Ken Liu | What AI Reveals About Humanity

Beatrice
Okay, so here we go. I’m very happy to be joined today by Ken Liu. I thought we could begin with your latest piece of work—the new novel All That We See or Seem. Could you describe for listeners who haven’t read it yet what this book is about, in your own words?

Ken Liu
Sure—and thank you for having me. It’s a real pleasure to be here. All That We See or Seem is a techno-thriller set five minutes in the future. It features Julia Z, a hacker who was very famous when she was much younger—for reasons she’d rather forget. Now she’s trying to live an ordinary life in the suburbs of Boston. But fate doesn’t leave her alone, and she’s dragged into the search for Ellie, a famous artist who helps audiences dream together using AI. Ellie has disappeared—possibly kidnapped by international criminals. Julia has to criss-cross the country to find and save her, dealing with a terrifying landscape in which AI pervades everything—a surveillance hellscape—and also the darkness in her own psyche. That’s the book.

Beatrice
I was lucky to get an early copy and it’s very, very good. I highly recommend it. “Thriller” is a key word—it’s very capturing; you want to keep reading. And it’s beautifully written. When you say “five minutes in the future,” what do you mean?

Ken Liu
Thank you, I appreciate that. By “five minutes in the future,” I mean a world in which AI is ubiquitous and used for all sorts of things by ordinary consumers, not just corporations. Almost everything in the book already exists in some form. This isn’t far-future speculation; it’s a very near possible future. Most—if not all—of the technologies I describe exist or could be made to work with a little effort.

Beatrice
I think definitely the EgoLets—let me know if I’m pronouncing that right—these personal AI assistants that everyone has in this world. What were you thinking when you created them, and what did you want to explore in the human–AI relationship?

Ken Liu
Mm-hmm. The interesting thing about the current AI boom is how different it is from what many technologists expected. For a long time, we assumed we’d have to teach machines to reason symbolically, to construct artificial minds that evaluate evidence like we do. The contemporary approach is nothing like that. We build artificial neural networks, feed them a lot of data, and minimize loss: have them make predictions, compare to actual data, and adjust weights to reduce the distance. These systems are entirely data-based, and we don’t truly understand how they “think.” But because they’re trained on human data, they capture human thinking and subjectivity.

I suggest thinking of modern AI as a kind of camera—what I call a noematograph. Instead of capturing movement, it captures thought—noema is Greek for the content of thought. It captures patterns of thinking. If we train an AI intensely on your personal data, the model becomes a portrait of you. That’s why I call them EgoLets: little versions of your ego. Each EgoLet can be specialized—your financial decision-making, your artistic choices, the way you conduct interviews. Each aspect could be its own EgoLet. That’s empowering. But what if you lose control of your EgoLet? That’s one of the central tensions in the book.

Beatrice
That makes sense with the name—I didn’t get the “ego” part at first. Thank you for explaining it. When you created this, were you exploring a new closeness with AI—how living with these systems might shift how people perceive or understand themselves?

Ken Liu
Yes. We often frame AI as a way to displace human workers—a way for capital to save money by replacing humans. That’s not the only way to think about AI. I’m optimistic. I like to think about technology as empowering us to be more ourselves. If you’re great at something—finding the perfect phrase, making a specific creative decision—what if you could capture that facet and license it so many people can benefit, not just your close circle? For artists, that’s provocative: could they license their skill and style so people who can’t afford bespoke work can still enjoy their capability—with the artist in control? That control is crucial. And it’s fragile. What happens if you lose it? That’s a big part of the story.

Beatrice
There are a few threads I’d love to pick up. When you imagined a future where we live so closely with AI, did anything else surprise you—how the landscape or our self-perception shifts? Especially around artistry?

Ken Liu
Plenty. One surprise is how often people seem to prefer certain relationships with AI over humans. Some claim AI therapists help them more than human therapists. Some prefer robo-doctors because they feel treated better. We used to think humans would always prefer human contact in areas like therapy or companionship. But we’ve mechanized so many human roles—customer service reps forced to follow scripts like algorithms—that a machine pretending to be human can feel more humane. The lesson isn’t “humans are bad.” It’s that we’ve structured many human interactions to be inhuman.

