Audrey Tang

Yes, so far. And in 2016, when people asked me — when the HR people in the cabinet asked me — what does a digital minister even mean? I said to them that digital in Taiwan means shùwèi — means not just digital, but also means plural. So I said actually I'm a digital minister to steer away from singularity, which is over-concentration of power, machines becoming super-intelligent and leaving us all behind, toward plurality, which is technology to foster collaborative diversity and swift trust in the society. So I wrote my own job description, which is a very short poem.

Yes, it's very short. I can recite for you. Goes like this.

When we see internet of things, let's make it an internet of beings. When we see virtual reality, let's make it a shared reality. When we see machine learning, let's make it collaborative learning. And when we see user experience, let's make it about human experience. And whenever we hear that the singularity is near, let's always remember the plurality is here.

That is exactly right. We have secondments from pretty much all ministries. And in each ministry we also have a satellite team called participation officers that is in charge of not broadcasting to people, but broad listening from people. In 2020, when the pandemic hits, many people predicted Taiwan, because of our vicinity to Wuhan, will not survive the first wave. Well, we only lost seven people in 2020. We never locked down in the three years, mostly because we run a very tight feedback loop. Anybody can call 1922, the rapid-response line, and suggest what to do next and what are we not doing well. And the people's ideas — like privacy-preserving contact tracing, visualization of mask availability, and so on — became national law and infrastructure within like a week or something. So the ability for us to listen broadly, I would say, is one of the best ways to counter against the unknown unknowns as posed by the infodemic, by the pandemic, and now by deepfakes and other malicious AI swarms and whatnot.

I call myself a servant to the public service. I was a national tech support — the OG DOGE. And the tech support, again, is not about disrupting as in breaking things, as in taking away bureaucracy. It was instead saying everyone in our society should have almost like a shared reality of the what and the how of policy-making, not just the what of the policies made. So to be very concrete: any 5,000 people joining in the online petition can force a ministerial conversation. If people feel that the tax-filing software takes too long, like three hours, they can propose how to shorten it to three minutes, and we made that happen. Very young people like 14-, 15-year-olds, even before they can vote, they go to our national participation platform and say we should go to school one hour later, because study shows one more hour sleep gets to a better grade than one more hour of study. That's a very good idea, and so they've got that as well. We ban plastic straws from our bubble tea, our national drink, boba takeouts. That was also driven by 16-year-olds. And many of them, like I was under 35, became what we call reverse mentors — cabinet-level advisors to ministers — who are under 35, to ministers who are over 35. But it is the younger people, like a Pygmalion effect. We expect them to steer the national direction, and they went ahead and did that.

Well, I would say that our 14-year-olds, according to the ICCS study a few years ago, are now the top of the world when it comes to — they are not just civic knowledge, but civic muscle. Way before they turn 18 they feel that they understand why the politics is doing what it is doing, and they feel that they can set a national agenda and steer the country to a better place. According to BTI, more than 90% of people in Taiwan think democracy is fairly good. And so even if we're still pretty polarized ideology-wise, we turn that polarization into fuel, almost like a geothermal engine — so not a volcanic fire to evacuate away from, but we can tap, harness that polarization into energy, into democratic renewal. I would say that's the legacy I'm the most proud of.

Yeah, definitely. We're still maintaining a very high voter turnout. But even setting that, like more than 70%, aside, it is the in-between elections — not just a low-bandwidth participation of one referendum vote, one bit, one presidential vote, maybe two bits and so on — but rather like every day there is an e-petition, there is a participatory budgeting, there is mini-public, citizen assembly, as I mentioned. Many, many participation happening in a way that trains the civic muscle in all high schools and in all long-term care centers and cultural centers in our Indigenous nations and so on. So this horizontal takeoff — that is to say, people seeing technology strengthening existing community bonds and linking together the disparate communities — I think that is the real democracy, where the real swift trust is happening, and the voting booth is included but not a particularly prominent part of it.

Yes.

