Thanks for taking the time to do this conversation. Is it okay if I record this just for my own reference when writing the story?

Yes, definitely. I’ll keep a transcript as well. Actually, I have a small request: after you publish the story, would you mind if I also publish the transcript on my website, assuming you’re comfortable?

You mean the full transcript? I’ve never actually done that before. Let me just check with my editor. That’s something I’d need to confirm.

Of course, that’s totally fine. You can edit out or rephrase anything you’d like. I ask because I have a website where, for more than a decade, I’ve been publishing transcripts of all my on-the-record conversations—with journalists, lobbyists, everyone. If you agree, you would become the 8,000th speaker in over 2,000 published conversations.

That’s like when you go to a store and you’re the 10,000th customer! I’ve always wanted to be that special number. I could be the 8,000th speaker.

Exactly.

I’m interested. Let me just start recording on my end… Recording in progress. Okay, cool. Well, for context, this is a story for JoySauce, a publication focused on Asian and Asian American stories. This particular piece will be a Q&A. I’ve been corresponding with someone from Seattle Arts & Lectures through your team. I work in tech myself, and have done AI-related work, though I’m speaking just on my own behalf and not for any company.

I’m really curious about generative AI and how quickly average people can use it—like you can use AI now to write an email or even an entire book. This changes what it means to be a writer, a storyteller, a creative person. So that’s my lens: how we use AI today in everyday life, in ways that didn’t exist before.

Right. It’s like having new “brushes.” AI can imitate your style, or anybody’s style, quite readily.

Yes, exactly. Where do you sit with the current state of AI? I’m in the U.S., and it seems especially heightened here—maybe because of the recent election cycle. But how do you see AI and its social ramifications right now?

What do you mean by “heightened”? Do you mean polarization? Tribalism? Something else?

I think about how technology companies, at least in the U.S., are investing less in fact-checking, for instance. Facebook has rolled back parts of their fact-checking programs. There’s also reduced support for DEI (Diversity, Equity, Inclusion). Many friends who do DEI speaking or training have had their funding pulled. I see this as linked to broader political shifts.

People often rely on journalism or on social media to “speak truth to power,” but when the companies controlling these platforms deprioritize fact-checking or facilitating truth, it becomes harder to find reliable information. Meanwhile, AI-generated content is getting more convincing—making it increasingly difficult to tell what’s real or who created what. That can create more opportunities for disinformation or scams.

I see. So you’re referring not just to creative or playful synthetic content, but also content used for scamming, political manipulation, or other malicious purposes. Indeed, it’s quite flexible as a “brush,” and that flexibility can cause trouble. It becomes difficult to distinguish photography from AI-generated artwork, for instance, and deepfake videos can simulate real people.

In Taiwan, we’ve seen that issue explode—particularly with social media ads. You might see a celebrity in a Facebook ad promoting crypto or something suspicious. If you click on it, “they” appear in video and somehow have time to speak to you in real-time. Previously, you could tell it was fake because of, say, six fingers or weird image artifacts, but this year those telltale errors have basically disappeared. Now it’s very convincing.

Until recently, a scammer had to puppeteer these models. But nowadays, with more advanced AI, it can run automatically, end-to-end, scamming people with no direct human in the loop. That’s a serious concern.

In Taiwan, we use AI plus broad social participation to tackle that. For example, we randomly texted 200,000 people asking them to join a massive online deliberation—like a giant citizens’ jury—and 450 people joined across 45 AI-facilitated breakout rooms. They collectively figured out how to counteract this kind of fraud.

The result is that Taiwan became the only country (as far as I know) requiring all Meta or YouTube ads to have a certified digital signature. If a celebrity supposedly endorsed something but didn’t actually sign off, and someone is scammed for a million dollars, then Facebook or YouTube is liable for that loss if they failed to verify a celebrity’s real digital signature.

It’s basically crowdsourced regulation. If a platform like Facebook doesn’t comply, our National Communications Commission can slow down connections to Facebook or block domain resolution—similar to how TikTok was threatened in the U.S. That approach has drastically reduced these scams.

