Prey (2002): What is a swarm and does Crichton’s flock of turkeys count?
Recommendation: Skip Prey - the nanorobot swarm is really a flock of turkeys
Michael Crichton’s 2002 bestseller, Prey, should be the perfect book to learn about swarms while relaxing after a long day at work or sipping an adult beverage. The premise is that an artificial intelligence company, amply endowed with the usual “move fast and break things” hubris, has created a swarm of nanorobots. The company releases the swarm into the desert thinking that will accelerate its self-evolution while avoiding any accountability or pesky regulations. Not only is this unethical, which the fictional company acknowledges, it is not artificial intelligence that they are propagating but rather a form of natural stupidity. The company scientists find themselves with the exact same problems as real-world machine learning researchers: Oops, we didn’t think through the learning objective function and look at all the unintended consequences as the AI system goes wonky by optimizing the wrong thing. In this case, the swarm is based on predator-prey models of evolution and has the ability to self-replicate, so the mistake has a higher consequence than regular machine learning research. Humans become the prey of a new predator with a Malthusian growth curve.
Not only should Prey be a good book to learn from, it should be one of my favorite science fiction books because Crichton cites one of my books as a reference. That book is Artificial Intelligence and Mobile Robots: Case Studies of Successful Systems, actually a volume that I co-edited with David Kortenkamp and Pete Bonasso. I don’t think I would have noticed that I was cited except I mentioned I was reading Prey to one of my university colleagues. He indicated that he was reading it too and what did I think about being referenced? Referenced? I was referenced?! That night when I got home from the office, I snatched my copy of the book, turned to the back, and there it was. I was referenced by Michael Crichton! How cool is that?! The next day, I asked my colleague about a plot point. He looked at me blankly, saying he hadn’t actually started reading the book yet. He had turned to the back of the book to look at references, just like you’d do with a scientific journal article. Hmmm, not the approach that I take with fiction, but it is an approach...
But back to the point that Prey should be a great book. Except it is terrible as entertainment and as science tome. The plot is less of a plot and more of a literary version of the parlor game where one person says one line of story and the next person has to build on it, then the next, and so on. Crichton seems to be playing the game with himself, approaching each chapter with a “now I must introduce technology concept X” The wild extensions are more pedantic than amusing or clever.
Crichton’s gift for explaining technology as a natural outcome of telling a story is missing; here it is all exposition and little story. Worse yet, there are no likable characters. Every single one, including the swarm, needs a good bitch-slap. Though in retrospect, that might have been deliberate as the book does not look kindly upon Silicon Valley, which, in 2002, was already showing its win-at-all-costs mentality decades before Elizabeth Holmes and Theranos became the poster child for techno-greed. The book is a page turner but mostly because the reader keeps turning the pages searching for that bit of literary magic that will pull all the threads together or wrap it all up with heartwarming connections. Spoiler alert: the magic never comes.
Prey doesn’t get the science wrong, but it gets the connections between the concepts wrong. A swarm of homogeneous nanobots without centralized control can certainly emulate insect and animal swarms, flocks, and herds. By the way, that is what AI roboticists mean when they say “swarm”- a large number of identical (in software and hardware) robots that do not have centralized control and are weakly coordinated because they may not even be aware of other members of the collective (ants are clueless about other ants), and use indirect communication (like pheromones or proximity). But there’s a large distinction between formations which optimize foraging for food, defends the swarm by balling up into a ball, or gets members to fly in a V to reduce energy and a robot swarm that spontaneously creates a formation that mirrors exactly the person in front of it. Likewise, the leap from insect collective intelligence to deliberative reasoning levels of intelligence is intuitively appealing. But, even now, and certainly not in 2002, there is no scientific support for how that leap would occur.
If you read Prey, you will be introduced to swarms, nanotechnology, predator-prey models, self-replicating machines, programmable matter, and the Game of Life for cellular automata. You can see the influences of Craig Reynolds’ classic paper on how simple it is to replicate swarm and flock formations as well as contributions of the larger artificial life (A-life) community. As we might say in robotics speak: Prey gets the individual technologies right but misses the aggregation.
If you want to learn about swarms, you might be better off just reading the section on swarms in chapter 53 in the Springer Handbook of Robotics, second edition. If you want a good techno-thriller, you might opt for the Revelation Space series with the Greenfly outbreak of terraforming swarm robots. Or skip swarms and just re-read the Andromeda Strain- it’s still Crichton’s best novel.
And if you want to learn more about AI robotics in general, check out my new book Robotics Through Science Fiction: Artificial Intelligence Explained Through Six Classic Robot Short Stories and remember to subscribe to our YouTube channel and newsletter so you won’t miss a thing!