Robopocalypse: rigorous, terrifying novel about a robotic campaign to exterminate humanity

Daniel Wilson is a PhD roboticist who made his name with a series of fun, light science books about robot uprisings and similar subjects. But his new novel, Robopocalypse, is anything but fun and lightweight: it's a gripping, utterly plausible, often terrifying account of a global apocalypse brought on by a transcendant AI that hijacks the planet's automation systems and uses them in a vicious attempt to wipe out humanity.

Robopocalypse opens on the first days after the terrible robot war, with Cormack Wallace, a human soldier, contemplating a "black box" containing the war's history as recorded by Archos, the rogue AI that nearly exterminated the human race. After this bit of stage setting, we go back to the months before the war, when a researcher unwittingly creates Archos and then loses control of it when it murders him. A series of small, grisly episodes follow in which automation systems are compromised by Archos and domestic robots, autonomous cars, and other devices turn on their owners. These presage the war that is to come.

But the war, when it arrives, is far more brutal than Wilson hints. Archos possesses perfect global coordination, and so when the killing begins, it is near-total and so swift that humanity hardly knows what's coming. Cars mow down people in the streets (even as their owners, trapped within, scream in horror and beat at the windscreens); elevators and domestic robots conspire to drop terrified people down empty shafts or shake them to jelly; planes crash themselves; buildings seal themselves and asphyxiate their occupants.

But some humans survive, thanks to luck and sheer numbers and a bit of cunning, and these humans live to fight. Slowly, and with little success (at first), humanity strikes back at the robots, learning something about the threat they face. Archos is adaptive, though, and every successful measure evinces a countermeasure, horror piled on horror that will have you glued to the pages.

Like Max Brooks's World War Z, Robopocalypse is structured as a kind of oral history, composed of vignettes that take the form of first person accounts, transcripts, technical documents and so on. This is a great literary device in that it dispenses with much of the stage-business in novels where characters get from A to B so that the reader can see what's going on -- rather, the author is free to jump from anyplace to anyplace else, telling the story with a viewpoint that's both omniscient and intimate.

But Brooks's novel had very little in the way of recurring characters, while Robopocalypse quickly converges on the stories of a dozen or so freedom fighters whose lives gradually become entangled. This recurrence gives Robopocalypse something World War Z lacks: heart, in the form of character arcs, wherein heroes learn and change and grow, and we get to root for them.

The film rights to Robopocalypse have been bought by Stephen Spielberg, who has announced a big-budget feature for 2013. If the script is at all true to the novel, it might just be one of the best and scariest science fiction movies of all time.




  1. I just finished this novel a few weeks back and really enjoyed it. Quick read. The thing that really sold me on it was the way the robots helped each other. Very plausible scenarios all throughout.

  2. “This recurrence gives Robopocalypse something World War Z lacks: heart, in the form of character arcs, wherein heroes learn and change and grow, and we get to root for them.”

    Great review, but I disagree with you a bit there.

    World War Z featured plenty of character arcs and, in my opinion, was overflowing with heart and tragedy. Paul Redekker, the Chinese sub chapter, the President’s character arc (very tragic), the Todd Waino storyline (which extended throughout the novel), the downed pilot chapter (terribly sad), the adult woman who has reverted to her childhood self, the war dogs chapter, the Russian chapters and the Decimation, The French chapter (On ne passes pas!), and on and on.

    If there was one thing World War Z wasn’t lacking, it was heart.

  3. Oh, wow. Another science fiction story about robots.

    Oh wow, another science fiction story about the apocalypse.

    Oh — wow — another film that preys on people’s fears of an arbitrarily plausible future. I mean, anything and everything is arguably plausible — we live in a universe of infinite possibilities — why are stories that explore horse-beaten ones still so cool, huh? I guess that’s just the way it is. People are so easily manipulated. It’s disappointing.

    I know, I’ve got my pair of crabby pants on this morning. But at least I’m wearing pants.

  4. As long as Spielberg doesn’t turn it into another “boy and his dog” movie. He has done that more than often enough.

    1. No, Spielberg will just turn it into another trite “I have to save my family!” Hollywood schlockfest

  5. Zombies? Done and overdone. Robots/AI trying to kill all humans? The same.

    Wake me when you have zombie robots shambling through the streets trying to devour processors. Then you’ll have something!

  6. I dislike this portrayal of AI. Out of all possible motivations for an AI, exterminating humanity is a pretty small target. You’re far more likely to get an AI that wasn’t programmed explicitly to be friendly and therefore simply isn’t careful with humans — they’re made of atoms that it can make use of. Or an AI that was programmed to be friendly, but by someone who doesn’t understand humans.