It's a great bridging mechanism. I totally agree. We're also experimenting in Taiwan not just with electronic budget participation, but also priority-setting participation. We also experiment with what's called quadratic voting, which is each person not only has one vote but have 99 points, and they can allocate across like 200 or so of projects. But the trick is: one vote costs you 1 point, two votes 4 points, three votes 9 points. So it's 99 points. You cannot actually vote 10, because that would cost 100. So people will have to discover the synergies. And just like compulsory voting, I would say that it again prioritizes the bridging voice, the overlapping voice, not the outrage.

Well, across the Pacific there are, last I checked, 50 labs of democracy. And some of them are becoming frontier labs. I'm talking about the state-level democracy. Just last month, California legislature is now going through — and I think it's out of the committee now — a bill that will make something like the Taiwan model, the e-petition model. They call it Engaged California, into a permanent fixture of their executive branch. They already use this kind of bridge-making algorithm to listen broadly, to overcome, for example, mitigation and prevention of future wildfire in Los Angeles. They also engage their staff within the California employees to make sure that state employees can introduce AI in such a way that strengthens it instead of taking away human judgment and human trust.

And they are now using it to figure out how to impose duties on the AI labs, frontier labs, housed many of them in California, so that they can not just replace human work, meaningful work. They have to somehow preserve and advance apprenticeship, belonging, care and dignity. So again this kind of direct digital democracy — Taiwan is no longer the largest polity. California and Japan both are much larger than Taiwan, and are advancing democracy. So while I agree with you that people are dissatisfied with traditional partisanship, once every three or four years election cycle, I would say that's just a kind of legacy version of democracy, and a new version of democracy upgrade is coming.

Yeah, I think it's really, really nice that people are now seeing that we cannot just fight social media from individual action. It's just not possible. Although I found a lot of personal success by turning all my screens grayscale. But—

That's exactly right. So now people are finally seeing it is the PPM that matters — and not just carbon parts per million, but also polarization per minute that is steadily rising in the antisocial corner of social media. So as soon as we shift from this, like individual grayscale, putting away our phones, switching to touch screens with stylus and not with our fingers — I mean, life hacks that people can deploy — but just like climate change, these are insufficient, and really the only way is to start measuring PPM. And once you measure, you can internalize the externalities and really make those companies pay, like the fridge-makers of Freon back when the Montreal Protocol was signed. The protocol became what's called technology-forcing regulation, because it says essentially in a few years if you continue to make fridge that destroy the ozone, you will be dealt with like a crime against humanity and environment as you are. Right. So we're now finding ourselves at the Montreal Protocol point when it comes to social media.

Yes, and just like ozone, brains can also heal. It's called neuroplasticity. We can also figure out how to, exactly. Yeah. Because I'm a very competitive sleeper. I sleep for eight hours every day. And if I get sufficient number of hours of sleep, I literally feel my brain renews itself. And so whatever pessimistic or stress-inducing news I got during the night, if I sleep well, I wake up thinking, oh, actually that's fuel. We can turn polarization into energy.

Yeah, exactly. Yeah, that's the direction. We need to become the kind of society that's not just that a decision-maker sleeps better, but everyone sleep better. And I really think that it will heal the social fabric if we can reorient technology toward that.

A few years ago we banned all small screens, touchscreen or not, for young people in schools. But we complemented that with providing one laptop per child, or large tablets. And we always use technology in shareable screens. So in classrooms kids learn about programming, about generative AI or whatever, always with other people. So again this is like a campfire that lights, illuminates our differences and allow for small-group conversation, and it's never the kind of wildfire that leaves young people shadowboxing with the outraged, caricatured version of some other people. So I think as long as it's coupled with something that is prosocial — instead of banning all media, you ban just the antisocial corner, the antisocial use of social media, and then you have good pro-social communitarian media — I think that's very good. And in a way that doesn't compromise privacy, that would be even better.

Yes, and also scroll less, sleep more.

Indeed, which is why I think pro-social media is a must. Not a nice-to-have. Imagine in Taiwan if we just ban screens altogether from schools and everybody goes back to, I don't know, pen and paper, not even calculators. Of course Taiwanese students wouldn't be able to retain our like number-three to number-five place on OECD PISA. Right. So of course we have to use technology in education. But again, the technology needs to serve the public interest, and therefore we need to have very clear rubric, including the PPM, polarization per minute, and measure what the technology is doing to social cohesion in such a fabric.