You can see how it’s not just “AI vs. democracy,” but “democracy working with AI.”

That’s really interesting. So you basically tapped into the wisdom of the crowd, using deliberative processes, to find a technology-based solution. Do you think that approach scales to larger countries with different political cultures? Or is it unique to Taiwan, which already has a culture of civic participation?

The U.S. has also tried some of this on a smaller scale. The open-source platform we used, called Pol.is, was tested in Bowling Green, Kentucky, and continues to be used there for certain local public-service deliberations.

Various U.S. states, such as Oregon, already use citizen assemblies and civil juries to deliberate on policy. California and other states are moving in that direction, too. If you search “citizen assembly” or “democracy next” or “broad listening,” you’ll find many such experiments.

At the national scale, we can glimpse hints of it. For instance, Meta’s “Community Notes” (formerly “Birdwatch” on Twitter, or “fact-checking with cross-party upvotes”) attempts to reduce tribal polarization. Community Notes only get traction on a trending post if it’s upvoted by both left-leaning and right-leaning people—people who usually disagree. This forces “both wings” to cooperate to bring the note up, reversing the typical race to the bottom of the brainstem (the outrage-driven approach).

If done well, it surfaces context or bridging facts. That approach can inform people better, preventing further polarization. It’s not perfect, but we’re actively experimenting with more open approaches—like in BlueSky—for what I call “pro-social media.”

Yes, in past interviews, you’ve sounded quite hopeful about community-based solutions to misinformation. Have you always been optimistic about technology-for-good, or about building community through digital innovation?

Yes. Because Taiwan was in a place much like the U.S. is in now. Back in 2014, trust in government was around 9%, and society was extremely fragmented. We had a president who, it seemed, every time he spoke, the majority was against him—no matter what he said. Eventually, we occupied Parliament. Nonviolently, but it was still an occupation for three weeks.

Yet that occupation was facilitated. We found a way to flip technology’s logic. If technology is used to maximize time-on-screen, it can degrade attention spans and social bonds. But if we use technology for “broad listening,” it becomes a tool for discovering common ground.

In 2014, during the Sunflower Movement, more than 20 NGOs each facilitated some aspect of online/offline conversation, all recorded and transcribed daily. We identified and publicized consensus points that emerged over time. As the protest continued, the consensus got stronger—until the speaker of Parliament said, “Okay, the crowd has the better idea. I’ll adopt that.”

That transformed me from cynic to optimist. I realized we could shape digital technology in a way that fosters democracy rather than fragmenting it.

You were already working in tech then, right?

Yes, I had been working with Apple’s AI team on multilingual capabilities for Siri, and also at Socialtext, which built collaborative enterprise software (wikis, shared spreadsheets, collaborative editing) before Google Docs or Sheets existed.

In large enterprises, you do not want employees staring at their phones indefinitely—that’s not helpful. The goal is calmer tech that fosters real collaboration. The “assistant” concept can free you for deeper work, rather than dragging you into bottomless scrolling.

LinkedIn is an example: it has a “feed,” but the ranking is quite different from Facebook or Instagram. On LinkedIn, you don’t feel that addictive, infinite scroll in the same way.

You’ll be speaking in Seattle soon, hosted by Seattle Arts & Lectures. Are you planning to talk about the state of AI, democracy, and so on?

I’ll be in conversation with Ted Chiang, whom I greatly admire. He’s brilliant, with deep insight into how machines affect human understanding—like his analogy of large language models as “blurry JPEGs of the web.” I’ll likely follow his lead and the audience’s questions.

But my main point is we don’t need to align humanity to AI. We can align AI to humanity. If we don’t, we risk losing meaning—losing what makes human connection, creativity, and communities valuable. The question is how to ensure AI truly serves our social fabric, not the other way around.

What would it look like to have an AI that aligns with humanity? Sometimes it’s described as an “assistive model” rather than a “replacement.” People ask me, “Aren’t you afraid AI will replace writers?” I say no—writing is more than stringing words together. AI might help me research or rephrase text, but I still do the real creative thinking.

