Name: Brainstorm AI London 2024: AI’s Impact On The Geopolitical Landscape
Uploaded: 2024-04-15
Duration: 17 min 55 s
Description: Emran Mian, Director General, Digital Technologies and Telecoms, Department for Science, Innovation and Technology (DSIT) Robyn Scott, Co-founder and CEO, Apolitical Guy Williams, Head, UK Defence, Palantir Technologies Moderator: Alan Murray, Chief Executive Officer, FORTUNE

Emran Mian, Director General, Digital Technologies and Telecoms, Department for Science, Innovation and Technology (DSIT) Robyn Scott, Co-founder and CEO, Apolitical Guy Williams, Head, UK Defence, Palantir Technologies Moderator: Alan Murray, Chief Executive Officer, FORTUNE

Brainstorm AI London 2024: AI’s Impact On The Geopolitical Landscape

Photography by Joe Maher / Fortune Brainstorm AI

Conferences

Emran Mian,, Innovation and Technology (DSIT), Robyn Scott, Apolitical, Guy Williams, Palantir Technologies, Alan Murray, FORTUNE

Thank you. We have a lot to talk about and a very short time to talk about it. Enron, I'm eager to talk to you about what can government actually do to mitigate the downside risks of A I, Rob? And I'm eager to talk to you about how can government use A I to just be better at what it does. But I, I think I have to start with you uh because you're in the defense business, uh you work for pun here but work closely with the UK government in the last 48 hours we've all seen on television, a pretty extraordinary uh event that Iran sent more than 300 drones, ballistic missiles, cruise missiles aimed straight at Israel. And we're told that 99% of them were shot out of the sky before they hit the ground, which is if true. Extraordinary. What is this event that we've all witnessed in the last 48 hours? Tell us about the changing face of national defense. I think the thing that's really highlighted for me is um if you go back in time, only, not even 10 years, but say four or four or maybe even three military operations are all about the speed of decision making decisions at the speed of relevance I think is a term often used. And historically, we've seen things like if you go all the way to before something is launched, it might be looking at satellite imagery to understand what things have moved, what changes have taken place. Now, with the introduction of computer vision models, you're able to take a two kilometer by five kilometer image. And within a second, identify all of the things that you might be interested in, whether it's changed or whether it is something that you, you recognize historically, that might have taken analyst hours to get to that point on one single image. And now it would take thousands of 300 of them with a pretty limited time. Yeah, exactly. And so the idea of identifying where these things might be coming from and deciding these are the ones of interest and then maybe tasking things, the collection assets that you might have, which could be sensitive or what have you and then passing that structured data through a model to then understand what are the interesting things do you care about all the things that might be good indicators? And more so it's a secret, it is a sea change. Totally, the compression of decision making times has been, you were directly involved. My understanding is you were directly involved in working with Ukraine to very much the same weaponry, by the way, both the offensive weaponry and the defensive weaponry, uh that Russia was using against Ukraine. What did you learn in that experience? Uh The Ukraine example is, is fascinating and very unique in the fact that you have seen, um, a complete tech sector nationalized into the defense sector and the speed of change that they've been able to deliver in terms of transformation and the use of A I and other technologies has been incredible. Well, they had some help, a little help. Yeah, but I mean, again, one of the, one of the earlier um commentators are talking about the, the getting your data in a healthy place and then run the models across it. And that's a very important part. The thing is we're fortunate in both of those cases that the defensive capabilities were so strong, but these are going to empower offensive capabilities as well, right? I mean, it's gonna be, it's gonna be a race. Yeah, I think the idea of strategic strategic competition is definitely, definitely with us and I think in terms of governments investing into future technologies and A I is, is gonna be pretty um pretty game changing, deterministic uh Emma and I, I want to back up a little bit and, and get to the broader question of what A I means for governments, uh which is the question you wrestle with every day in your position with the UK government, I guess. The thing is someone said this from the stage earlier today is this is moving so fast, so fast. How can government hope to keep up with the pace of change and mitigate the negative consequences in any way? Yeah, so a big part of our approach has been to do things differently. So we've been much more open to bringing in expertise from outside of government than perhaps sometimes we are as so we brought in lots of expertise from A I firms from academia, including Ian Hogarth who chairs our A I Safety Institute I think was speaking earlier in the conference and we've been much more open to the expertise and sometimes we are as government. We've also moved really, really quickly and in order to move as quickly as that, you really need political support all the way up to the top of government. And we've absolutely had that we've had a prime minister who's chosen to prioritize this set of issues. And in particular, he lent into saying that the UK would host the first A I Safety Summit which we ran in November last year. And I think that's another important lesson for us in terms of dealing with some of the risks or some of the potential downsides of artificial intelligence. We've got to do it with other governments. We've got to do this together. We've got to build international governance around this and that's why it's been so important to us to kind of begin to lead that conversation. So