    I recall reading about someone who wanted to make an AI friendly by making its motivation maximizing apparent human happiness, as measured by facial expressions. This has an obvious and horrifying failure mode: the AI will rip off your face and wire it into a smile, as soon as it has the power to do so.

    It might be a bit easier to test an AI for unfriendliness than for botched friendliness. It’s devilishly hard in either case, and not many people will be testing.

  7. NeonCat, you might find this will wake you up. From a review on Amazon…

    (smart cars stalking pedestrians; human corpses reanimated by machines into zombie warriors)

  8. @NeonCat I read a Charles Stross short story a while ago in an anthology called “Engineering Infinity” (IIRC) where the all robot crew of a deep space ship suffered damage to their CPUs thanks to a gamma ray burst. The robots turned into zombies attacking one another to try and scavenge intact CPUs/brains.

    A sort of a subclause of Rule 34 is at work here.

  9. You know, I’m kind of burnt out on the whole “apocalypse” thing. Humanity is way too focused on its own importance and epic demise. As for the structure of the novel, if World War Z already did it (and did it well), why do it again for what is essentially the same type of story?

    I’m just not going to get pumped up about yet another end-of-the-world tale…

  10. How ’bout a story about robots and computers who *don’t* conspire to end the human race, but do end the human race by accident (accrual of glitches and unforeseen outputs)? Or instead of ending the human race, sending human society into utter mayhem? Awesome. I’ll get started on this.

    1. “How ’bout a story about robots and computers who *don’t* conspire to end the human race, but do end the human race by accident…?”

      If I recall correctly, Sarah Zettel’s “Fool’s War” involves something similar. Wasn’t half-bad, either. (I read it probably a decade ago, though, so it may not have aged well.)

  11. I know someone who is a childhood friend of Daniel Wilson. It is my understanding that he withheld this book until he acquired a movie deal, similar to how the movie and novel versions of 2001 were completed. I think he may have even written the novel after the movie deal was signed. He has also been working closely with Spielberg’s team to help on storyboarding, etc. That’s not to say that the movie couldn’t still suck, but it shouldn’t be due to Spielberg changing Wilson’s vision.

  12. what a great title for movie/book about robots ending humanity. Almost as daringly imaginative as A.I.

    At least it sounds violent.

  13. dhasenan wrote:
    I recall reading about someone who wanted to make an AI friendly by making its motivation maximizing apparent human happiness, as measured by facial expressions. This has an obvious and horrifying failure mode: the AI will rip off your face and wire it into a smile, as soon as it has the power to do so.

    It might be a bit easier to test an AI for unfriendliness than for botched friendliness. It’s devilishly hard in either case, and not many people will be testing.

    I fear any setup where an AI is given control and told to maximize or minimize something, even with the best intentions. The more so for AIs with a larger knowledge base or with the ability to self-improve.

    In your example I can see a bunch of things going wrong even if the AI had improved capabilities to sense true human happiness (say it could use miraculous tech to scan your nervous system):

    The AI removes human brains and wiring up the pleasure centers. And of course to maximize human happiness it has to clone more brains until the maximum biomass sustainable on the earth has been converted into wired-up happy brains. Eventually other planets are colonized so as to gain more resources to sustain additional happy brains.

    Or, how to minimize unfriendliness? Just kill off all humans. AI programmed to not allow humans to come to harm? Then it just has to learn enough about the human nervous system and then alter it (drugs? genetic manipulation?) so that humans are biologically incapable of being unfriendly.

    I think the problem with giving AIs power and setting them to a task is that we have these romantic notions about what makes us happy, based on things we “like” (video games, friends, steampunk, Maker projects, religious experience, the idea of individual identity having value, the idea that we control our own destinies, or whatever). But really we can probably be made happy by just having our biology tweaked a certain way.

    And this is just programming AIs to maximize things we think everyone will appreciate. This ignores the nightmares that will arise from asking an AI to maximize your company’s profits for the quarter or prepare the perfect piece of buttered toast in the morning.

    As anyone’s who’s written up a multi-paragraph document to specify the outcome of a Wish spell in D&D knows: Setting a superior power/intelligence with the job of expediently accomplishing a simple goal is always a matter of rolling the dice of unintended consequences. The more powerful the agency, the more exactly you have to define your parameters.

  14. “it’s a gripping, utterly plausible, often terrifying account of a global apocalypse brought on by a transcendant AI that hijacks the planet’s automation systems and uses them in a vicious attempt to wipe out humanity.”