Exactly. This interview is a great example. Our conversation is fully human-to-human. Yet behind the scenes, AI is already doing things like noise cancellation or background replacement so I don’t need to wear headphones or tidy my room. That’s a helpful assistive layer letting us focus on each other.

That’s different from sending an AI avatar to speak on my behalf. That would remove the authenticity of real human connection.

A simple heuristic: if it’s machine-to-machine, we can gradually automate it. If it’s truly human-to-human, the machine should only assist, not replace. If there’s a role in between—like a software engineer writing code (the “machine” side) and designing user experiences (the “human” side)—AI can automate the purely mechanical tasks, freeing the designer’s time for deeper creativity.

What about more “gray area” cases, like auto-generated YouTube channels summarizing content? People are using AI to generate quick video summaries of books or movies. In some sense that’s helpful—maybe I just want a spoiler of a scary movie. But it also removes a bit of the human commentary or personal perspective.

Yes, it’s a great example. Suppose I want a short summary or a quick translation of something. In that sense, the AI is purely translational, with no personal agenda. It’s a convenience. People already used to hire translators transactionally, with minimal personal connection. So letting AI do it doesn’t necessarily break a human-to-human bond.

But if the point is to subscribe to a particular human’s perspective—the intangible nuance of their life story—then you don’t want that replaced by AI. An AI has no genuine life story, no embodied vulnerability or experiences. So it depends on whether we seek a purely transactional “please summarize X” or a relational “tell me your viewpoint.”

As AI-generated content gets harder to distinguish from human-made content, do you worry about trust or about verifying authenticity?

Yes, which is why Taiwan’s approach to digital signatures is important. We made digital-signature infrastructure a public good. You can bring your phone to a household registration office, prove your identity, and your phone is then equipped to sign content cryptographically. You can also selectively disclose. For instance, prove you’re over a certain age without revealing your exact birthdate.

Then, if someone posts content claiming to be me, people can see if it’s digitally signed. If not, or if it’s forged, the platform has liability. Likewise for ads featuring celebrities.

We need two safeguards: (1) the state must not surveil or track each signature use, so you’re not forced to reveal every conversation. (2) interoperability, so we can verify across borders. Many places—U.S. states, Japan, Australia, Singapore—are testing systems for easy cross-border verification that you’re human.

From my time in big tech, everything often feels proprietary rather than open. Does that pose a challenge to building these signature or identity-verification systems at scale?

Historically, cryptography was treated as a military secret in the U.S. That was eventually relaxed as people saw you could print the algorithm on a T-shirt—so was that arms export or free speech? Over time, open cryptographic standards became the norm.

That’s why even proprietary software tools rely on open-source cryptography underneath. No one trusts a “black box” cryptosystem these days; a system with a hidden backdoor is too risky. We’re at a similar moment with large-scale infrastructures like Ethereum or other blockchains. They typically must be open source for people to trust them.

Similarly, “OpenAI” was open initially but then got more proprietary. Sam Altman himself recently suggested maybe they’re on the wrong side of history and that open source is the future.

So the “open source” approach can let people verify. Is that push coming more from users or from the industry?

It’s both. Communities demand transparency, and open-source innovators get “free cycles”: outside developers volunteer improvements, donate hardware for training, etc. Many major breakthroughs—browsers, operating systems—emerged from open source.

From everything you’ve done—Apple, Socialtext, now in the public sector—it sounds like community and togetherness have always shaped your work. Is that a fair statement?

Yes. In Taiwan, democracy and the internet basically arrived at the same time. We ended martial law in the 1980s and had our first direct presidential election in 1996—the same year the World Wide Web exploded in popularity. We see them as intertwined: new ways for society to co-create its destiny.

I want to believe technology can help democracy in the U.S., too, but it feels like democracy is eroding in recent years, and that big tech is fueling polarization. I still trust in communities, but I’m also a bit anxious about how low we’ll sink before we rebuild.

In Taiwan, you’ll notice that many leaders with advanced education also maintain spiritual practices. We have no single dominant religion—my father studied Tibetan Buddhism, my grandparents were Catholic, and I practice Taoism. There’s also folk religion, Shinto-like beliefs, etc.

