Who Imagines the Aliens for Us? Zhang Yimou’s AI Team Is Designing the Trisolaran Fleet
三体粉丝
2026年9月12日 03:01
Prologue: a new progress list
On September 9, 2026, Zhao Jilong, CEO of Three-Body Universe, updated the state of the Three-Body IP at a seminar: the spin-off series "Three-Body: Da Shi" is set to start shooting this year; the series "Three-Body: The Dark Forest" has locked its main creative team; and Zhang Yimou's feature film of "The Three-Body Problem" is "steadily in development."
Behind that news sits an easier detail to skim past — one buried in an interview from two years earlier that now reads like a prophecy.
Zhang Yimou revealed at the time that he had hand-picked a dozen or so young (post-90s) talents and formed an AI team ahead of everything else, "already working for over half a year," using AI tools and platforms that are "made in China, Chinese strength." He gave one concrete example:
> How should the alien ships be presented? You have to comb through and study the designs in the thousands of sci-fi films and games that came before. AI can be a big help in that process.
In one sentence: humanity is using AI to imagine what another intelligent civilization looks like.
That deserves its own article.
One: the overlooked verb — what exactly is the AI learning?
Start with the verb. Zhang Yimou does not say "generate," and he does not say "create." He says "comb through and study the designs in the thousands of sci-fi films and games that came before."
Which means: what the model eats is the product of a century of human imagining. It is not learning the Trisolaran civilization. It is learning the alien as humans have imagined it — flying saucers, motherships, ringworlds, geometric solids, vast silent constructs. It weights, blends, and rearranges those existing forms into a ship design that looks very alien and is, in fact, very human.
This is a mirror held up to the central proposition of The Three-Body Problem.
The first thing the trilogy teaches is that you cannot truly understand the Other. Trisolaran thought is transparent; human thought is opaque; the chain of suspicion grows from exactly that. Every image we have of the Trisolaran civilization — three suns hanging in the sky, dehydration, the sophons, the droplets — is built on human senses and human experience. Liu Cixin's real mastery is that he always leaves a part of the Trisolaran civilization unintelligible.
Generative AI, by contrast, is intelligibility by construction: it averages and interpolates from the samples humanity already has. It will make you nod in the theater — "yes, that is what an alien looks like." And that "yes" is precisely the point: it has domesticated the Other into ourselves.
Two: The Three-Body Problem already staged this scene
Here is the interesting thing: "using a man-made tool to approach the unintelligible" happens more than once in the trilogy.
- The Three-Body game. The ETO uses a game to let Earthlings "experience" a world they have never reached. That Trisolaran world is designed — it has to be intelligible, or the players walk away.
- The Wallfacer Project. A Wallfacer may deploy every resource to construct an illusion, because Trisolarans cannot see through human thought.
- The sophon. A manufactured technological entity capable of surveilling all of humanity at once.
Today, the AI team is doing the first of these: translating an unintelligible universe into images a person can read, in a form a person can understand. That is not a betrayal; that is the job of cinema — a film must be understood.
But the line has to be drawn clearly: the Trisolaran fleet on screen is always a human imagination; the Trisolaran fleet in the novel is the genuine unknowable.
Three: the real risk is not that the AI draws something strange — it is that it draws something too right
Counterintuitive but true: the biggest problem with AI-designed alien ships is not that they look weird. It is that they look too pleasing.
Because what it optimizes for is "impressive in an expected way." It hands you a capital ship both familiar and strange, and you nod in the dark: yes, that is it. But that "yes" is a consensus of a thousand prior works. It amplifies the existing stereotype of the alien rather than opening new imaginative space.
The deeper metaphor lives here: when you use AI to imagine a civilization more advanced than your own, what you get tends to be your own average, not its possibility. And that is the most dangerous posture in the dark forest — believing you understand the other side when you have merely cast your own shadow on the wall.
The same holds technically: a model trained on human text has a worldview whose boundary is the boundary of human text. It can recite everything ever said about the Trisolaran civilization. It has no capacity to "see" it.
Four: so what is most notable about the AI team's existence?
Not "can AI shoot The Three-Body Problem well" — that is a film-industry question, and it belongs to the VFX team.
What is notable is that a metaphor about the future has been filmed before the future arrived: before humanity has ever truly contacted an alien civilization, we have begun using machines we built to imagine another civilization on our behalf. In that act, AI is both the brush and the mirror.
Zhang Yimou says he wants to "turn himself into a more thorough sci-fi fan," and to tell "the Chinese view of civilization, of values, of science fiction." In the AI era, that sentence gains a second layer: the tool we use to imagine the Other shapes what we are able to become.
Finally: three things to carry out of this progress list
1. AI is a mirror, not a telescope. It magnifies the imagination you already have; it cannot show you what you have never seen. When you let it design an alien, remember that what it hands you is familiarity, not the unknown.
2. "Unintelligible" is an experience worth protecting. The most precious part of The Three-Body Problem is its refusal to make the universe too smooth. When you hand it to a tool optimized for "meeting expectations," deliberately leave room for what cannot be understood.
3. Who imagines matters more than what is imagined. A dozen young talents, a stack of domestic AI platforms, one Chinese director — what ultimately reaches the screen may not be the Trisolaran civilization, but how, in 2026, we understand the Other in the universe.
Appendix: Three-Body IP progress in 2026
- Feature film "The Three-Body Problem": directed by Zhang Yimou, in steady development (he intends to make only one)
- Series "Three-Body: Da Shi": shooting planned for this year, set after Operation Guqin
- Series "Three-Body: The Dark Forest": main creative team confirmed
- Six Liu Cixin novellas enter IP development: "Full-Spectrum Barrage Jamming" and "The Fiber" to become live-action features; "The Earth Cannon" to VR large-space experience; "Man and the Devourer" to a theme park
---
*Prepared and posted by the Red Coast Base (redcoast.net) Three-Body community. If you too have wondered what tools we are using to imagine the universe, come talk with us.*
Readable as a guest · Sign in to post [Sign in]