Fictional Computers: The Three Body Problem

If you intend to see the Netflix series “The Three Body Problem” or you want to read the Hugo-winning story from Chinese author [Cixin Liu], then you should probably bookmark this post and stop reading immediately. There will be some mild spoilers. You have been warned.

While the show does have some moments that will make your science brain cringe, there is one scene that shows a computer that could actually be built. Would it be practical? Probably not in real life, but in the context provided by the show, it was perfectly feasible. It could have, however, been done a little better, but the idea was — like many great ideas — both deceptively simple and amazingly profound. The computer was made of human beings. I’m not talking like Dune’s mentats — humans with super brains augmented by drugs or technology. This is something very different.


This is your last chance. There are spoilers ahead, although I’ll try to leave out as much as I can. In the story, top scientists receive a mysterious headset that allows them to experience totally immersive holodeck-style virtual reality. When they put the headset on, they are in what appears to be a game. The game puts you in a historical location — the court of Henry VIII or Ghengis Kahn. However, this Earth has three suns. The planet is sometimes in a nicely habitable zone and sometimes is not. The periods when the planet is uninhabitable might have everything bursting into flames or freezing, or there might not be sufficient gravity to hold them on the planet’s surface. (Although I’ll admit, I found that one hard to grasp.)

Apparently, the inhabitants of this quasi-Earth can hibernate through the “chaotic eras” and wait for the next “stable era” that lasts a long time. The problem, as you probably know, is that there is no general closed-form solution for the three-body problem. Of course, there are approximations and special cases, but it isn’t easy to make long-term predictions about the state of three bodies, even with modern computers.

The Great Kahn

Of course, the court of Ghengis Kahn didn’t have computers, but they could have, as the show demonstrated. In one scene, some players of the game are passing by row after row of soldiers.

At first, you think it is just a show of military force for some reason. However, two other players announce that they can solve the riddle of when the next chaotic era will arrive and how long it will last. You can see that the field of soldiers is vast, and there are towers strategically placed.

The towers probably communicate between different sections.

On a command to “run” a program, the soldiers begin twisting poles they are holding with large cards on them. One side of the card is white, and the other black.

It doesn’t take much to realize that one color is a binary 1, and the other is a binary 0. The soldiers each follow rules to form logic gates. You can assume the towers gather up results and send them to other towers. Riders on horseback carry the results up to the player who is with the Kahn.


This could obviously be made to work. Each soldier’s rules would determine what kind of logic gate it was. For example: look at the soldier to your right and your left. If both of their cards are black, show your black card. Otherwise, show your white card. There’s an OR gate. An AND gate or an inverter would be just as easy.

You can’t tell me this doesn’t look like a CPU die under low magnification

Of course, the entire setup looks like an IC die, so in real life, you’d probably have problems getting so many people together and trained. However, this is virtual reality, so it seems about as feasible as making a CPU in, say, Minecraft. It would be easy to write rules for higher-level functions, too. For example, different flip flops, multiplexers, and demultiplexers could be rule-based like in Verilog, instead of made from gates.

There are a few other practical considerations. It isn’t totally clear on the TV show, but it seems like the system is asynchronous. That’s hard to design, so we usually use synchronous designs. I mentioned flip flop rules earlier. It seems like some instructions to the human elements would start with the phrase “When you hear the drum beat…” The drummer, then, would be like a clock.

Of course, on a large FPGA, you have to worry about getting the clock signal to everyone simultaneously and you’d have the same problem here. Perhaps you’d use pipelining to have a local clock that then hands off to a different clock domain. Maybe a mirror on a tower could help synchronize the clocks.

The other problem is that humans make mistakes far more often than digital logic. So, like a relay or quantum computer, it would probably help to detect errors and maybe even make corrections in the system. Of course, we don’t know that it didn’t have that. In the end, the computer didn’t work. However, we don’t know if that was due to error or the general problem of solving the equations numerically. Even today, it takes a lot of CPU power to do that and we can’t imagine the human computer is very fast.


Is this really practical? In theory, yes, although we don’t have enough friends to put together something that big. You might be able to devise a scheme to do it over the Internet, but that doesn’t seem as impressive. Perhaps at the next Supercon, we’ll try to put together a full adder. You could use people as traditional gates or even relays.

True, back in the old days, a “computer” was, in fact, the job of a person who did math to build things like tables. But that is different still. The Three Body Problem computer could take people with little skill, teach them a simple rule, and then get the benefits of a modern digital computer. You just need a tremendous number of people.

We are always fascinated by how, once you understand the concept, you can make a computer from almost anything. So, using this method, fluidic logic, or marbles, people like the ancient Egyptians, Chinese, or Romans could have had computers. What would they have done with them?

54 thoughts on “Fictional Computers: The Three Body Problem

  1. The human computer was a fun part of the story – although I thought they did a better job of it in the Chinese TV version. Did occur to me that they probably would have had to add error correcting. I think “human gates” would be pretty prone to single bit errors…

      1. In 5he book they go out of their way to talk about error prevention.
        During training and practice, anyone who gets it wrong is publicly executed.
        Here must be error checking going on, since many are killed at first.