The result is that technology did not displace or overshadow those spiritual/civic roots. It served to connect them. Our public officials sometimes talk openly about their faith—like our former Premier Chen Chien-jen, who is Catholic and even recognized by the Vatican. He was once asked in a parliamentary interpellation, “Does God talk to you?” and he replied, “Yes, if I pray sincerely.”

We have no single majority religion, so he wasn’t imposing on others; it was just a sign of vulnerability and personal connection. That’s part of diversity. In many U.S. tech contexts, you see cultural or ethnic diversity but not always spiritual diversity. Reconnecting to that might help unify people across divides rather than locking them into tribal conflict.

It’s interesting because here, especially in tech, “spirituality” can sound antithetical to “logic,” or at least some people see them as opposites. Meanwhile, you’re describing how your Taoist perspective literally informs your approach to innovation.

Yes, exactly. Taoism, as I practice it, is about balancing and seeing how yin and yang contain bits of each other. Even my approach to “seeking uncommon ground” is partly derived from the Tai Chi symbol—seeing the white dot within the black half, and vice versa. It’s a design principle, not just a concept.

Could you say more about how Taoism became central in your life?

Sure. I was born with a heart defect. One of my earliest memories is a doctor telling my parents that I had only a 50% chance of surviving until I could get heart surgery (which I eventually got at age 12). Those first 12 years, I never knew if I’d wake up the next day. It felt like flipping a coin nightly.

So I learned Taoist breathing and qigong meditation from my parents, basically as a survival technique: if my heart rate got too high, I’d faint and potentially end up in the hospital. Those exercises let me restore homeostasis. It became second nature before I even read the Tao Te Ching.

How does that look for you today? Sometimes mindfulness can seem at odds with the pressure to be super-productive.

Yes, “letting go.” My solution is never to postpone things to the weekend—because who knows if a weekend will come? I publish before I perish, so to speak. That’s why I record these interviews and ask to publish them. I started doing diaries on tape as a child. If I don’t wake up tomorrow, at least I’ve left something behind—my ideas are in the public domain so others can carry them forward.

The idea of open source resonates with that: I don’t have to wait to make a “perfect version.” I can publish a draft, go to sleep, and if I don’t wake up, it’s already available. Similarly, when I co-wrote a book called Plurality with over 60 contributors, we put it in the public domain so people can make manga adaptations, translations, documentaries—whatever—no permission needed. They don’t have to wait 75 years for copyright to expire. That’s the notion of being a “good enough ancestor.”

That sounds quite liberating. I often tell people who want to write that they should just write and share it. But they’re afraid to publish. Do you find it freeing to just let your works exist in the world?

Absolutely. Ever since Creative Commons came around, I’ve licensed my written works under it. It’s analogous to open source, but for creative works like text. Wikipedia uses a form of Creative Commons called “ShareAlike,” which is effectively “copyleft”—if you translate or remix it, you must also share your result under the same license.

“Copyleft”? That’s a fun term.

Yes—imagine the © symbol mirrored so the “C” faces left. It’s an established concept in free culture circles.

We’ve talked about Taoism. Are there other aspects of your identity that inform your work?

Yes. I’ve experienced two puberties. I was born with a testosterone level that tested like that of a 70- or 80-year-old man. So I never fully developed as a boy in my first puberty, and I never fully developed with estrogen in my second puberty either. I’m physically non-binary, so from my perspective, I don’t see half the population as “like me” and the other half as “unlike me.” Everyone is equidistant.

That also extends to how I refuse to list gender on official forms—just “N/A”—and similarly no party affiliation. It’s somewhat Taoist, but it’s also a systemic, non-binary approach to bridging differences. Instead of seeing an issue as “my side must extinguish the other,” I see it more like two eyes working together to produce stereoscopic vision. Different viewpoints can yield a deeper understanding.

It reminds me of “queering” an institution, like going beyond the default assumptions. But sometimes people equate “non-binary thinking” with “I don’t see color” or ignoring differences. How do you respond?

