Latest / Diplomacy and International Relations / AI Framework and Diplomacy

Transcript

AI Framework and Diplomacy · Diplomacy and International Relations

• 0:00
  Hello and welcome to Diplomacy and International Relations by
• 0:03
  Bastian Freiborg. I'm your host.
• 0:05
  And I'm your expert. It's great to be here.
• 0:07
  Today we are tackling a subject that honestly feels like it's
• 0:10
  pulled from a cyberpunk novel, but it's actually happening in
• 0:14
  embassies and situation rooms right now.
• 0:16
  Oh, absolutely. It's not science fiction
• 0:17
  anymore. Before we really get into it, I
• 0:19
  just want to encourage you to follow and subscribe wherever
• 0:23
  you're listening. It really helps.
• 0:24
  And you can also subscribe to the podcast on Spotify or Apple
• 0:27
  podcast. There's a small monthly fee so
• 0:30
  it helps us keep the lights on and gives you access to extra
• 0:33
  episodes. Yep, the link for that is right
• 0:34
  there in the description. So today's topic, it's huge.
• 0:38
  We're looking at how artificial intelligence isn't just another
• 0:41
  tech trend, it's fundamentally shifting geopolitical power.
• 0:46
  Exactly. I think when most of us hear AI,
• 0:49
  we think of, you know, chat bots writing poems or those funny
• 0:52
  images of the Pope in a puffer jacket, right?
• 0:55
  But the material we've got today, it paints a much, much
• 0:58
  more serious picture. We're seeing this transition
• 1:01
  from what used to be called information power.
• 1:04
  To something new. Algorithmic power.
• 1:07
  OK, that's the key distinction, isn't it?
• 1:09
  It really is. In, let's say, traditional
• 1:11
  international relations, power was about who had the
• 1:15
  information, who knew where the troops were, who had the
• 1:17
  economic data. But algorithmic power is
• 1:20
  different. It's not just about having the
• 1:21
  data anymore. It's about controlling the
• 1:23
  logic, the algorithm, the processes, all that data.
• 1:27
  So it's like the difference between owning a giant library
• 1:31
  and owning the librarian who decides which books you're even
• 1:34
  allowed to see. It's a perfect analogy, and that
• 1:37
  brings us to this really fascinating concept from our
• 1:39
  this is called algorithmic diplomacy.
• 1:41
  Which sounds incredibly cool, but also maybe a little bit
• 1:46
  ominous. It could be both.
• 1:47
  The core idea is that the act of feeding the algorithm, giving it
• 1:51
  data, is now a deliberate act of geopolitical communication.
• 1:55
  States are realizing this. So if they can influence the
• 1:58
  data and the models. They can influence how the
• 2:01
  entire world is interpreted by these systems.
• 2:03
  So that's our mission for this session.
• 2:05
  We're going to unpack that. We'll look at the big players,
• 2:08
  the US, China, the EU and how they're trying to regulate this
• 2:13
  beast to their advantage. And we'll also look at the tools
• 2:15
  diplomats are actually using right now, and then get into the
• 2:18
  really heavy stuff. AI, colonialism and the fight
• 2:22
  for digital sovereignty. It's a lot to cover.
• 2:25
  The sources are incredible though.
• 2:26
  Let's get into it. Let's do it.
• 2:28
  First up, the great divergent this Rand analysis bridging the
• 2:34
  artificial intelligence governance gap.
• 2:37
  It frames this whole thing as a huge philosophical split between
• 2:41
  the US and China. Is this just another arms race?
• 2:44
  You know who's got the fastest computer?
• 2:46
  That's the surface level. And look, speed definitely
• 2:49
  matters, right? But what the Rand analysis
• 2:50
  suggests is that goes so much deeper.
• 2:52
  We're talking about fundamentally different
• 2:54
  philosophies of governance. Yeah, if you look at the US
• 2:56
  approach, it's very market driven.
• 2:59
  It's the Silicon Valley effect. The whole move fast and break
• 3:01
  things idea. To an extent, yeah.
• 3:04
  But from a regulatory standpoint, the US is focused on
• 3:07
  two things, the models and the hardware.
• 3:10
  The hardware being the computer chips.