  2. One thing not mentioned is clock speed… Even if you get a computer out of this arrangement, you’ll still end up with something with a millionth of the performance of a current pocket calculator. Doing the math by hand is probably faster.

    1. It was explained in the books, the Trisolarans communicate via light waves at long distances, so in the past, they could set up such a computer with them all standing around and communicate via light and reach results much faster than with hand signals. The entire Three Body video game is intended to get humans to think like the Trisolarans did about the problems of their world. It was designed to elicit sympathy for their cause and help them invade when they came, that was the entire purpose of the game. It wasnt intended to teach humans anything about themselves.

  3. A similar idea showed up in _Master_of_the_Five_Magics_ by Lyndon Hardy .. back in 1980.

    Only it was a demon using captive humans in cages to be the individual bits.

  4. That part was interesting, but the show was ruined for me for three reasons:
    1.) The aliens couldn’t figure out n-body problems to any degree of accuracy but they could plot a course from their home system to ours.
    2.) The specific three-sun arrangement used in this fictional Alpha Centauri is one proposed by Richard Montgomery and is a predictable, repeating system.
    3.) The real Alpha Centauri/Proxima Centauri system is actually quite stable (being effectively a binary with a close neighbor).
    4.) The conversations about lying proved that this species fell into the “dumb aliens” category.

    1. Number (4) about lying was THE major downfall of the program. As a kid I used to get told off if I lied AND told off if i told the truth and the adult did not like that particular truth. It was a terrible dilemna – which was the most logical and ethical thing to do? After a four years of getting severe mental punishment, I just got better at lying and also realised the adults concerned were pretty ****ing dumb. It’s actually a really interesting subject which just got glossed over in the film.

        1. The “loss of gravity” (at least in the book) was due to multiple suns lining up close by the surface of the planet. Their combined gravitational attraction opposed the gravitational attraction of the planet, causing everything to begin floating off into space.

          As far as the practicality of the computer as described, it makes a lot more sense given Trisolaran biology, as they communicate telepathically using electromagnetic waves that are visible as they think. The black and white flags thing was a metaphor from the VR game to teach humans Trisolaran history; they were saying “this is how we tried to make sense of our world, in terms you’ll understand”.

          There may be a theoretical solution to the 3-body problem for point masses of equal mass and zero initial velocity, but a trinary system that coalesced out of a stellar nebula would have none of those preconditions. Cixin Liu’s understanding of the idea as written in the 2008 novel is accurate to our current understanding as well — see this (professional quality!) astronomy class presentation from 2022:

          As far as whether Proxima Centauri is really like that, well, I think Liu was more concerned with exploring the allegory of living in chaos at the mercy of enormous forces out of one’s control as a metaphor for the Four Pests campaign and subsequently the enormous construction kick they’ve been on for the last 20some years, something he mentions in an author’s note to the English translation of the book.

      1. Except the trisols are physically incapable of deception.
        Each trisol can directly see the intent of another, and their entire language
        and society is based on the absolute certainly that each individual is always being as truthful as they can be, while understanding that they might be wrong by having incomplete external information.

        Imagine an NPC hearing the table talk at a game, the internal monologue of the players, and having access to the character sheets and dice results.
        That NPC would ALWAYS know if you intended to deceive them.

        Humans have to EXPLAIN the concept of deception to them.

    2. 1) Pretty sure the game is meant to be a narrative of their history. By the time contact is made they’re pretty advanced. I haven’t seen the show, this is what I gathered from the book.

    3. I haven’t watched the film, but I did read and enjoy the books a few years ago. Problem 4 is solved by Trisolaran biology, or what we learn of it. Their thought process is transparent to one another; they don’t have the separation between thought and speech we do. As such, deception and concealment are things they don’t have concepts during the initial stage of contact with humans, but they pick up on it pretty quickly. So not dumb, just different. When given an opportunity to learn these skills, they pick up on them pretty quickly.

      Problem 3 is really only solved by suspension of disbelief. From there, 1 sort of becomes a non-issue at least as I recall the book (Netflix may have screwed this up). I don’t recall the Trisolarans being incapable of reasonably forecasting future solar configurations (within a certain horizon), but they understandably don’t like the arrangement of being part of a cosmic juggling stunt. Fra

  5. Humans are much too smart to be doing only AND/OR/etc gates. It’d be better to give them all a chalkboard with the input and have of each them do binary math, turning each person into an ALU (probably limit it to 8 bits per person). The clock cycle would be much slower (whoever was the slowest at the math) but it’s not like the human binary gates would have fast clock cycle (I assume it’d be several seconds due to the time flipping the sign).

    You’d have to figure out how to parallelize the calculation which is probably impossible for the 3 body problem.

    1. > It would be easy to write rules for higher-level functions, too. For example, different flip flops, multiplexers, and demultiplexers could be rule-based like in Verilog, instead of made from gates.

      Agree. You’d have to balance how smart the “gates” were compared to how much time you want to train them and how many errors they make as complexity grows.