Saying “I see no color” flattens everything to sameness, removing the dynamic that differences bring. I do see color—all the colors! The key is not letting difference become an excuse for conflict. If we say, “Oh, both sides are equally correct, so we do nothing,” that’s not helpful either. We need synergy, like wings of a bird; they must coordinate, or the bird can’t fly. If left and right eyes don’t work together, you lose depth perception. That’s quite different from “both-sides-ism,” which can lead to inaction. Instead, differences create energy—our job is to harness it productively rather than let it explode destructively.

Any advice for organizations that want to celebrate differences in ways that actually benefit the community and not just check a diversity box?

Tap into existing traditions. Don’t invent a new “idol,” whether it’s AI or anything else, out of nowhere. In Taiwan, many people have some spiritual or meditative practice—could be Buddhism, Taoism, Catholicism, yoga, Zazen. Even if you don’t call it a religion, it’s a form of spiritual practice. Those contain deep wisdom about community-building, mindfulness, and bridging differences.

It reminds me of how “networking” can feel transactional. I tell people to think of it instead as relationship-building—bring curiosity, genuinely learn about others. That’s more fulfilling.

Yes, exactly. Another practice is not to jump immediately to judgment. In Chinese, we have a term “傾聽” (qīng tīng)—to tilt yourself fully in the direction of the speaker so you can see the world from their perspective. You suspend judgment for a time, then evaluate after hearing them out. If you judge too soon, you won’t fully absorb their perspective. It’s akin to reading a book: if you’re criticizing every other paragraph, you won’t immerse in the writer’s worldview.

I learned that in journalism school: open-ended questions and letting people fill in the gaps. But sometimes you do have to steer a conversation toward certain themes.

Right, my parents were both journalists, and they taught me that journalism is a journey of discovery, not a final pronouncement. There’s always more data, more angles, more voices to include. A good journalist has to bridge differences so that people with opposing views can see what they still share. Opinion pieces are fine, but the overall ecosystem should reveal the larger picture.

We’re trying to embed that bridging approach into “pro-social media.” Without it, platforms devolve into echo chambers reinforcing different rabbit holes, fracturing society.

Could you talk more about “pro-social media?” Because major platforms seem to prioritize ad revenue, which can amplify extreme content.

Yes. If a platform is advertiser-driven, it lumps people into micro-cohorts for targeting. That is invisible to ordinary users. As a result, people may see a viral post and think it’s “consensus,” when it might just be popular in a specific silo. Conversely, truly unifying content might remain hidden if the algorithm doesn’t highlight shared resonance. This leads to a false-consensus effect (where we overestimate how widely an extreme view is shared) plus a hidden-agreement effect (where people don’t realize how many others share a moderate stance).

Pro-social media tries to correct that. Community-driven curation, like Community Notes, or specialized platforms such as the MIT-based “Cortico,” can surface surprising common ground. In Taiwan, we use Pol.is for textual sensemaking; Cortico can do it with audio, creating “medleys” that let people hear clips from diverse participants. When you realize how others feel—and that you share more than you thought—it’s powerful.

That’s fascinating. Does it give you hope that we can move to these new social media paradigms?

Yes. Ten years ago, many people believed more connectivity = good by default. Then we saw how it could supercharge polarization. Now there’s near-universal agreement that we want more bridging, not more division. That broad consensus gives me hope.

Besides the power of communities to curb disinformation, how else can communities advance democracy?

“Democracy” with a small “d” can be done locally—no need to start federally. Some neighborhoods in San Francisco effectively practice direct democracy in shared-living arrangements, co-ops, or “intentional communities,” not unlike occupying spaces in Taiwan. Everyone shares child-rearing or decision-making; maybe they take a quick straw poll on WhatsApp. Below about 150 people (Dunbar’s number), direct democracy is very doable.

That fosters civic muscle. You don’t need elaborate digital verification if you already recognize each other face-to-face. So if folks start at that community level, they learn the habits of self-governance. Then they can scale that experience to bigger contexts.

In Seattle during the 2020 protests, there was the “Capitol Hill Autonomous Zone,” an experiment in local self-governance that lasted a bit before the city came down on it.