• 3:12
  Exactly. The executive order 14110 is all
• 3:16
  about compute thresholds. OK, hold on compute thresholds,
• 3:19
  break that down for us. What are they actually
• 3:21
  measuring? Think of it like a speed limit
• 3:23
  for processing power. The US government is basically
• 3:26
  saying if your computer system is powerful enough to perform X
• 3:31
  quadrillion operations per second, we need to know about
• 3:35
  it. So they're regulating the
• 3:36
  capacity to build the AI brain? That's it.
• 3:39
  They assume if you're building a massive engine, you're building
• 3:41
  a race car and they want to register that car.
• 3:44
  And this is where those export controls on chips come in, which
• 3:47
  we keep hearing about. Because that's the geopolitical
• 3:50
  lever. By restricting China's access to
• 3:52
  the most advanced chips, they're trying to physically slow them
• 3:55
  down. It's a hardware choke point.
• 3:57
  Cut off the fuel, the engine stops.
• 3:58
  Pretty much. And diplomatically, the US
• 4:01
  prefers these exclusive clubs, the G7 partnerships with allies.
• 4:06
  They're not too keen on broad, messy UN frameworks for this.
• 4:10
  O EU strategy is control the hardware, control the power and
• 4:15
  do it with your friends. How does China's approach
• 4:17
  compare? The sources say their focus is
• 4:20
  on social. Stability, yes, and it's a world
• 4:22
  apart. China's approach is state led
• 4:25
  and it's all about the content, the outputs.
• 4:27
  So while the US is worried about how powerful the chip is,
• 4:31
  China's worried about what the AI is actually saying.
• 4:34
  Precisely. They want the AI to stay on
• 4:36
  script. Their regulations explicitly say
• 4:40
  AI has to reflect socialist core values.
• 4:43
  It can't subvert state power. For them, it's about harmony
• 4:46
  control. If an AI generates text that
• 4:49
  questions the government, that's a regulatory failure.
• 4:51
  That is a massive difference. One side is measuring math, the
• 4:53
  other is measuring meaning. And here's the interesting part.
• 4:56
  Internationally, China pogus for AI governance through the United
• 5:00
  Nations. That seems counterintuitive.
• 5:02
  If they want tight control, why go to the UN?
• 5:04
  It's huge and famously slow. It's a very clever strategic
• 5:08
  play. In the UN, China has huge sway
• 5:10
  with the global S Yeah, so they frame it as democratizing AI.
• 5:14
  They argue that the US wants to keep AI for the rich countries,
• 5:18
  while China wants to bring it to the General Assembly where
• 5:20
  everyone gets a vote. That is very smart branding.
• 5:23
  It positions them as the champion of the developing
• 5:25
  world. Exactly against the quote UN
• 5:27
  quote elitist West. So to boil it down, the US is
• 5:32
  trying to control the chips, the hardware, and China is trying to
• 5:36
  control the narrative, the content.
• 5:39
  That's a great summary. Which brings us to Europe.
• 5:42
  For listeners who are interested in this from a European or maybe
• 5:46
  a Danish perspective, where does the EU fit in here?
• 5:49
  Are they just watching these two giants fight?
• 5:52
  No, Europe is trying to carve out what you could call a third
• 5:55
  way. There's an analysis from
• 5:57
  Brookings that talks about the Brussels effect.
• 5:59
  I love that term. It sounds like bureaucracy as a
• 6:01
  superpower. It kind of is.
• 6:03
  The Brussels effect is this idea that the EU single market is so
• 6:07
  huge and so valuable that if the EU sets a strict standard,
• 6:11
  global companies will just adopt it everywhere.
• 6:13
  It's easier than making two different products.
• 6:15
  So they're not trying to build the biggest AI, they're trying
• 6:18
  to write the rule book for everyone else's AI.
• 6:20
  Exactly, the EUAI Act is the first big attempt to regulate
• 6:25
  this stuff across the board and they use what's called a risk
• 6:28
  based approach. OK.
• 6:30
  What does that mean? Risk based?
• 6:31
  They categorize AI not by its power, but by its potential for
• 6:35
  harm. You have unacceptable risk.
