it interesting that he put that in, in part of the Peace Corps and, and connected it explicitly to the American AI exports program. Well, if you remember when Kratios came to CSIS and he was talking about why we needed that development program, he mentioned the frustrations that America had in competing with Huawei in telecommunications where there was this period, I mean, ultimately Huawei's like 5G technology became, you know, as good or better and cheaper than what the west had in some cases. But there was this period of time where American technology was better and at least from a total cost of ownership perspective was cheaper, you know, and so why didn't countries in the Global south, you know, buy American as well as, you know, American critical partners, many European and Japanese companies? And the hypothesis was that the Chinese government provided attractive financing options so you can, can have the, the 5G or the telecommunications infrastructure now. And once that starts generating the profit, that's when you pay us back. You don't have to pay us up front. Well, he's talking about mobilizing all these government agencies to provide attractive financing options to buy the American AI stack, like the Development Finance Corporation is just one example. And the second thing he's talking about is this Tech core. And there again a lot of the benefits of flawway Way was we're not just going to like build this equipment and leave your country. We're going to build this equipment and we're going to train the 10,000 people your country needs to operate it yourself. Right. So there was this horror, there's this whole attractive solution that Huawei would integrate for those countries where it's like we are your one stop shop. Everything you need is going to be from us. And that includes training your domestic workforce so that you're not dependent upon our engineers living in your country for forever. I see some echoes with that, with this Tech Core initiative linked to the American AI exports program. Again, this is an announcement. All of this stuff is pretty nascent. We kind of have to see how the rubber is going to meet the road. But I thought that was super interesting.
- [46:10]
B
So let's finish up the episode with talking about President Emmanuel, French President Emmanuel Macron's remarks. What did you find most interesting about that and what can you tell us about it.
- [46:21]
A
Yeah, I think Macron is an interesting place to end it because he obviously had the last summit and so he is the sort of connective thread about where we are. So there's a few things that I think are interesting. The first is I want to say around the importance of international collaboration. His quote said the old world said you compete or you lose. The new world says you connect or you fall behind. Behind. Now that's really interesting. I think, I think it's a thinly veiled criticism of America. Right. You know, being very explicitly and harshly competitive in many circumstances and saying that actually connection is a key part of it. He's also just like Michael Kratos, however, going against the self sufficiency hypothesis. Right. Countries are not going to do this alone. And, and then China is really, really stands out as the, as the country that really has embraced the self sufficiency initiative and approach. The second thing is around sovereignty and independence. This is kind of interesting. I don't quite understand what he means here, but I still think it's an interesting quote. India chose granular and smart and Europe chose sovereign and scaled. But both chose independence and both were right. Right. Well, here's the thing, like India has sovereignty requirements, including in building out like cloud infrastructure locally. So what he means when he says Europe chose sovereignty and India didn't, I don't know what that means, but both chose independence. And maybe, maybe that, that's getting it, you know, sort of the, the self sufficiency versus other distinction, I don't quite know. But later he says France and India share a common vision, a sovereign AI used to protect our planet and foster prosperity for all. So sovereignty, I think is like one of the buzzwords that's really reaching a crescendo here at this summit. And then finally he has a defense of the European approach, which I think really is interesting because last year he was pretty critical of the European approach. I mean, he was, he was fighting against the EU AI act and I think in important ways succeeded in changing not just the way the act was written, but very, very importantly, especially around last year this time, the way the act was implemented, because the regulators do have quite a bit of flexibility. And so here's what he said as a defense of the European approach, quote, and opposite to what some misinformed friends have been saying, wink, wink, I think we're talking about America. Europe is not blindly focused on regulation. Europe is a space for innovation and investment. But it is a safe space and safe spaces will win in the long run. I'm sure of that. And so he's basically saying, you know, what ultimately countries, companies, consumers are going to want is AI that works reliably, safely as intended over its use cases. And, you know, to the extent that Europe has chosen regulation, it's in pursuit of that and it's nowhere near as bad as I think he's thinly attacking the Trump administration has been saying. So look, that was a tiny fraction of what went on, of what went on here at the India AI Impact Summit. It was already a lot, but hopefully folks found that useful as as what we certainly took away from a really awesome week.
- [49:52]
B
Well, I think you hit the most important parts and helped everyone understand, but I've really enjoyed being here with you and talking more about it. So it's almost time for us to head to the airport, but literally, yeah,
- [50:03]
A
we're about to go get on the
- [50:04]
B
plane, but thanks so much for listening to this episode of the AI Policy Podcast. Greg will be back next week with more episodes, so stay tuned.
- [50:12]
A
We've heard those of you out there who are crying for more News Roundup episodes, so we'll be back in the not too distant future next week with two episodes, one News Roundup and one interview. So stay tuned. Thanks for listening to this episode of the AI Policy Podcast Podcast. If you like what you heard, there's an easy way for you to help us. Please give us a five star review on your favorite podcast platform and subscribe and tell your friends. It really helps when you spread the word. This podcast was produced by Sarah Baker, Sadie McCullough and Matt Mann. See you next time.