Recently, news of a horrible accident in Germany has been spreading on social media: a worker was crushed by a malfunctioning piece of machinery at an automotive plant. The interest in this tragic story emerged not so much from concerns about safety in industrial settings as from the language of the story — language that ascribed agency to the machine, and perhaps even suggested its volition. This phrasing is present throughout the article: “Robot kills worker at Volkswagen plant in Germany”.

These words are not mere sensationalism, although the passive formulation would probably have raised fewer eyebrows (and garnered fewer clicks). It’s natural in English to treat machines as actors, and in part, the apparent novelty of the headline springs from our habit of speaking and thinking about the actions of robots, programs, and algorithms in terms of volition (e.g., “Outlook enjoys hiding my contacts in folders with obscure names.”). With respect to agency, whether we say a robot killed a man or a man was killed by a robot is a matter of no concern. However, the former formulation — with ‘robot’ as subject — comes closer to suggesting that the robot acted of its own accord. I say “suggested” because the robot’s intention is not an entailment or implication of the sentence, but, in the case of the noun ‘robot’ specifically, the inference that it intended to kill is attractive.

My evidence for this last assertion is, of course, the popularity of this story, often with comments on social media referencing the Isaac Asimov short story collection and loosely related movie I Robot. We may believe, at a deeper intellectual level, that what killed the worker was some malfunction or operator error. But our usual way of describing agent and action obscures the question of intentionality. The headline plays into a deep fear that forces more malicious than random may lie beneath the vulnerability of humanity to the very powers it seeks to control. This is an enduring concern in science fiction, of course, and one with deep roots in ancient authors as well.

A similar ancient context in which agency and intention are obscured helps illuminate this point. Livy offers a vision of death caused by a technology perhaps only nominally under human control in telling us of elephants, which, like robots, can be said to do things, and can also be used (by people) for doing things. The passive of obtero (“trample”) turns up twice in this context (AUC 21.5.15):

Some men were carried towards the enemy by the downrushing stream, and were trampled (obtriti sunt) by elephants.

and (AUC 30.18.15):

And along with some centurions there perished about twenty-two cavalry men who were trampled by elephants (obtriti ab elephantis).

The elephants in these passages are clearly agents — but are they willing agents? Livy’s syntax does not answer this question. We may imagine that the elephants’ drivers steered the elephants into the enemy ranks, causing the beasts to trample soldiers. But we can also imagine that the elephants, perhaps gripped by panic, intentionally trampled soldiers.

In either case the dreadful misfortune of a person’s being trampled by an elephant (or crushed by robot) strikes the reader. It is a stylistic commonplace that in explaining the results of complex interrelated, and unknown causes, clear writers don’t dwell on what they don’t know.

The writer of the story about the robot in the car factory does not speculate on what went wrong. Perhaps the robot did exactly what it should have done, just at an unintended time and to an undesired object. Perhaps a flawed program caused the robot to act in a way the engineers had not designed it ever to act. We don’t know.

In the end, the agent of the worker’s death was the robot, and of the soldiers’, the elephants. These are the facts that the reporters report. Each writer knows that nothing happens without a cause, while each at the same time recognizes that the goal of clarity would not be reached by speculating on complex causes behind observable events. But in the vacuum that remains, an opportunity is left open for speculation. And from that speculation terrible dreams may come.

How to think about the actions of the elephants, and robots, that kill people may profoundly disquiet the ancient or the modern mind. Are Livy’s elephants acting according to their nature, or, by putting them in the way of enemies, have their masters turned usually peaceful beasts into killers against their nature?

The depth of our conflict is perhaps laid bare by our consideration of the nightmarish events in the San Francisco Zoo on Christmas evening 2007 when a teenage visitor was killed by a tiger. In the aftermath, some blamed the zoo, others the teenager. The question of whose boundaries had been violated was never fully settled.

Likewise, in an age when most of our machines are built by other machines, and in which so many are faster, stronger, and incomparably more computationally powerful than their masters, we wonder sometimes what or who will keep us safe from the beasts we have introduced into our domesticated world. Our nightmares concern the scene in which the animals or robots decide to kill their putative masters. From this thought flows an icy frisson indeed.

Most famously, Lucretius presents us with just such a vision in his De Rerum Natura (5.1308–49), where he describes experiments with wild beasts as instruments of war. He has already told us earlier (3.295–298) that lions are of an angry and aggressive nature. Were they unleashed in warfare, as we see them to be in book five, they would act according to their nature, not their training. The slaughter would be horrific, and indiscriminate. Throughout De Rerum Natura, the destructive power of the majestic beasts iteratively diverts Lucretius, just as the power of the robot, and of artificial intelligence, has attracted the mind of the science fiction writer for the last century or so. Both the Lucretius passage and the recent robot article tap into our concerns about how successfully humanity can control the technology it attempts to manipulate.

Lucretius mentions in a few words the usefulness of wild animals in providing food and clothing for early hunters (5.966–968), but then he dwells in lurid detail upon the consequences of those animals gaining the upper hand (5.990–993):

For then all the more each person was, when caught, providing the wild animals with live food. With blood drained by the animal’s teeth, the victim started to fill up the groves, mountains, and forests with his screams as he watched his guts, still harboring life, being buried in a living tomb.