• 6:38
  Like what? Things like government run
• 6:40
  social scoring systems. You know, tracking citizen
• 6:43
  behavior. The Black Mirror episode.
• 6:45
  Exactly like that. Those are just banned out right
• 6:47
  in the EU. Then you have high risk AI in
• 6:52
  healthcare, policing, border control.
• 6:54
  You can use it, but you have to pass all these strict
• 6:57
  transparency and accuracy tests. And then there's lower risk
• 7:00
  stuff like spam filters or video.
• 7:02
  Games, which gets a much lighter touch yeah, so the strategy is
• 7:05
  you want to sell your AI product to our massive market, you have
• 7:09
  to play by our human centric rules.
• 7:11
  And in doing that, they effectively export European
• 7:13
  values. If a company has to change its
• 7:15
  algorithm for Brussels, that change probably rolls out in
• 7:19
  Copenhagen and Tokyo and New York too.
• 7:21
  That's the goal. But isn't there a risk here?
• 7:23
  I mean, if you regulate this heavily, don't you stifle your
• 7:26
  own innovation? I can imagine there's a big
• 7:28
  debate about that inside the EU. Oh, that's the huge internal
• 7:31
  debate. The sources talk about this
• 7:34
  tension between the Parliament's live committee, which is all
• 7:37
  about civil rights, and the IMCO committee, which is focused on
• 7:40
  business and the market. So a real fear that while Europe
• 7:44
  is busy writing the perfect ethical rulebook.
• 7:47
  The US and China are busy building the future.
• 7:50
  You can't regulate something that you haven't even built yet.
• 7:52
  That's the classic European dilemma, safe and slow or fast
• 7:56
  and risky. OK, let's shift gears a bit.
• 7:58
  We've talked about the high level policy, which is
• 8:00
  fascinating, but what about the people on the ground?
• 8:03
  The actual diplomats? Is AI really being used in
• 8:07
  embassies today or is it still mostly hype?
• 8:09
  This is where it gets really practical.
• 8:10
  There's a great source, Diplomacy, by Doctor Cindy
• 8:13
  Jogrijic, that shows this shift from what he calls diplometrics
• 8:18
  to AI diplomacy. And it mentions something called
• 8:21
  the Bilateral Navigator that sounds like a gadget from a spy
• 8:24
  movie. It's a.
• 8:25
  Very real tool. It's an initiative that produced
• 8:27
  over 18,500 detailed analysis of bilateral relationships between
• 8:33
  all the UN member states. 18,000 A human analyst couldn't even
• 8:37
  That would take a lifetime. Exactly.
• 8:39
  And that's the power. It's about scale.
• 8:41
  It helps diplomats understand these complex relationships
• 8:44
  through data, and it can identify what it calls hidden
• 8:47
  coalitions. Hidden coalitions?
• 8:49
  What do you mean by that? Well, a human diplomat knows
• 8:52
  they're obvious allies, right? France knows it generally aligns
• 8:56
  with Germany, but an AI can scan decades of voting records and
• 9:00
  spot subtle patterns like why do country A and country B, who
• 9:04
  have nothing else in common, always vote the same way on
• 9:06
  maritime law? It reveals these leverage
• 9:09
  points, these alliances that were totally invisible before.
• 9:12
  That's incredible. It's like Moneyball for the UN.
• 9:14
  You're finding value in statistics nobody else is
• 9:16
  looking at. In a way, yeah, But it goes
• 9:19
  beyond just analysis. We're seeing generative AI being
• 9:21
  used in the actual craft of diplomacy.
• 9:24
  Writing speeches and memos for them.
• 9:26
  That's the simple stuff. Think bigger sentiment analysis
• 9:31
  using AI to monitor in real time the public mood in a foreign
• 9:35
  country. So instead of just reading the
• 9:37
  morning paper. You're analyzing millions of
• 9:39
  social media posts to get an instant, precise read on public
• 9:42
  opinion or crisis simulation running scenarios to predict how
• 9:46
  a conflict might escalate if you make a certain move.
• 9:50
  And what about this Negotiator by ChatGPT concept?
• 9:54
  Yeah, the idea there is you simulate your counterpart before
• 9:56
  walking into a high stakes negotiation.