Yes, it reminds me of the Sunflower Movement’s occupation of parliament. The difference is that in our case we had an explicit facilitation process, so it felt less chaotic. But overall, it’s that idea of “small-scale direct democracy in action.”

I think those are the main questions I had. Is there anything else you want to add or emphasize?

We’ve covered a lot. I appreciate your open, curious approach. Perhaps I’ll just point out one additional angle: a more spiritual exploration of these same themes.

When I was in Davos, the Buddhist monk Shoukei Matsumoto, who translated the book The Good Ancestor into Japanese, and I had a discussion about Taoism, funerals, and “the passing of human supremacy.” We discussed how AI humbles humanity’s sense of supremacy (especially in certain tasks) and can reconnect us to a broader, more-than-human world. That prompts us to ask: “If AI is bridging human and nonhuman, like analyzing whale communication or birdsong, what ethics are involved?” Are we ready to communicate with those other intelligences?

It’s another big frontier AI opens up.

Has AI made you rethink your own connections or your own work?

Absolutely. I’m involved in a project called globaldialogues.ai that crowdsources questions from around the world, then asks them back to a global community. Some people want to pose the question: “If AI can translate between human and animal languages—like whales or certain birds—what should we ethically ask them?” That’s mind-blowing because a couple of years ago it seemed sci-fi, but now we’re inching closer to possible translations.

Listening is one thing, but responding raises ethical considerations. This is where the canvas opens up—AI as translator to the more-than-human realm.

That can feel overwhelming—like infinite possibilities. Do you ever feel that overwhelm?

If I get eight hours of sleep, no. (Laughs.) I operate with what I call a daily “checkpoint.” I wake up each morning with one day’s worth of ego, and it dies when I go to sleep. Yesterday’s Audrey is already my ancestor, tomorrow’s Audrey is my reincarnation. If I just do the best I can today and publish those efforts, then if I don’t wake up, the open-source world can continue.

Do you feel you’re a “good ancestor,” then?

A good enough ancestor. Not perfect. If we try to design a flawless future for everyone, it’s actually hubris. It robs future generations of their own creative role. “Good enough” leaves some cracks for light to get in, as Leonard Cohen would say.

When you look at your career so far, is there a single accomplishment you’re proudest of?

Convincing about 8,000 people to share their stories in the public domain, at least for the span of one conversation, so future generations can use it however they wish. That’s a large community of people who, for that moment, relinquished any claim to controlling how it’s used. That “good enough ancestorship” is probably my favorite outcome—shifting from “Control” to the “Space” bar on the keyboard, so to speak, creating more shared space.

That’s a lovely way to put it. Last random personal question: since you’ve done thousands of these interviews, do you have a “favorite” one, or are there qualities that make an interview especially good from your perspective?

I appreciate interviews that include random, personal questions—like the one you just asked—because it shows we’re actually connecting, not just following a script. If an interviewer only wants a set of standard statements I’ve already given 1,000 times, they could just query an AI that’s memorized all my past interviews.

I love trying to ask questions people haven’t heard before. Have you ever been genuinely stumped or surprised?

Yes—this conversation about “queering” social contexts is something I rarely get asked directly. In 2018, I had a somewhat related dialogue with philosopher Paul B. Preciado about “queering museums,” but you brought up a broader notion of queering everyday relationships or default assumptions. That made me pause and think. I appreciated that!

I’ve thought a lot about it as a queer person, especially around “straight passing”—like people assume heteronormativity by default. Queerness grants you freedom to reject that baseline assumption. But it can also be exhausting to re-explain yourself. So I’m always exploring that tension.

Yes, it’s powerful but can be a repeated challenge. It helps us see each other by our shared values, not by some “prototypical human” yardstick.

Thank you. This has been so interesting and enlightening—and not a traditional, formulaic interview.

Thank you. I really enjoyed this. Thanks for labeling potential leading questions and giving me space to answer non-binarily.

Of course. Last detail—do you typically want to review the story draft before it goes live?

I generally only review the transcript for accuracy (typos, clarifications). Then once your editor finalizes the published piece, I share the transcript—assuming we’ve agreed on that plan.

Perfect. Thank you!

Thank you. Take care. Bye!