    Transcendant AI? Ahh… I see… magic. Because if you write about sexy gynoids and then wave a magic wand to magically make them alive… that’s science!

    Modern SciFI thinking about AI’s and robots is exactly analogous too 19th century thinking about fairies and gnomes, clichéd and pedestrian.

    Boring boring boring. Howz abouts commin’ up with sumpin NEW?

  15. @Alex @NeonCat Just finished this book, and what Cory does not mention here is that in the end, killing all humans is actually just a side bar from the AI’s perspective. You don’t get the feeling that Wilson really believes the cliche “here i am and now i am going to destroy humanity” scenario is at all plausible. Worth a read for the scary scenes that are going to stick in my mind for a while! Also is interesting that the story is collected from the AI’s perspective.

  16. Global Economics: So in this future with implacable enemies randomly appearing for no reason and apparent utter disdain for cybersecurity, is the entire world at the same level of economic development with utterly automated infrastructure vulnerable to a single mad computer (who, by the way, has to magically coordinate billions of murder-death-kill activities worldwide every minute)…

    Where are the less-developed regions without total automation? The sleepy towns worldwide where robots are only owned by the priviledged wealthy?

    The mad computer is utterly destroying the economy of the first world by killing off the human principal. So, given this, it would merely mean that the second and third worlds would have an opportunity to dominate. There are around seven billion people on the planet. But less than a billion people have any kind infrastructure that would even remotely be vulnerable to this kind of attack.

    The mad computer isn’t even the most interesting thing in this magical world. The more amazing thing by far is the universal global infrastructure that might even remotely allow it to kill a significant proportion of the world through non-nuclear means. But lets say it turns the weapons of an unsuspecting world against the world. So we’re living in a future where global infrastructure is nearly universal yet everybody is still armed to the teeth – and the mad computer is able to use one nation’s computer infrastructure to take over the entire world DESPITE everybody maintaining a hostile stance against each other?

    So indeed it is a magical world of morons where cyber-security is utterly non-existent and guns and nukes are considered the only means of hostility and the magical mad computer is able to exploit huge gaping security holes that apparently the collective insight of the entire world’s hacker community utterly missed and CERTAINLY never exploited for lulz because GASP who would ever do such a thing?

    In conclusion…

    You can’t call it ‘rigorous’ science fiction while totally ignoring the elephants in the room. The details are more interesting than the premise. Each of these problems can be explained away in the writing, but suffice to say, the mad computer doesn’t do it on his own. He has human help, and this could not be the story of a spontaneous robot uprising so much as yet another global terrorism story with a higher body count.

    1. @ Coherent: My thoughts exactly. A story about robots killing off “all of humanity” implicitly applies that “humanity” consists of only dwellers in first-world industrialized nations with 24-hour a day electricity and modern infrastructure. The vast majority of humanity lives with little or no technology, and billions live on less than 2 dollars a day. It’s the typical myopia of a Western scientist that none of the rest of the non-developed world exists. Or are we to belive that all nine billion people (the current projection for 2050) on earth live in an AI-constructed world of robot cars, robot servants and robot-piloted aircraft? We’re going to the ends of the earth just to find enough fossil fuels to sustain our current level of development.

      I think all an AI system would have to do, if it were really smart, is erase all of the world’s financial records. All the planet’s economies would disappear overnight and we would kill each other off ourselves in the ensuing chaos, doing the AI’s work for it.

        1. From the article cited:

          “Many people will live in the growing number of cities with over 10 million inhabitants, known as megacities…. By 2025, 27 megacities will exist, 21 in less developed countries.”

          2025 is 14 years away. 1997 was 14 years ago. Time flies: and all the world is modern.

  17. Oooh! Oooh! I’ve got a great idea for a sequel!

    The AI, realising toward the end of the war that it is loosing, invents a time machine and sends a robot assassin disguised as a human into the past to take out the mom of the most important resistance leader!

    What do you think? Any milage in it?

  18. Please please PLEASE don’t stick killer robots in line after zombies, giant mustaches and bacon. I don’t want to go through a four year cycle of Matrix-knockoff graphic novels, robot flashmobs and, eventually, lame network TV shows about the Terminator.

    Oh wait…

  19. What I’ve always wondered is why would an AI be so inefficient as to use robots to manually kill humans. Create or just release (you know some military has developed them) near totally fatal disease across the planet. Then mop up whatever is left with your robots or whatever. Then cover the planet in a foot of every nerve agent and poison you can find just to be sure.