• 9:58
  You can literally argue with an AI that's been trained to think,
• 10:01
  react, and even get offended like the specific person you're
• 10:05
  about to meet. So you get to spar with the bot
• 10:07
  before the real fight. You test your arguments, you
• 10:09
  find the weak points, you refine your strategy.
• 10:11
  You can ask, OK, if I threaten sanctions, how does the AI
• 10:15
  representing this foreign minister react?
• 10:17
  That sounds incredibly useful, but I don't know.
• 10:20
  Doesn't it risk making diplomacy too robotic?
• 10:24
  If everyone is just gaming out the optimal move, do we lose
• 10:27
  that human element? That is the big danger, and the
• 10:30
  sources are very clear on this point.
• 10:31
  This isn't about replacing humans.
• 10:33
  The goal is what they call augmented diplomacy.
• 10:36
  Augmented. Yeah, you combine the raw
• 10:38
  processing power of the AI with uniquely human skills like
• 10:42
  empathy, judgement, cultural nuance.
• 10:45
  The diplomat as a Cyborg, kind of.
• 10:47
  The digitally enhanced statesman, maybe.
• 10:50
  But this brings us to the darker side of all this.
• 10:52
  Because if data is the new oil, who's getting exploited for it?
• 10:57
  This is the AI colonialism concept from that Arcs FF paper.
• 11:01
  This one really really stuck with me.
• 11:02
  It's. A critical lens, especially if
• 11:04
  you're looking at global inequality.
• 11:06
  The dynamic is it's pretty stark.
• 11:09
  You got the global N the US, Western Europe, they have the
• 11:12
  data centers, the infrastructure, the capital.
• 11:14
  And then you have the global S developing nations who provide
• 11:17
  the raw data. And often the cheap human labor
• 11:20
  for things like data labeling. The people clicking this is a
• 11:22
  stop sign 1000 times so a self driving car in California can
• 11:26
  work. Exactly that.
• 11:28
  It's a form of extraction. The global S is feeding the
• 11:31
  algorithm, but they don't own the models that result from it.
• 11:36
  They're the resource providers, not the architects.
• 11:38
  Which has led to this push for data sovereignty.
• 11:42
  Countries are waking up. They're realizing that having
• 11:45
  all their citizens data stored on servers in, say, Virginia or
• 11:49
  Dublin is a massive strategic vulnerability.
• 11:53
  So they're passing these data localization laws.
• 11:56
  Demanding that data generated inside their borders has to stay
• 11:59
  inside their borders. Which is creating what the
• 12:01
  sources call a splinter net, right?
• 12:03
  A fragmented digital world. Correct.
• 12:05
  And it makes global cooperation so much harder.
• 12:08
  The UN tried to tackle this with the Global Digital Compact, the
• 12:11
  GDC, back in September 2024. What was the goal with that?
• 12:14
  It's this comprehensive deal to try and close that digital
• 12:17
  divide to to make sure AI benefits everyone, not just
• 12:20
  Silicon Valley in Beijing. Did it work?
• 12:23
  Is it binding? That's where it gets
• 12:24
  complicated. The Southcenter source we read
• 12:26
  points out it's non binding, it's a set of aspirations and
• 12:29
  you see that same tension we talked about before.
• 12:31
  The global S versus the US. Right, the global S sees the UN
• 12:36
  is the only place they can get a fair shake.
• 12:39
  The US is deeply skeptical of UN LED tech governance, fearing it
• 12:43
  will slow innovation and give authoritarian regimes too much
• 12:46
  say. So we've got this fractured map.
• 12:49
  the US is building fences around hardware, China's building walls
• 12:53
  around content. Europe is building a fortress of
• 12:56
  rules. And the global S is trying to
• 12:58
  build a fence around their own data to keep it from being
• 13:01
  taken. That's a perfect summary, and it
• 13:03
  leads to maybe the most subtle part of all this narrative
• 13:07
  seeding. This is from the eternal source
• 13:09
  and it's mind bending. It's not just propaganda
• 13:12
  anymore. Deeper than that.
• 13:13
  So much deeper traditional propaganda tries to convince you
• 13:17
  the person narrative seeding is about shaping the training data
• 13:20
  of the large language models themselves.