Well. I would continue on the work of Community Notes, which was already imported from Twitter, now X.com, the algorithm that we deploy in Taiwan that surfaced the unlikely consensus from Uber supporter and taxi supporter, and has inspired a new wave of bridging systems including Community Notes. So if you look at X.com and other social media, you see a context — that context, that note, are done by people on both sides that kind of red-team each other and finally found something they can all live with. And in X.com, they're already getting Grok and other language models to draft such notes. So instead of something that comes too little too late, they would now attach almost immediately when Elon or somebody makes a controversial post. And so I would immediately take that to Facebook, and then also change the feed so that we can say, dear algo, I want to see bridging posts first. And then the language model will be able to translate between, for example, climate-justice language and biblical creation-care language, and ensure that both are now seeing each other kind of eye to eye. And they see that they are doing the same object-level actions. And we have worked with the New Pluralists Institute in the US and Google Jigsaw to show that the US, after they get into the social translation, in the social layer, actually more than 97% of people agree on the core ideas, and even the most controversial ideas get almost 70% people support. So again, the polarization is largely an illusion, a function of the high-PPM feed design. So just having a better information diet goes a very long way.

Yes, I think that is a big part. And the good news is that people are going toward conversational interface now. I think across all age groups people are leaving infinite scrolling in droves, which is, I guess, a good thing, because at least with the conversational interface we can have people look at each other's contributions through translation, instead of the infinite scrolling, which is almost at peak slop now. There's no relational nutrition anymore, because a huge amount of it is machine generated synthetic media.

Well, as I mentioned, Taiwanese people are extremely divided on our relationship to Beijing. I remember when I first became digital minister, there was an 8,000-strong petition to change our time zone away from plus 8, which is Beijing time zone, to plus 9, which would be Japan time zone, and moving us closer to Australia. And another 8,000 people petitioned to keep us in plus 8. Of course, we cannot strike a compromise.

Yeah, yeah, exactly. That would be a compromise and leave everybody much worse off. So it's like the downwind, not the upwind, of political solutions. So we deploy the bridging algorithm, inviting the leaders from both petitions in the same room. And we figure out actually they both want Taiwan to be seen as more unique in the world, to have wildly different cost-benefit ratio analysis. So we calculated exactly how much it would cost recurringly for the time-zone change, and say, okay, if we're ready to spend this money on making Taiwan seem more unique in the world, surely there are better ways. And so they start brainstorming, and we say, oh, actually, let's give a Gold Card visa for people who contribute to open source for more than three years. Let's give them free residency and universal healthcare. Very good idea. It doesn't cost much. Let's house the international human rights conference, the RightsCon, to advocate for our friendliness to marginalized groups, so on and so forth. So all together, this actually spent less than the time-zone change, but both sides are much happier. So my point being: consensus is not just finding the plus-0.5-hour compromise. It is actually stepping back a little bit and saying what is the rare undiscovered common ground about the shared values, not just the shared instrumental actions. And once people elevate from this utility-based instrumental conversation into a more care-based conversation, then people can actually see each other much better, and also entertain their neuroplasticity and heal the social fabric.

Certainly. I think we the people are already the superintelligence. We have this latent ability to self-organize, not just within community, but across large distances in different communities. We have this innate ability to empathize and to do collective action. And so I think language models in particular can translate not just languages, but also cultural and ideological distances, like the climate change and the biblical creation care. We have seen many successes in Taiwan and now elsewhere deploying this. Like in Japan, there's now a new political party, the Future Party, Team Mirai, that explicitly runs plurality, the language-model-based social translation, as their way to have a covenant with the future — to translate the interest of future generations who cannot vote yet with the current generation so they can agree with each other. I think that's a very beautiful thing. Previously it was very difficult to do that, to close the loop, to make sure everybody feels that my two cents actually changed, moved the needle in this direction. And I understand it in my own language, my own culture, exactly how my contribution made sense. Previously we had to speak really good bureaucratic English or bureaucratic Mandarin in order to know exactly how your ideas changed your polity. But now this can be done almost with no marginal cost. So I'm very optimistic about the judicial use of language technologies in narrow models — not a super-giant model that consumes electricity to fold proteins and fold laundry and for a lot of other things, but rather specific small models that foster this kind of inter- and intra-group cohesion.

