Audrey Tang

Yes, I'm very surprised that here almost nobody knows that ChatGPT has a group chat mode. They rolled it out as a pilot first in Taiwan and South Korea, New Zealand and Japan, but now it's generally available. People just don't use group chats with a language model, but actually in group chat mode, it's much more pro-social. It's not trying to lure you in. It's trying to restore the relational health between people.

Yes, free red teaming.

Well, I think for many people in democracies, we hear this refrain: democracy only leads to chaos and never delivers. And every year, there's a variation on the theme. So, for example, in 2020, we received a lot of information manipulation that says, you know, only lockdown, top-down, shutdown, takedown can contain the virus. And democracies only lead to chaos. And Taiwan, of course — we're a very liberal democracy — will suffer a lot. It's inevitable that we will succumb to the virus. So, if the people do believe that, like a prophecy, then they give up.

But actually, Taiwan only lost seven people to COVID-19 in 2020. And throughout the three years, we never locked down any city. And then we reported historically high GDP growth during that period, because we never locked down any city, so TSMC kept running. So, my point being, democracy actually coordinates better than autocracies. But autocracies want people in democracies to believe that only autocracies can control the virus — cyber virus, bio virus, or whatever. And in fact, because …

Right, exactly. And because the virus actually mutates very quickly, the steering wheel needs to follow how the virus is mutating. So, actually, when Omicron came, Taiwan steered very quickly to counter that. Whereas, like in Shanghai, where they still had lockdown, they had a very terrible situation vis-à-vis the virus. So, although top-down, shutdown, takedown is very quick, it also steers less flexibly. And I think that's the counter-message we have. That is to say, democracy is a social technology for coordination. And the more people invest their time and energy in it, the better it gets, like any other social technology.

So, I think the information manipulation can come from anywhere that believes autocracy is the only solution to contain emerging threats. And we need to counter it by showing people that civic coordination can actually steer better.

Have you read the AI 2027 report? It's by a former OpenAI researcher, and it says by 2027 — I think they revised it to 2029 or 2030 now — AI systems will be able to recursively self-improve. So, it won't need humans anymore to train the next generation of robots. And at that point, it will just escape outside of control. So, my point being, the entire world is now looking at a timeline. And people in Silicon Valley truly believe, I think many of them, that we're in a takeoff. And so whether it is about the timeline for kinetic escalation or for cyber escalation or for bio escalation, many people are feeling like these following few years are the very short window where we still have the potential to steer things right — outside of which AI is a runaway.

So, I think my job as cyber ambassador — cyber, by the way, comes from the Greek, kybernan, which literally means to steer, right — is, in the next few years, to have everyone imagine a better landing. So, it's not just a safe takeoff, but also where to land. And we really need to land — to strengthen trust not just within communities, but across communities, so that we can wage peace together. We will not fall into, like, false escalation just because of some misunderstanding. We'll be able then to socially translate — like the climate justice communities and the creation-care communities. So, they realise: oh, we're working on different segments of the same wall. We're building the same wall together.

I think the Pope just last week in the encyclical said that we shouldn't just build this Tower of Babel, like a superintelligence — that we should build this wall together. I think this is a wall by many, not a tower by a few. I think it's a very strong message.

Well, I mean, for me, any autocracy that truly believes "democracy only leads to chaos and never delivers" cannot be countered on its own terms. If you do that, you're already thinking like an autocrat. Instead, think about how we can train a better system that responds to emergencies even faster than the autocracies. So, some position it as a race. Somebody has to build a Tower of Babel quicker than the other Tower of Babel. A tale of two towers. But I think that's fundamentally not where we're going. Not Taiwan, not Europe, and not, I would say, most people in either Beijing or in Washington.

I think most people would much prefer if we put technology into service — AI in the loop of humanity, not humans in the loop of AI, like a hamster on a wheel. I think this massive acceleration toward zero-sum or even negative-sum racing is bad for everyone, which is why I don't counter autocratic messages on their own terms.

Like, "we need to be following a strong leader in order to counter the threats." And then, "maybe it's okay to give up some liberty because the virus is very scary." If you start on that path, then you're already thinking like an autocrat.

Well, I think that they first and foremost cherish freedom, right? So, the freedom of expression, of thinking, of association, to them, as I would argue for most people in all democracies, is what the job of democracy is about. It's not just about uploading three bits every four years, namely voting. But rather, whenever you have a good idea of how to make a community better, you can voice it without waiting for four more years. So, that's the point of having civic muscle in a democracy.

