Last week, Argentina’s government unveiled something it called a “social digital twin.” The idea: a massive AI simulation of the country’s economy, demographics, and public sentiment — a crystal ball for policymaking, capable of predicting the consequences of every tax hike, subsidy cut, or welfare reform before it hits the streets. The announcement came via a slick video. And within minutes, anyone with half a brain and a pair of functioning eyes could see it was a mess.
Let me be clear: I’ve been writing about AI and government tech for over a decade. I’ve seen the UK’s NHS spend billions on digital flops. I’ve watched the EU try to regulate its way into innovation. I’ve tracked the metaverse hype cycle from peak delusion to the trough of “wait, who’s still wearing a headset?” But this? This is a new kind of special.
The video — posted on X, because of course it was — features what appears to be a deepfake of Argentina’s Minister of Deregulation and State Transformation, Federico Sturzenegger. The real Sturzenegger is a respected economist. The video version? A slightly waxy, slightly uncanny simulacrum, mouthing words that may or may not have come from the actual man. The government hasn’t confirmed it’s a deepfake. They also haven’t denied it. Which, in the world of PR, is a confession.
But the deepfake is almost the least of it. The video is littered with typos. Grammatical errors. Misspelled words. The kind of stuff that would get a high school student a red pen massacre, let alone a national government announcing a flagship AI project. One slide reads “invesment” instead of “investment.” Another misuses Spanish syntax in ways that suggest the script was run through a free translator and no one bothered to check. The AI couldn’t predict a typo. It couldn’t even predict that someone would read the screen.
I think that’s the part that gets me. Not the hubris — we expect hubris from politicians. Not the technical overreach — that’s par for the course. But the carelessness. The sheer, sloppy indifference to looking competent. If you’re going to announce an AI that will simulate the future of 46 million people, maybe — just maybe — spellcheck the PowerPoint first.
A Digital Twin of What, Exactly?
Let’s back up. What is a “social digital twin”? The term comes from engineering, where digital twins are virtual replicas of physical systems — a jet engine, a factory floor, a power grid — updated in real time with sensor data. They’re used to run simulations, predict failures, and optimize performance. Boeing uses them. Siemens uses them. The idea of applying that to a whole country is… ambitious.
Ambitious in the way that trying to build a scale model of the ocean in your bathtub is ambitious. The complexity of a national economy — with its millions of actors, irrational behaviors, black markets, climate shocks, and sheer chaos — is not something you model with a few terabytes of past data and a neural network. It’s not even something you model with a supercomputer. At best, you get a toy. At worst, you get a tool for confirmation bias: you feed in your assumptions, the AI spits them back as predictions, and you call it science.
Milei’s government, which came to power on a wave of anti-establishment fury and promises to “chainsaw” the state, seems to think this is the future of governance. The social digital twin, they claim, will allow them to test policy changes in a risk-free virtual environment before implementing them in the real world. Sounds good, right? Who wouldn’t want to see if a tax cut actually stimulates growth or just inflates asset bubbles before making it law?
But here’s the problem: the model is only as good as its assumptions. And the assumptions are made by people — people with political agendas, ideological blind spots, and, evidently, a troubling disregard for proofreading. If the AI is trained on historical data from Argentina’s last few decades of economic turmoil, it will learn that Argentina defaults on debt every few years. That’s not a prediction. That’s a pattern. It won’t tell you how to break it.
Deepfakes and the Credibility Gap
The deepfake element is worth dwelling on, because it reveals something about the government’s relationship with truth. In a normal administration, if you want to announce a major policy initiative, you have the minister stand in front of a camera and talk. If the minister is unavailable, you have a spokesperson read a statement. You do not create a synthetic video of the minister and pretend it’s real. Unless, of course, you’re trying to normalize the idea that official communications can be generated by AI.
I’ve been covering deepfakes since the early days, when they were mostly used for non-consensual pornography and celebrity pranks. Now they’re being deployed by governments. The technology itself isn’t evil — it’s a tool. But using it to simulate a real person delivering a policy announcement, without clearly labeling it as synthetic, is a dangerous precedent. It erodes trust. It makes citizens wonder: Did the minister actually say that? Is this real? And once you start asking those questions, you stop believing anything.