When I was flying to Madrid from Taipei yesterday, I was watching the monitor, and there's a small line underneath the airline cabin monitor. It says, and I quote, "stow and latch your monitor during takeoff and landing." So my main fear is that we would be stuck in doomscrolling and forget to stow and latch our monitors during the takeoff of artificial intelligence in the society. Because in that case, then AI would amplify the distrust, the hate, the ideological dunking of each other. And we will lose the ability to coordinate intra-group and across group. And the next pandemic or infodemic or cyber attack or bioterrorism or whatever will leave very difficult-to-recover damage. So I think still stow and latch our monitors during takeoff. It's very important. And then we can think about where are we landing.

That's a great question. I think it's very similar to the situation around the pandemic. Well, first of all, don't panic. Because when people rush into decisions and policies during a pandemic, it usually backfires, and actually the solution may be worse than the problem. So don't panic. That's number one. Like, scroll less, sleep more. But that aside, I think one of the main constructive things we — especially in free and open societies — is to advocate the use of what I call Civic AI, or bounded, locally answerable models, where the data is treated as soil, so we can regenerate it. If the Taiwanese Indigenous nations — there's 16 of them, 42 languages — feel our national sovereign model doesn't do the Indigenous nation justice, because the entire pipeline is open and the local data is not extracted, so they can always retrain and make sure that the alignment happens to their community context within like a day or two, instead of having to wait for a year for some extractive companies somewhere else on the planet to finally hear their plea or something. So the refocus on locally answerable bounded models is happening now. And the more we say data is not oil, data is soil, the more steerability we have on AI's impact to the society. So just think of the deepfake-scam case that Taiwan has successfully drawn the red lines. We used small language models accountable to the Taiwanese population to draw the red lines around the large deepfake models, the malicious AI swarm, and you can repeat this exercise to pretty much every aspect of AI's impact to the society. And in doing so we exercise our civic muscle. Our community gets stronger, so that as AI scales toward even larger capability gains, so does our governance. And this kind of scalable governance will then take us in a global horizontal takeoff, instead of a vertical takeoff that blows up and leaves everybody behind.

Yes, so panic together. I think that's the point: to make it common knowledge, and then turn that into collective action.

Yeah, that's a great question. I would always say that younger people know more about the future than we do. So please lead the country and we will follow. That's what I always say as minister to our young people. And in 2016, when we were doing our curriculum redesign, that was also the year of AlphaGo. So we also saw Move 37, done by, well, a Taiwanese researcher who could play pretty good Go but not quite at professional level. So he trained AlphaGo to play at a much better level. And we learned from Dr. Aja Huang and Demis's work that very soon AI systems will be able to excel beyond humanity when it comes to the utilitarian — that's to say, score better on the Go board — and the deontological, that is to say, to follow the rules of Go. And it's not even close. But we do not have to chase the horse, so to speak. We can ride a horse, right? So we change the curriculum to say we shift away from literacy, like memorization, the rote test that Taiwan and East Asia are famous for. We discard all these together, and then say only three things are going to matter now, and they are curiosity, collaboration and civic care, and nothing else matters. Which is why the emphasis on civic knowledge, civic competency and media literacy shifted to media competency, so people can produce meaning together. So the point here is that the future of jobs will be created by positions that emphasize apprenticeship, belonging, care and dignity. And if students and young people over-identify with the deontological following of rules and the utilitarian gaining of individual competitiveness, then of course they will panic — rightly so, because machines are going to take over that. But if they learn to ride a horse and focus only on the civic-care parts, I think they are going to have a much more fulfilling life.

Yes. I think the intelligence within a brain is overweighted. The genius is always just a small fraction of the genius, which is a genius of the collective of the organization. And AI systems can serve as connective tissue. We don't have to be just mitochondria of the artificial organisms. We can invert that situation and achieve what many people are calling symbiogenesis, the growing of our seniors with AI systems.