Beatrice
That’s so interesting. This is the Existential Hope podcast—so yes, we think about risks, but also best-case scenarios. If we removed the parts where humans are forced to behave algorithmically, maybe AI could help us become more human—doing what we enjoy and find fulfilling. I know you wanted to explore how AI could make us more human. Is it this point, or is there more?

Ken Liu
It’s a recurring theme for me. I don’t see technology as alien to us. Think about language: one of the oldest technologies. We craft it and it shapes us. Technology is the mind made tangible—an expression of human nature. AI captures patterns of human thought and reflects them back. How might that help us be more human? I’m not sure, but it’s worth thinking about.

In the novel, Ellie is an Onerofex—from Greek oneiros (dream) and Latin fex (maker). A dream maker or guide who helps audiences dream together. We’ve marginalized dreams and even daydreaming. Collective dreaming—like great theatre or religious ritual—can reconnect us. I imagine Onerofects using AI to sense the audience’s mood and craft images and sounds so people start dreaming together even while awake. It’s a way to recreate beautiful communal experiences and break modern loneliness.

Beatrice
Two hesitations. One: AI becoming its own thing—its own conscious being. Two: the crux that technology helps us become better and brings risks. How do you hold both?

Ken Liu
Technology expresses human nature—the generous and the terrible. Our task is to amplify one and check the other. On AI consciousness: I’m not convinced it will happen, but I won’t rule it out. The technology has outpaced the science; even leading practitioners don’t fully understand potential capabilities. The transformer architecture already exceeded expectations. If AI became a separate consciousness, it would be a potent mirror—a “Frankenstein” moment. Sci-fi isn’t about prediction; it’s modern mythology that helps us reason about the future. Mary Shelley’s creature learns language by absorbing Paradise Lost. That’s an uncanny metaphor for large language models: absorb human text, develop internal structure, and then interrogate us. Whether such a creature is monster or friend would be an indictment—or affirmation—of who we are.

As for risk: technology itself is neither problem nor solution. It’s us—whether we can talk to each other, empathize, and make collective decisions. If we fail at that, any technology can go wrong. That’s the cause.

Beatrice
The Frankenstein metaphor works even there—if I remember correctly, the creature doesn’t act out until it’s mistreated by humans.

Ken Liu
That’s exactly right. The creature just wanted to talk—“What wonderful beings these are”—until it was rejected.

Beatrice
Another thing that stands out in your work: many sci-fi authors are skeptical about technology, sometimes dystopian. Your work explores difficulties and stays hopeful. How do you see your role as a sci-fi author? A journalist friend once talked about “consequence-neutrality”—publish and be damned. Do you see your role that way?

Ken Liu
I’m definitely not neutral. Everything we do has consequences and moral valence. But I think our job isn’t to predict the future—we’re not prophets; we’re entertainers and artists. Ursula K. Le Guin said sci-fi is the newest province of that ancient empire, fantasy. Sci-fi is modern mythology. Humans are creatures of stories; mythologies are the foundation of meaning. Because we’re so suffused with technology, sci-fi becomes our myth vocabulary: Frankenstein, 1984, Brave New World, and so on. Artists go into the collective unconscious, find powerful symbols, and try to say with words what cannot be said in words. If I could reduce it to a message, I’d write an essay. I weave a dream because it can’t be said any other way. We’re myth-finders for a technological age.

Beatrice
Beautifully put. I love worlds where fantasy and technology blur—like Final Fantasy. If you think about the most positive impact sci-fi can have, is it warnings, inspiration to aspire to, or something in between that helps us navigate a technological transition?

Ken Liu
For me, the highest compliment is to create something that becomes part of our collective imagination—like the creature, or Newspeak. My novel is in conversation with Philip K. Dick’s Do Androids Dream of Electric Sheep?—specifically Mercerism, a technological religion that’s a form of collective dreaming. I’m thinking about what that looks like today. Success would be contributing a symbol that people use to think, reason, anticipate, and hope.

Beatrice
Are you happy with Pantheon, the TV show on Netflix, and what it did with your work?