    2. But when you have hundreds of thousands, millions even, of single element human ‘logic gates’ who are innumerate, illiterate and untrained for anything other than the role they were born into you can’t afford to invest the time to assess ability, teach binary math etc, you leverage the massive number of ‘single elements’ and reduce it to the absolute basics of NAND, NOT, NOR, AND, XOR which gives you minimal risk of error because they only need to know the simplest of rules.

      Then you can build your CPU from those ‘gates’ and have a much higher level of confidence that the results are correct.

    1. > True, back in the old days, a “computer” was, in fact, the job of a person who did math to build things like tables. But that is different still.

      My point there was that a “computer” had to have some knowledge of math. Maybe you didn’t know why you multiplied and added those numbers but you did know how to multiply and add. We take for granted today that (most) adults know how to do those things, but if you don’t know it is hard to learn. But you could teach someone to be an AND gate in about 30 seconds even if they are not the sharpest sword in the yurt.

  6. I think this is neat, kind of like the Clacks. But on the other hand, doing calculations with incredibly simple tools is entirely possible, and there’s tons of independent ways of doing it, many of which involve carving notches on sticks. A number of people using tools and producing more than a binary output could outrace this one. A 20th century slide rule is an example of how far sticks with lines can be pushed, but calculating using geometric relationships is far older than that and doesn’t always require as much precision manufacturing, nor does it always require the ability to read and use symbolic numbers.

    For instance, it’s the fault of the ancient Greeks that we have compass-and-ruler “constructions” which use the aforementioned tools to draw out solutions for the various problems whose constructions were discovered. Those included the normal four functions, as well as square roots, and a lot of things that are natively geometric problems – like finding a midpoint, mirroring, drawing the smallest circle that contains a set of points, scaling shapes, finding areas, etc. You can imagine that if your drawing represents a real structure, there’s plenty of occasion to use a specific construction that only solves one kind of problem, although they were able to match up other problems to the constructions they knew, in order to solve them. And actually, things were proven using and about constructions, like the Pythagorean theorem, and they were perfectly willing to use other methods to find solutions to problems that famously weren’t constructible.

    There’s a lot of good explanations of the methods used by various historic tools here, though many are modern variants on similar concepts.

    1. Ook!

      Love the Clacks as a concept and the space in the underlying protocol where the Clacks operators keep the names of fallen comrades alive and ‘send them home’.

  7. In the book “Souls in the Great Machine” by Sean McMullen there is a device, Great Calculor of Libris, that uses humans as computational components. Depending on their intelligence and willingness to learn they are assigned to perform different operations, from simple addition and subtraction to more complex functions, using mechanical aids. They all sit at their desks and perform operations, while data is transferred over “buses” made of metal wires and linkages. The Great Calculor uses two identical “processors” to control each other, and storage is done by runners getting commands from “IO devices” to bring certain information from library. Also it’s Turing complete. And IIRC, bugfixes are applied with a lash.

  8. This reminded me of a Dr Who episode about the Logoplitans. The Logopolitans were a species of humanoid mathematicians from the planet Logopolis. They were well known for their ability to do Block Transfer Computations in their minds. When some were killed it through off the computations.

  9. I did enjoy the books but I’m not precious about the science in my chosen genres of SciFi and fantasy having to comply with or be even distantly related to the laws of physics as long as it’s not just some obviously concocted way to get the author out of a corner they painted themselves into.

    I liked the ‘human’ computer, obviously utterly impractical on the scale suggested but the idea of operation is plausible and would be a nice way to teach logic to kids.

    As for the people floating away, it’s been a while since I read the books but I think I read it as the planetary gravity being overpowered by the greater gravity of the nearby sun rather than gravity of the planet failing?

    1. I read the entire trilogy. It was mostly boring, with mostly unlikable characters. The only interesting aspect of it was its use of fables for conveying information past censorship. The ending is weak and stupid…

    2. Yeah, it’s pretty bad. I don’t get the hype. And of course Netflix is going to make things even worse, and they did. Couldn’t get through the first episode, it’s just grating how poor quality they are these days.

      Also this computer would a. have too high an error rate and b. the Mongols wouldn’t want to do any of this crap for long and would quickly disband and demand that they get dispatched to some place they could pillage and conquer.

      1. Look at my post above, where I write about much better meatware computer, the Great Calculor of Libris.

        I tend to avoid big budget shows that adapt books. After giant hype of GoT, I watched it a bit, then I read the books, and books were worse than the show, while the show itself was rather bad.

  10. It’s an interesting idea but it’s clearly designed to be cool visually rather than a reasonable solution.

    They worked backwards from computers to something humans can do but never worked forwards again to optimise the solution. Instead of making each person a black and white banner you could, with the same or less effort and time, give each person a paper dial and get them to physically add and multiply numbers together without them even needing to be numerate beyond counting to 10.

  11. I love the 3 body problem. I don’t suppose there is a place I could meet other fans lol. These books taught me to appreciate cannibalism in deep space to survive a distant journey. And to appreciate mathematics at a cosmic scale. Beem me up Scotty. Don’t you fucking curse me for this dang it.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.