• 13:22
  Wait, so you're not trying to convince the user, you're trying
• 13:24
  to convince the AI so that the AI will then convince the user?
• 13:28
  Exactly. You were trying to define the
• 13:30
  reality that the AI perceives and then regurgitates.
• 13:33
  If an AI is trained mostly on US centered text, guess what?
• 13:36
  It's going to produce US centric world views.
• 13:38
  A bias towards liberal democracy, capitalism, Western
• 13:42
  history. And if you're sitting in Moscow
• 13:45
  or Tehran, that looks like a cultural weapon pointed right at
• 13:48
  you. Which is why we're seeing this
• 13:50
  push for informational Atarki from places like Russia.
• 13:53
  Atarki self-sufficiency. They are trying to insulate
• 13:56
  their digital space. They want to build sovereign AI
• 14:00
  that doesn't rely on Western data, Western narratives.
• 14:03
  They don't want their citizens using a chat bot that thinks
• 14:07
  Crimea is part of Ukraine. They want a chat bot that
• 14:10
  reflects their official truth. We're seeing it in the Gulf
• 14:12
  states too, with the Falcon model in the UAE.
• 14:15
  They're building high quality Arabic language models.
• 14:19
  Yes, and it's not just about translation.
• 14:21
  It's about projecting cultural influence, about modernizing on
• 14:24
  their own terms, without having Silicon Valley define their
• 14:27
  digital culture for them. It's like a new kind of space
• 14:30
  race, but instead of racing to the moon, everyone is racing to
• 14:33
  define reality itself. That's what algorithmic
• 14:36
  diplomacy is. The diplomat of the past
• 14:39
  negotiated borders on a physical map.
• 14:41
  The diplomat of the future might negotiate the weights and biases
• 14:44
  in a neural network. That is a heavy thought, so
• 14:48
  bringing it all home, what does this mean for you?
• 14:50
  Our listener may be sitting in Copenhagen or anywhere else
• 14:53
  watching this unfold. I think it means we have to get
• 14:56
  used to the idea of digital borders.
• 14:59
  We're used to showing a passport to cross a line on a map.
• 15:02
  Now our data has to show a passport.
• 15:04
  And for anyone working in this field, diplomats, analysts, it
• 15:07
  sounds like the job description is changing overnight.
• 15:10
  It absolutely is. The diplomat of the future needs
• 15:13
  AI literacy as much as they need to speak foreign languages.
• 15:17
  You can't negotiate a trade deal if you don't understand how the
• 15:20
  algorithms monitoring that deal actually work.
• 15:23
  And if you don't understand it, you're going to be manipulated
• 15:25
  by the person who does. Precisely.
• 15:27
  Which leads to one final thought.
• 15:29
  The biggest dilemma in all of this for me is the question of
• 15:32
  meaningful human control. The human in a loop.
• 15:35
  Yes, as we hand over more analysis, more drafting, more
• 15:38
  simulation of these AI systems, we have to make sure that moral
• 15:42
  and legal responsibility stays with a human.
• 15:46
  These algorithms are often black boxes.
• 15:48
  We don't always know how they reached a conclusion.
• 15:50
  So if a diplomatic bot makes a recommendation that escalates a
• 15:54
  crisis, who's responsible? The coder?
• 15:58
  The diplomat who used it? The country.
• 16:00
  That is the trillion dollar question.
• 16:03
  We can't let ourselves get to a place where we can just say the
• 16:06
  algorithm made me do. It we can't.
• 16:08
  But the temptation to rely on these incredibly powerful tools
• 16:12
  will be immense. Well, on that slightly
• 16:15
  terrifying but also exhilarating note, I think that's where we
• 16:17
  have to leave it for today. It is a brave new world.
• 16:20
  It certainly is. If you enjoyed this analysis and
• 16:23
  want to understand more about these huge shifts in global
• 16:26
  power, please make sure to follow the podcast on whatever
• 16:29
  platform you're using. And hey, if you have a moment,
• 16:31
  rate the show 5 stars. It really does help us reach
• 16:34
  more people. Thanks for listening to
• 16:36
  Diplomacy and International Relations by Bastian Freiburg.
• 16:39
  We'll see you next time.