That's a great question. I would say that the Sunflower Movement in 2014 was really a watershed moment. In 2013 or earlier, if you ask a random person on the street, there would be like, you know, I don't want to hear about politics, participation doesn't work at all. It's elite capturing all the profits and all the decision-making power, and what can we do with it. So I think this is a very common refrain in the 2010s in Taiwan, especially among younger people. But what the Sunflower Movement have shown, however, is that we do not have to topple anything. In the words of Mr. Fuller: we do not need to destroy the old system. We can just make a new one and render the old one obsolete naturally. So the beauty of the Sunflower Movement, I believe, is that it shows just by participating in a social-media-like interface, just by showing up in a video-conference-like interface, just by participating in a group-chat-like interface, almost magically, it results in better mutual understanding and better policies. And they are very transformative. So people who were in the street — half a million people — or online participating, it creates a peak experience that people then afterwards, whether it's on Uber, Airbnb, Bitcoin, whatever challenge of the day is, people instantly say, oh, let's go to this kind of digital democratic platform. So all it takes is just one peak experience, one success story. And then you can then ripple it out in all the high-school classrooms and many other places.

Well. In a sense, yes, because Taiwan, as you know, is the youngest tectonic island on earth — we're only four million years old. So it's literally plate against plate. Every day there's three earthquakes somewhere in Taiwan, small ones, but still earthquakes. But Yushan, the tip of Taiwan, was pushed up from — almost 4,000 meters now, I think reaching 4,000 meters, by half a centimeter every year — because of this plate against plate. So I think, yes, there is this existential opportunity in Taiwan that we have to innovate. Otherwise we get absorbed by the authoritarian regime that loves to say democracy never delivers, only leads to chaos. So we have to prove them wrong. But also, I think the fact that we celebrate diversity and see them as fuel — the conflict as fuel — plays a very large part in it. Because had we only tolerated diversity, then we would not be able to withstand the earthquake in 2014, like both literally and also metaphorically.

You have a very high-resolution camera, as we're having now over video conference, in the synchronous participation. So in Taiwan, broadband is a human right. No matter how rural you are, the remote islands, high up in Yushan, you're guaranteed to have the kind of bandwidth, and we have a universal service fund to ensure that telecom will invest in places where it doesn't make economic sense to do so. So this sounds very basic, but this really is the guarantee that no matter where you are in Taiwan, if you're mobility-challenged, if you need real-time transcription and translation in sign language, which is part of our national language, and so on, you have affordable, almost universal-healthcare-level access to all these communication services that close the distance.

I agree. I totally agree. And so I think — and this is an idea that I'm calling techno-communitarianism — which is to use technology to facilitate the existing in-person communities that we're all part of, like sports, fashion, food and faith also. In Taiwan, the more highly educated you are, the more likely you openly practice some spirituality. I am Taoist, which is 25% of the Taiwanese population. Buddhism, another quarter. Christianity, Islam, folk religion, animism, and you name it. And because there's no dominant religion, it has become a kind of favorite pastime without carrying darker connotations to talk about faith and religion. So once we use technology to foster not just intrafaith gatherings and convenings, especially for younger people to participate, but also interfaith communications, including with atheists, I think this then encourages people who would otherwise retreat into the wildfire of energy-consuming antisocial media into the in-person pro-social gatherings. So in a sense, the prosocial deliberation is almost just an excuse for you to get into this mode of conversation. And then next Sunday or Saturday or whatever, you can apply it to your own community gathering.