And so I think they mostly feel that Taiwan maybe is a kind of demonstration, a demo, that we do not need to fear technology. Because around the time they gave birth to me, Taiwan also gave birth to the personal computing movement, the IBM PC. Many of them were made in Taiwan, PC compatibles. And so prior to the '80s, people didn't really have computers. People had terminals. You typed into a terminal. Everything you typed went into this cloud mainframe — a large financial institution or state or whatever, research institution. The system operators saw everything you typed. There was no privacy. And then you were at their whim as to what software to install.

But then came personal computing. And people could compute in private — their spreadsheets, desktop publishing. So, it's like having a printing press for everyone, right? And it's very powerful. And so I think the point they really made to me is that it is this freedom of expression, of association, this communal civic muscle, that is the true backbone of Taiwan's democracy. It is not just about voting every four years. I think this message is demonstrated conclusively, I would say, in Taiwan. And now we're seeing many democracies really taking this to heart.

Yes, exactly. And that's why the work with SPRIND is so important. Because they literally just ask people "What sucks?" And it's very easy actually to complain. It's the most democratic question. Yes, and the trick here is using Civic AI. You can make people feel truly heard, like closing the loop. If the federal government or your state changes some policy because of your input, then for the first time you can have not just broadcasting, as what we did, but broad listening. It is because of these things you said that we did this. And it's thanks to you that now we are asking you another follow-up question. That's truly massive multi-scale listening. It's finally possible and easy.

And so we need to upgrade how the civic muscle works in the age of AI. Otherwise, it's just people broadcasting messages — just, like, slop. People would not even pay attention, because there's no relational nutrition. But if you're truly heard and the loop goes back to you, then people are much more willing to share problems and ideas on how to make things better.

Well, at Computex, which is just now in Taiwan, Jensen Huang called this the reinvention of personal computing, right? The point is that instead of a personal computer requiring you to install some app, type in some commands and set it up and get IT support or something, it should actually be something fiduciary, loyal to only you and your relationships. So, it's a form of personal intelligence, and I think it's a great vision, because it means, just like the origin of personal computing, no matter how many data centres are out there, they're never going to extract oil like our personal data. We're going to till our data as soil. And this soil regenerates itself. And I think this vision is far more preferable to the Tower of Babel. And I'm very proud that Taiwan is now making this possible — to make it really happen for a lot of people.

Well, I mean, even the Catholic Church has changed, right? Since the Industrial Revolution, the church has emphasised the idea of subsidiarity in Catholic social teaching, because at the time they also saw capital accumulating, aggregating far quicker than what unions and associations and co-ops could handle. So, they basically said subsidiarity — the ability for local people to settle local matters using only local tools — is what the church is now about. So, this over-centralisation has changed. Now, the Pope also says something very similar in the age of AI.

So, I think people by and large — like, you're using an iPhone and you're recording, and the iPhone can make a transcript of this recording in airplane mode. Right. It doesn't really need a cloud. Everything stays in your iPhone.

And if you have a Mac computer, you can run, as I do, local AI. Again, using Apple chips, you can run faster than Claude Sonnet, more accurately than Claude Sonnet, and it only serves you.

So, on quality, on speed, on ease of access, there's really no reason to use cloud models if you're not doing a very complex world simulation or things like that. For personal and communal uses, the local models are more than enough. And so I think it's just awareness. It's not there yet.

That's right.

We were touching upon that just a few minutes ago, right? I think many people confuse content moderation and censorship, whatever, with this interoperability commitment. If you confuse the two, then it all sounds like stifling innovation or something. But actually, interoperability is good for innovation. If people can change from one telecom to another without changing their number, then of course the telecom has to innovate more. If you can switch from Bluesky to W, then of course Bluesky has to innovate more to keep my business. So, the freedom to move in the cyber realm, once it's protected as a human right — most of the other digital rights follow from that interoperability. So, I think it's very good that Europe is also looking into interoperability laws.

But I think many people in the policy space confuse that with censorship, with stifling innovation and so on. But I would say an information superhighway without an off-ramp is just a hamster wheel. It's not going anywhere.

I think for many people in Taiwan, because we grew up with personal computers, we see computers and tech in general as serving our communities first. Whereas here in the UK and Europe, sometimes people correctly fear that the technologies are really serving some masters in the cloud. So, I think it is also about social media. In Taiwan, many people's first experience with social media is PTT, which is like Reddit, but it's run by National Taiwan University. So, it's part of the academic network. It has no advertisers, no shareholders, no extraction at all. And the people make their own rules in self-governance.