we hosted the first summit last year, we're co hosting the next summit with Korea next month, we are also trying to collaborate really extensively on A I safety testing with other countries. So our A I safety institute just signed a partnership agreement with the same institute in the US. And we're talking to these capabilities as they begin to be developed in other countries as well. So kind of that collaborative approach I think is really fundamental. And so what is sort of the top three things that you think you can accomplish with this A I safety effort? So I think the first thing is to bind this set of countries together, to build that connective tissue between the countries which have frontier A I firms and which you would expect to be at the frontier of A I being deployed across the wider economy. I think the second thing I'd say is there's still a lot to do in terms of understanding the science of A I safety. So there's a lot of discussion about what's the right approach to regulating frontier A I. And our view has been that the best thing to do first is to figure out what does good A I safety testing look like and set standards for that and to do that in collaboration with other people who are at the frontier of doing that work. And that's what the A I Safety institute been focused on. We want to understand what really good safety testing looks like, do that in collaboration with the firms and do that in collaboration with other countries. And is the end result, you give a seal of approval to certain models and not to others. What does this look like? Five or 10 years down the road? I think it's hard at this point to say what it looks like. Five or 10 years down the road. I think our approach at the moment is let's establish the basic science, let's figure out what good safety testing looks like. And alongside that, let's use our existing regulatory system for the harms that are manifesting potentially here. And now, and I think that's another important part of our approach. The digital markets that we're talking about here are markets that we already regulate in important ways. We have just passed legislation to allow us to better regulate for harms that occur online. Um Many of those harms will now also be created with A I, but we've, we've just regulated for that. So we should use that regulatory framework to deal with A I generated harms in the same way as we were planning to use it for other things. Yeah. Robin. Uh uh you're working uh from a different angle on these issues. You're trying to figure out how can A I make government better at what it does. Can you talk a little bit about what you're doing and where you're seeing results. Yeah. So the perspective I bring to this is running a political, which is a global network used by about a quarter of a million public servants. We also do online training and as part of that, we have a government A I campus and we've trained already about 10,000 public officials around the world. So that's our lens. The headlines are the potential is immense. The Turing Institute came out with a report recently which speaks to speak to this in the UK context, but it's generalisable. There are about a billion transactions a year between government and citizens. About 100 and 40 million of those are complex but repeatable tasks. And of those uh the the institute estimates that about 84% can be uh automated with A I. So incredible potential to cut into the work of government, improve citizen services, improve the return on taxpayer dollars. There's also a secondary reason for doing this and for adopting A I, because better users of A I are better regulators of A I and they're better buyers of A I. And I think, you know, the UK really has shown leadership on the general adoption front in government as well. And is there a willingness to do this gets back to the retraining issue? We were talking about a few minutes ago. Is there a willingness in government to really embrace the effort? Yeah, absolutely. So it's very interesting to look at government's current data sort of position on A I. First of all, somewhat surprisingly, your average public servant, we've done a lot of polling on this is more positive than negative on the potential of A I that doesn't hold necessarily for prior technologies. And given all the talk about concerns, we've been surprised how generally optimistic they are. There is much more optimism in emerging markets versus developed markets. The conversation in emerging markets is all about this is going to solve personalized education. This is going to bring great health care. The conversation in developed markets tends to have much more of a bent to work towards safety and harms and economic nationalism is also uh becoming a more and more important uh force, sorry A I nationalism rather. And we've seen big plays recently in the UAE and India and expect to see many more around the world. So all those are context for adoption, but there is general willingness to use it. The vast majority of public servants have already played with generative A I sometimes I think under approved guidance, probably sometimes somewhat off the books and that's only going to increase pretty dramatically. Yeah, I'm glad you mentioned economic nationalism because one of the things that makes this conversation about geopolitical effects uh interesting and challenging is the existence of China not represented on this, on this stage, China is certainly an A I superpower. There are some people out there who seem to think it will become the leading A I superpower. Uh and it has a very different background set of values that could be on the other side of any uh uh actual uh uh warfare. Um It, it takes a very different approach to many of the questions uh to use it a simple example. Uh The UK has said very clearly, we're not going to use facial recognition or other technologies to pick people out in a massive crowd crossing the street. I believe China has done the exact opposite so that we can make great gains in social cohesion if we have the ability to do that. So how, how does China's