I would say that as long as you're not vendor-locked in, we should be free to experiment with pretty much any technological application in a group setting. Because even if those technologies may carry downsides, the group setting is the only setting, in a peer-to-peer way, where the young people will nonetheless be exposed to those harms, with other people, and panic together and figure out how to overcome it. There's just no other way. Right? So I think the group-setting experimentation with technology in a way that does not lock people in is very important, which is why in Taiwan we work on procurement rules to ensure that the data again is soil, not oil. So the personal data is not extracted outside of Taiwan, outside of their communities, but the community, including classrooms, own the data, and they run, for example, AirBoxes that measure air quality and water quality, noise level, and so on, and learn about data stewardship, learn about distributed ledgers, learn about collective sense-making and then decision-making. Maybe start some referendums together. Some schools actually did do that — water quality. And all this stems from the fact that the data for them is regenerative, is soil, not oil. So if we make this the procurement standard, if the teachers in the classroom learn to work with this kind of censorship-resistant, open, private and secure technology, then there's just no wrong road. You may be zigzagging a little bit, but that is good exposure for the young people. What I mostly worry is that there's some top-down edict saying only this solution is good. And then you sacrifice all your kind of communitarian sovereignty in the service of the solution of the day. And it turns out that it's not a good solution. That is the monoculture, the singularity, that I want people to be wary of.

Well, in a sense, I'm singing the praise of interoperability, which nowadays includes data and source code, but we're also in an age of agentic engineering. So I think source availability is one part of it, but the larger part of it is the care — the commitment to be continuously attentive and responsive and responsible to the various downstream applications that we as builders and technologists simply cannot anticipate. If we think that we're somehow above other people and extract their data as oil, then we will want to move fast and break things. But if we see that actually we're caretakers, and we're also caregivers through technology, then we think in data soil, in regenerative technologies, which include open source and open data, but more importantly, it carries a more caring maintenance mindset. So training the next generation of public-interest technologists in this ethics-of-care tradition, I think, is equally important as source availability.

Thank you, Toby. And scroll less, sleep more.

A growing number of organizations, from Silicon Valley tech outfits to a host of other big corporations, have begun linking AI use to performance evaluations. For many, survival in the workplace now involves investing substantial time and resources in learning how to collaborate with AI.

Yet, many may soon discover that these efforts do not necessarily make work smoother. Instead, they start to feel like obedience to a set of "prophecies."

Consider claims such as: "AI will become your closest work partner," "Everyone will have an AI assistant," or "Companies that fail to adopt AI will be eliminated." On face value, these statements seem to describe trends. In practice, they often discipline behavior: You better surrender your attention and judgment now, or you will be left behind.

My Oxford colleague Carissa Véliz, in her new book "Prophecy," reminds us that the power of prophecy lies not in accurately predicting the future, but in shaping it.

For example, suppose a business owner believes that "AI will replace 80 percent of the workforce." They may redesign performance systems, restructure teams and ultimately dismiss most of the employees. This does not prove that the prophecy was accurate. It proves that the prophecy is self-fulfilling.

Once the language of inevitability is accepted, people begin adjusting behavior accordingly, becoming part of the force that brings the prophecy into being.

Taiwan knows how to derail such narratives. In the 1980s, technology-intensive industries were widely seen as impossible to develop on our island of resilience. Today, we have TSMC. And at the beginning of the pandemic, some predicted that Taiwan would not survive the first wave. What later became clear is that many things described as "inevitable" were merely scripts that seemed plausible at the time.

When facing a prophecy, what matters most is always how we choose to respond.

If we want to free the future, the first and most important step is looking for collaborators. Whether the semiconductor industry or the pandemic response, it was collective action that overturned the prophecy.

So, is it inevitable that AI will replace human labor?

As infrastructure, AI may lower certain barriers to collaboration as it is used more frequently. But when the AI debate becomes a machine for manufacturing anxiety, the real questions we should ask are: What exactly do we want to use it for? For whom are we building it? And which values must we protect?

This is also why Silicon Valley tech outfits are inviting philosophers to take part in discussions on AI alignment, safety and institutional design. When people form emotional attachments to AI, or begin debating whether machines have consciousness, these are no longer merely engineering questions. They are questions about what kind of society we want to take shape.

If there was only one path into the future, none of these discussions would be necessary. It is precisely because multiple futures remain possible that prophecies deserve to be challenged.

So, whenever you hear that "a trend is irreversible," the more important question to ask is this: Is it truly a prediction, or is it a command?

Refusing inevitability does not mean rejecting technology. As long as we remain willing to deliberate, govern and build together, the future need not become the extension of a prophecy. It can become something we write and free — together.