But for many people here, when they first encounter social media, it is running in some abstract data centre in some other country, and it wants to figure out what keeps you addicted and keeps feeding you polarising things.

Right, exactly. So, like with PTT and other civic spaces, public squares, the more you participate, the more steering power you have. And you grow the civic muscle. Whereas in the extractive social media, the more you get sucked in, the less control you have over where it's going.

It's actually very easy, because the cloud companies did all the interface design. So, if you switch to Proton Meet, Proton Mail, Proton Drive, Proton Pass, it works exactly the way you imagined it would work.

Yes, so you don't need to personally install Linux or something. There are Taiwanese entrepreneurs in Switzerland — Andy Yen of Proton. And Signal, if you use it, is the same as WhatsApp. Actually, for most of the things that we use, there are European and civic and free-software equivalents that have exactly the same user experience. And so really there's no friction to switch.

Well, I think many people, if they enter a market that already has this social-licence-to-operate idea, are willing, of course, to negotiate and even be good model citizens and follow the social licence to operate. I travelled to India a few months back, and even the Tibetan exile people can use the UPI, the Unified Payments Interface, to transact, to open a business and so on. And that's digital public infrastructure. The point is that the Indian people said to Google Pay, Meta Pay, PayPal, whatever pay: if you want to enter the Indian market, all your QR codes must be scannable by every competitor of yours. You cannot lock anyone in. There's no vendor lock-in whatsoever for payments. And it's just a ground rule.

So, it's not like they say, oh, this is a national champion. No, this is about saying: we regulate you like a utility, and this is our ground rule. And this is what Europe can do, obviously.

Yes, definitely. I think no democracy is an island, not even Taiwan. So, we need to keep working hard and free the future — together.

Exactly. It's not just a shield of chips — or of bubble tea — but also of trust. And to give no trust to your people is to get no trust. It's a very simple lesson. And here in the U.K., we're also seeing, like, the National Strategy Project and so on really looking at Taiwan and taking some of those deliberative polling ideas to work here in the U.K.

Well, I travelled to 28 countries in the past couple of years.

I'm geo-flexible.

Well, actually not more than a couple of weeks in each place. So, I'm going in a couple of days to Gothenburg, Sweden, and after that, Berlin — many, many places.

It's for bio:cap, so in four or five days.

I think it's up to my speaking agent.

During a recent visit to Dharamsala in northern India, I came across a scene that has stayed with me ever since.

Tibetan merchants, who had long found the doors of formal financial services closed to them, can now — with a payment QR code printed from an e-wallet — plug their small shops into India's Unified Payments Interface, or UPI. This is more than payment convenience; it is digital public infrastructure, or DPI, at work.

Around the world, a growing number of displaced people are kept outside formal financial systems; they have little choice but to eke out a living in the grey economy. In India, the state lowered the barriers to participation, allowing people to build on the trust and community ties they already held.

This is the reminder the AI era needs most: Meet people where they are, not where a platform wants them to be.

Many worry that AI is assuming the shape of a "digital coloniser." A handful of tech giants control compute, models and platform rules, while the rest of us simply rent intelligence. Once a platform pushes an update or adjusts its pricing, even a business may find that its own workflow has been quietly "extracted."

Underneath this entirely reasonable anxiety lies a deeper question about how we understand the nature of "data."

Treating data as oil leads to one logic: extraction and concentration. I prefer a brighter picture: Data is not oil, it is soil. AI models are crops, with the key questions: Who tends the data ecology and who mends the system in the event of an error.

Taiwan's richest soil lies on the shop floors and factory floors of every trade. The owners of many small and medium enterprises, or SMEs, carry irreplaceable tacit knowledge. When a senior technician states that a machine "sounds off," that wisdom is usually understood only by those who work alongside the machine. That touch, that judgement, is the core capability Taiwan's SMEs have built up over decades.

The problem is that much of this knowledge is lost as the masters retire.

If AI remains under the control of a few large platforms, SMEs are left buying solutions designed by others. But imagine public infrastructure for data innovation, model-training frameworks, interactive interfaces and evaluation benchmarks, which any workshop could use — the same way UPI is available to any merchant with an e-wallet.

The masters need not surrender their trade secrets. They can instead calibrate the model according to their own standards, turning it into a site-specific "machine apprentice." For perhaps the first time, those judgements that are almost impossible to express in words can be fully passed on.