strength in A I affect these conversations, Emran, I'll start with you. Yeah. So our approach has been to include China in the international conversation about safety. Uh So China will represent it at the A I Safety Summit in Bletchley Park. And we think that's important because China produces a lot of fundamental research in A I. Uh and it has a very busy and a very active and a very interesting A I ecosystem. Um So having China as part of this conversation, we think is important. But at the same time, we've been really clear as a government that where China is engaged in activities that affect or harm UK interests, we will say so clearly and plainly. Uh so we've done that in the past few weeks we have called out where China has been engaged in uh offensive cyber operations in the UK. And we think that those two things are entirely consistent. Well, there are a lot of people in the US these days who think maybe they aren't consistent and that should we be sharing all our research with China if we're going to be in conflict over whether it's cyber or actual warfare? Yeah. And that's why our approach will not be to share all of our research with China. We're very clear on there are areas of research that you, that we will protect and that we do protect Robin. How do you think about this? We're looking very much at the practice, the applied side of A I in government. And one of the interesting general admissions, I think there's a general admission around thinking around capability in government. There's a lot of talk about, we must do this, there's a lot of talk about adoption, but there isn't the focus on building the kind of diffusion capability. So you can actually roll it out across government. So we look at that and then we look at how the exchange of best practice happens because the field is moving so fast, you are never going to win. If you just take top down guidance or uh you know, directives, you've got to talk to others who are in the field. And I think one of the interesting dynamics is actually the vast majority of countries are quite good at collaborating and sharing best practice. We see a lot of that on our platform, China kind of doesn't take part for the most part explicitly or at least in the same way in that sharing of best practice. So the question I'm interested in is is there going to be a siloing around how A I is rolled out in the Chinese government versus elsewhere? And should we be trying to get more permeability so we can, so we can be mutual learning around non controversial things? Like how do you get great services to citizens? How do you bring down the carbon footprint of these A I technologies? We heard about that earlier. Yep and Guy in defense, how, how do you think about China? Um Well, I mean, as you point out that facial recognition is reportedly like fantastic, but when it comes to things like large language models, apparently there's a slight limitation because they um restrict elements of the internet available to specific areas. So there is a, there is a theory that they might be slightly behind but to pick up Robin's Point there, I think what's really important is almost not ignoring it but just more internal focused of looking at an organization and in parallel to the regulatory and the safety side is like, how do you really push the use of these technologies to win? Like what are you doing day in day out to make the most of this tech to actually improve your, improve your work. And there's often this concern about losing jobs, but we like to view it as um in a marvel context of Team Iron Man versus Team Ultron. And it's how do you bring this technology to humans so that they can be 1000 times better at their job? And with the view to how you actually win in the long run, was that a national defense answer? Yeah. Well, I, I think so. I think if, when you're an analyst working in the defense in defense intelligence, if an operational commander is, how do you, how do you bring? But you, but you have to be thinking about, do our cap are our cap, do their capabilities match ours are our capabilities, period. That's the nature of defense. And I think that's the important thing of, of balancing the innovation and driving innovation forward in parallel to the safety side rather than letting the safety side restrict what you do in terms of innovation. Ok. Last question, I'd like to ask all of you to answer. If A I ran the government, what would we be doing differently today? What would we be doing that we aren't doing? And I'm going to make you go first. Um, we'd be doing meaningful work, not busy work. It would be my blunt answer. Still too much busy work in government. I think, I, I think it, it's very hard and, and A I can help us get rid of it. I hope so. Yeah, Robin. So there's been some data suggesting this is an aside that a majority of people, I think it's younger people in Europe actually want A I politicians. So they want A I to run the government as opposed to the politicians, as opposed to the flesh and blood runs of the coming, coming to government sooner than you may think. But in terms of what that would look like, assuming it's benevolent and aligned A I, I think the the potential the dream is that you get much better, much more personalized services, hyper personalized services to citizens and you can invest a lot more in those because you are saving money because of so many efficiencies that you're also finding Iran final word. Yeah, I'd add, I think government is a complex activity and it's too often con noted as being an elite activity. And I think A I gives us some real opportunities to, to deal with that. I think the way in which A I can help to simplify and explain things to citizens, I think has real power to it. I think the way in which it can help to simplify interactions with citizens, I think has real power to it. Um So I think breaking down that elitism that I think a lot of people experience in their interactions with government, I think, yeah, I can help us with that. Well, thank you all, you've given us a lot to think about. Thanks for being part of brainstorm A I.