Argentina already has an inflation rate north of 200%. Its poverty rate is over 40%. People are desperate for solutions. They don’t need to wonder if the person telling them about those solutions is a digital puppet.
The Typo That Tells the Story
I keep coming back to the typos. They’re funny, sure. But they’re also a symptom. A symptom of a culture that values the appearance of innovation over the substance of it. Of a government that thinks AI is a magic wand you wave over problems, not a discipline that requires rigor, testing, and — yes — basic copy editing.
Consider the optics: a video touting the most advanced AI simulation ever attempted for public policy, and it can’t even get the word “investment” right. It’s like unveiling a spaceship and forgetting to put doors on it. It’s a metaphor for the whole project: big claims, flashy visuals, but the fundamentals are hollow.
I asked a friend who works in AI policy what she thought. She laughed. Then she sighed. “This is what happens when you let tech bros run a country,” she said. “They think you can just GPT your way to good governance.” She’s not wrong. There’s a growing movement of “effective accelerationism” — e/acc, for those of you keeping score at home — that believes AI should be deployed as fast as possible, consequences be damned. Milei’s government seems to have bought into that philosophy wholesale.
What Could Go Wrong?
Let’s imagine, for a moment, that the social digital twin actually works. Not perfectly, but well enough to influence real policy. What happens when the AI predicts that cutting welfare will boost GDP? The government cuts welfare. The AI was right. But what if the AI was right only because its training data reflected a neoliberal bias embedded by its creators? Then you’re not optimizing policy. You’re optimizing for your own ideology, wrapped in a cloak of mathematical objectivity.
What happens when the AI predicts that raising interest rates will curb inflation, but the model doesn’t account for the fact that Argentina’s central bank has no credibility and people hoard dollars under their mattresses? The policy fails. The AI gets retrained. But in the meantime, real people lost their savings.
This is the danger of algorithmic governance: it looks scientific, but it’s often just prejudice with a graph. And when the AI makes a mistake, who takes responsibility? The minister? The developer? The algorithm itself? In a democracy, accountability is supposed to be clear. Elections are supposed to matter. But if you’ve outsourced decision-making to a black box, you’ve outsourced accountability too.
I’m not saying AI has no role in government. It does. It can help analyze traffic patterns, optimize energy grids, flag fraudulent transactions. But using it to simulate the entire social and economic fabric of a nation, and then making policy based on that simulation, is a leap too far. It’s not just technically questionable. It’s politically dangerous.
The Hype Cycle Comes to Buenos Aires
We’ve seen this before. Every few years, some government announces a moonshot tech project. A digital ID system for all citizens. A blockchain-based land registry. A city built from scratch in the desert. They promise efficiency, transparency, a break from the past. And then, quietly, they abandon it when the money runs out or the vendor fails to deliver.
Argentina’s social digital twin will probably follow the same path. A few pilot programs. A lot of press releases. Then, when the model fails to predict the next debt crisis or the next protest, it will be quietly shelved, and the officials responsible will move on to consulting gigs. The taxpayers will foot the bill. The tech press will write a few retrospectives. And the cycle will repeat.
But what strikes me here is the timing. Argentina is in crisis. People are hurting. They need real solutions — not virtual ones. They need policies that address the root causes of inflation, not a digital mirror that reflects their suffering back at them. The social digital twin isn’t a tool for change. It’s a distraction. A shiny object dangled in front of a desperate public.
And the typos? They’re not just embarrassing. They’re a tell. A sign that no one in that government actually believes in this project enough to do the boring, unglamorous work of getting it right. Because if they did, they’d have hired a copy editor before they hit publish.
I write this as someone who genuinely believes technology can improve governance. I’ve seen it happen — in Estonia, in Taiwan, in places where digital tools are deployed with care, transparency, and a healthy dose of skepticism. But Argentina’s approach is the opposite of that. It’s hype over substance. It’s shortcuts over systems. It’s a deepfake minister selling a half-baked simulation to a country that deserves better.
So here’s my advice to the Milei administration: scrap the video. Fix the typos. Be honest about what the AI can and cannot do. And if you really want to predict the future, start by listening to the people who are living through the present. They’ll tell you more than any digital twin ever could.
Further Reading
Read the original story at Decrypt
Original source: read the full article