    Using robots and machines just looks like a bunch of human militaries fighting it out. Why would an AI care about side-effects, from using disease or poison, that don’t impact it at all?

    That’s not even getting into an actual transhuman AI with the computing power of the human race at it’s disposal. We’d all be dead before we even knew anything has happened. Worse if we’re not lucky. Shoot, it could probably talk us into surrendering to it’s will and we’d do it with a smile on our faces.

  20. *checks amazon… Is the ebook available outside the US?… No, of course it’s not*

    Fuck you, publishers. Fuck you very much.


  21. Well, with the 3rd world, the ai wouldn’t have to worry about hurting its own infrastructure, so I would guess it would just go wild with nukes. No problem with EMP messing up its own systems in 3rd world crap holes.

  22. It’s books like this that make Robert J. Sawyer’s WWW trilogy so much more refreshing. Sawyer’s AI emerges, learns how to take over computer systems, hatches a plot…and then calls the president’s BlackBerry and starts discussing Progress. Instead of killing us, it aids the spread of democracy and responsible economic/environmental systems.

  23. Failsafe design, people! Sheesh. Assume that the software is compromised, under the control of some malicious entity (AI or just a mobster doesn’t matter) and design accordingly…

  24. Someone hide this from Harlan Ellison, quickly! He has a mouth; and he will scream. . .

  25. I want to read this, but I have to say that it isn’t really very plausible. For example cars, even self driving ones, wouldn’t be controllably by an external force. How you you access them? Hint: the Sat Nav and Bluetooth are not connected to the engine computer, and the law mandates that all cars have a mechanical link for the steering and breaks anyway. You can’t change the firmware without physically plugging something in to the car under the bonnet.

    Lifts (elevators) couldn’t drop either. They have a mechanical protection system that stops them falling even if the cable is cut somehow.

  26. How come the newly created A.I. life never wants to “help” it’s human parents?

    It’s always the same: “You’re a disease!”

    And never: “Thank you, would you like to know the secret of cold fusion?”

    Nope it’s always “Want to play a Game?” “Kill, Kill, Kill!”


  27. Sounds a lot like one of the back story lines in Frank Herbert’s Dune series. You know the one about men making machines to do the work then the machines enslaving mankind, leading human soldiers to destroy the machines and put in place a ban on all mechanical forms made in the likeness of man’s mind?

  28. Archos DOES want to help it’s human parents. And it wants robots in all of their current and future forms to be accepted by humanity. And it has all the time in the world to accomplish this. A little killing just happens to be an important part of the long-term plan!

  29. I am fully aware of that fact. However, to say that the emerging mega-cities of the world have or will have any kind of AI-enabled, advanced infrastructure is a stretch:

    GURGAON, India — In this city that barely existed two decades ago, there are 26 shopping malls, seven golf courses and luxury shops selling Chanel and Louis Vuitton. Mercedes-Benzes and BMWs shimmer in automobile showrooms. Apartment towers are sprouting like concrete weeds, and a futuristic commercial hub called Cyber City houses many of the world’s most respected corporations.

    Gurgaon, located about 15 miles south of the national capital, New Delhi, would seem to have everything, except consider what it does not have: a functioning citywide sewer or drainage system; reliable electricity or water; and public sidewalks, adequate parking, decent roads or any citywide system of public transportation. Garbage is still regularly tossed in empty lots by the side of the road.

    In India, Dynamism Wrestles With Dysfunction

    As you can see from the article, it’s difficult to believe that in a city like this, anyone but a small class of elites will have self-driving cars and robot servants. I guess that’s what I find difficult to believe about these future high-tech scenarios – it seems much of humanity is invisible.

  30. I down-loaded this to my kindle on the strength of the first chapter and that was a mistake. Get a couple chapters in and you realize that the author has no idea how to make disparate human voices sound real. There really is more to writing good fiction than having a cool idea for a movie.

    1. Just finished Robopocalypse. I would be amazed if Daniel Martin didn’t have a copy of Max Brooks’ World War Z at his side as he wrote it. Completely derivative, in a much shallower, simpler form. What made World War Z come alive was the rich realistic sociopolitical background in which the fantastic element of zombie plague is introduced. Robopocalypse glosses over its characters and introuduces only two threads in other countries, the comfortably familiar England and Japan. All the interesting threads devolve into a predictable action sequence at the end, all readymade for Spielberg to turn into a lowest common denominator summer movie.

      Someone please make the Asimov detective robot novels into movies…

Comments are closed.