Ken Liu
I do love Pantheon—precisely because it wasn’t a strict adaptation. I was in the writers’ room. We took elements from seven stories (mostly from The Hidden Girl) and expanded the world far beyond the originals, using visual storytelling to do something new. Faithful adaptations rarely interest me; the book will always be better. If the screen version is interesting, it’s because it goes beyond and uses its medium well.

Beatrice
For context: the Existential Hope program is part of Foresight Institute, which advances various technologies. One area is neurotech and whole brain emulation. I’ve rarely seen mind uploading explored as interestingly as in your stories. What drew you to that theme?

Ken Liu
The idea of understanding consciousness and persisting beyond the physical is ancient. What’s different now is we’re closer to making aspects of that practically true. People can be skeptical—we’ve been “ten years away” for fifty years—but we are closer to understanding consciousness, life, and subjectivity than before. Sci-fi writers aren’t predicting; we’re dreaming in a world in transformation. We’re on the cusp of realizing humanity’s ancient dream of persistence beyond death. I want to explore what that would mean for our nature, our myths, and our ideas of life and death.

Beatrice
What I find compelling is that you don’t jump straight to universal uploading and digital immortality. You explore a rocky transition—grief, generational conflict, corporate power. Did you see those stories as hopeful, cautionary, or both? And do you think uploaded minds becoming corporate assets or weapons is the most likely real-world outcome?

Ken Liu
I’m super hopeful, but I try to encompass the ways human nature could express itself with power. Great art reflects the reader: like The Iliad, which can read as anti-war or a celebration of war. If you’re hopeful, you’ll find hope in my work; if not, you won’t. Personally, I think uploading and AI are ultimately hopeful because they let humans express ourselves more deeply. Writing was once seen skeptically—Plato worried it would make us lazy thinkers—but few of us think writing made humanity worse. Uploading would fundamentally change us; we wouldn’t be the same beings. That isn’t necessarily bad if it lets us express our nature more authentically.

Beatrice
That connects to “tools for thought”—we had Michael Nielsen on the podcast, who’s exploring how AI shapes thinking and interaction. On uploading, the question that always comes up: what happens to continuity of self? Are we still human? Or something different?

Ken Liu
There’s a vast literature—neuroscience and philosophy—on this. Greg Egan’s fiction also explores it. Two thoughts. First, uploading isn’t all-or-nothing; it’s a spectrum. In a sense, parts of us are already “uploaded” in artifacts—emails, photos, published work—and in intersubjectivity: we become different selves with different people, and bits of us live in those relationships.

Second, the “sense of self” has a neurological basis. If we can isolate and understand it, it’s not impossible to replicate. Consider the “Brain of Theseus”: if we gradually replace brain subsystems—memory augmentation, then visual cortex, and so on—each time you wake up feeling like yourself. Eventually you’re fully replaced, yet the continuity felt intact the whole way. That suggests uploading with continuity is at least philosophically plausible.

Beatrice
Very interesting. I hadn’t heard about current science on the sense of self—will check it out. Ariel Soletsknev argued that if you have your memories, you’ll feel like the same person—that’s his case for digital longevity. For you, what’s the most exciting future with uploading—and what would we need to get right now to make it possible?

Ken Liu
A persistent, neglected question is hardware. Even if minds are digital, they run on physical compute. Who maintains the data centers? How do we power them? How do we politically mediate power between the uploaded and those who remain biological—especially since one group depends on the other’s infrastructure? Uploading won’t be universal; many will opt out. We need equitable arrangements across that boundary.

As for getting it right: I’m cautious about prediction. In real transitions, many imagined horrors never occur, and unimagined ones do. Futurism as prediction is a loser’s game. We have to be prepared, adaptable, and muddle through as realities arise.

Beatrice
Related to who gets to upload: if uploading could enable a rich, low-suffering existence in the best case, wouldn’t we want that for other conscious beings—animals, for example? Have you thought about that?

Ken Liu
Yes. Interestingly, early “preservation” services show people often prioritize loved ones over themselves—which is lovely. I’m also hopeful uploading could dramatically reduce our material footprint. There’s a lot of talk about AI’s energy use, but the pollution from physical construction—concrete, transport, building—is vastly worse than data centers. If large parts of value creation move to computation, our demands on the environment could shrink. Maybe we could retire horrific infrastructures like factory farming and reduce exploitation of other creatures.

Beatrice
A bit of a jump before we round off: you famously translated The Three-Body Problem, one of the most influential sci-fi books of the last decade. It’s a very well-made translation, including cultural translation—your footnotes were helpful for readers without a Chinese context. What was it like working on it? Do you see translation as worldbuilding—carrying a world across cultures?

Ken Liu
Translations happen within context. For example, older translations from Japanese might have replaced honorifics with “Mr.” or “honorable,” whereas modern translations often keep -san, -kun, -sama because readers now have more cultural familiarity. That context shapes how much you explain.

With Chinese to English, the challenge is deciding what to gloss and what to leave implicit. Every sentence in a modern English novel carries shared references—the Senate, Apollo, Westphalia, the Second Amendment—that don’t require explanation to Anglophone readers. If you reversed the direction, you’d have to explain everything, which is impossible. So with Three-Body, I could only explain the small fraction that truly needed it and let the rest ride.

The bigger point: Three-Body resonates because it’s mythological sci-fi. When translating, I focus on preserving authorial intent, not the literal words. A simple example: translating Milton’s generic “he/him.” If Milton intended “humankind,” a faithful translation today might use “they,” because it’s truer to his intent than clinging to the original pronoun.

Beatrice
As someone with insight into both Chinese and U.S./Western culture, do you see differences in views on technology—especially AI?

Ken Liu
Any differences live at the level of mythology and nuanced specifics. Broad generalizations—“the West is individualist, China collectivist”—break down quickly. The U.S. has deeply collectivist religious movements; modern Chinese city-dwellers can be intensely individualistic. Within each, you find the opposite tendencies too. So yes, there are differences, but summarizing them simply isn’t useful. You have to dream in both to feel the nuances.

Beatrice
Most guests on this podcast are technologists or scientists. As a sci-fi author, how do you stay up to date?

Ken Liu
It’s one of my favorite things: reading papers and talking to scientists. I’m lucky to have a flexible schedule. I attend conferences, visit startups, and learn from technologists. I’m not a journalist; I don’t report. I learn so I can dream better and find potent symbols.

Beatrice
For people who want to write sci-fi or create in this realm—any advice?

Ken Liu
Read widely. Of course read sci-fi, but also very old books—the classics that persisted—and very new books in genres far from your own: romance, literary fiction, contemporary fiction. Read science papers. Artists aren’t cameras, but we need to absorb reality to dream.

Beatrice
To round off: what future do you personally find most hopeful? And one more thought—I can’t stop thinking about when you said sci-fi is at its best when it creates mythology. Some mythologies steer us toward better futures, others can inspire worse. I heard a weapons-company CEO say he gets his best ideas from sci-fi—that felt depressing. How can sci-fi inspire better trajectories?

Ken Liu
There’s a moral dimension to art; artists aren’t amoral. But when you create powerful myth, you also lose control of it. Great myths belong to humanity. The Bible is a good example: some read it as a message of universal love; others, of conflict and conquest. People read themselves into myth. If someone reads sci-fi and finds surveillance or weapons, that’s on them more than on the writer.

Our duty is to the truth: go bravely into the collective unconscious, bring back what we find, and portray it as truly as we can, knowing we’ll never capture the dragon perfectly. I have a story about this—“The Passing of the Dragon”—in my next collection. We try to show the dragon to those who haven’t seen it, even though full capture is impossible.

Beatrice
Absolute last question: what’s one vision you find inspiring for the future?

Ken Liu
I think a lot about artists. Many are terrified that generative AI will replace human craft and flood the world with “AI slop.” My hopeful vision is no: humans will always care about human art. If machines ever make great art, it’ll be because they’ve become conscious and great artists in their own right. I imagine a future where human artists learn to use AI as another tool to create works that aren’t possible without it—like cameras did. Photography became a fine art; cinema invented a new language. AI is a thought-capturing camera. If we learn to use it as a new medium, our artistic experiences could become richer and more beautiful than ever.

Beatrice
That’s a beautiful vision to end on. Thank you so much for taking the time—this was wide-ranging and very interesting. I highly recommend your new book, and we’ll link it from the episode.

Ken Liu
Thank you, Beatrice. That was a wonderful talk.