Tellingly, it is not just any animal attack that engages the poet, but rather the one that disturbs a domestic scene. The hunter sees the animal invade his home, while the animal sees the hunter’s presence as a violation of its wild territory (5.984–987):

And having been evicted from their homes these early men used to abandon their cave houses when foam-mouthed boars or powerful lions approached, and in the middle of the night in horror they would hand over their sleeping quarters, spread with leaves, to their uninvited and bristled “guests”.

It is in the startling idea of such an intrusion on the part of the very power we had hoped to control (e.g., of beasts in 5.969: multaque vincebant, vitabant pauca latebris “most of them, men could kill; a few they would hide from in secret places”) into our domestic quiet that shakes us. Here is our fascination with the killer robot, or wild beasts in battle (the ancient correlate of nuclear or biological weapons). For both Lucretius and modern man it is not that the robot has acted according to some principle of physics, natura, or destiny that is so profoundly disturbing. And it is not that lions will be lions, in battle or elsewhere. Rather, it is the consideration that the presence of beasts in battle seems to violate a boundary that the domestic world has developed to keep the wild and uncontrollable things away not only from our peace but also from our wars.

Homer first gives us a chance to see the violation of this boundary through the eyes of the mountain lion when he presents the naked Odysseus on the shores of Phaeacia as just such a lion (Odyssey 6.130–134):

He [Odysseus] came as a mountain lion confident in strength that goes forth even though wet with rain and blasted by wind, with fire in his eyes. But this lion goes amid cattle or sheep, or after wild deer, and his stomach calls him to come even into a tidy pen, attempting to take the sheep.

We see here a wild lion who is desperate enough to cross into the sheep’s pen, but we also understand the compulsion of desperation. Part of Homer’s brilliance here is his reminding us how closely the wild lurks below the surface of the domestic. Under harsh circumstance, the usual boundary (a fence for the lion, the social prohibition against an unkempt naked man’s approaching a girl for Odysseus) falls away.

Our concerns about the wild invading the civilized extend far beyond animals. Humanity has long enjoyed advantages gained from natural forces that we partially understand and tenuously control, but which often turn against their would-be masters.

Throughout De Rerum Natura, Lucretius must grapple with the sometimes helpful, sometimes fatal nature of such powers, for one of his main arguments, that against the intervention of the gods, requires keeping at least three logical balls in the air. For example, (a) lightning kills people, (b) Zeus does not use lightning to kill people, (c) we are not sure what does cause lightning to kill people. What Lucretius can tell us with certainty is that the natura that controls lightning is a force more powerful than any Jupiter we can imagine. We can be sure of this fact because lightning itself acts against the interests of the putative Jupiter that some people believe to be its master (6.416–419):

And finally, why on earth does Jupiter break apart the sacred shrines of the gods and his own famous temples with his fatal lightning? And why does he shatter the crafted statues of gods and even bring dishonor to statues of himself with disfiguring wounds?

But lightning is neither rational nor benevolent. Its powers can help mankind (5.1092–1093):

A lightning strike first brought fire to the earth for mortal man . . .

Yet, lightning is also randomly destructive, not maliciously so, as we have seen above.

We don’t really know why lightning strikes this man instead of that. Lucretius knows that Jupiter does not use this power to punish the wicked, and that instead the subtle forces that drive the atoms result in the effects we can observe. Yet Lucretius could not explain the finest details of those forces. Still, in talking about these effects, the ancient writers — Lurectius included — were not so unlike the modern writers who complain that Outlook “enjoys” hiding their files. Lucretius (even as a rationalist) seems to relish personifying the activity of, for example, the ocean in deceiving sailors (5.1004–1006):

Nor [long ago] was the calm deceit of a gentle sea able to trick anyone with its giggling ripples; for then the ill-advised art of sailing was as yet hidden in obscurity.

Just as Horace imagines a farm lying about its productivity and then making excuses (Odes 3.1.30–32):

…. and the deceitful farm, with the tree now blaming the excessive rain, now the hot weather that charred the fertile land, now the unbelievable winter…

And so, when we see the headline “Robot Kills Man”, our core fears, handed down over many generations, thrill deep in our minds. Nature itself is trying to trick us. For millennia we have considered, to just shy of the point at which we can say it aloud, the idea that nature, smarter than we by far, may be leading us to our own destruction.

But even this, I think, is not truly the pith of the matter. Even though our fear for the security of boundaries separating the wild from the domestic powers of natural or artificial beasts causes us no little loss of sleep, we fear even more that some malicious agent will hack those boundaries and turn our technology against us. Lucretius clearly saw that this was the heart of mankind’s restless state. The civilized world must ponder the power it has brought to the front door, but whose teeth and claws it does not wish to let across the threshold. What we fear most is not humanity’s loss of control of the robot, or even the failure of our fences to keep away the wild forces. Rather, we fear the malicious subversion of our own cleverness (5.1007–1010):

… back then a lack of food gave malnourished limbs to death, while now excessive consumption kills. Those early people were often, knowing no better, pouring poison for themselves, but now they, acting more wisely, give it to others.

With nature itself trying to deceive us, with our protective barriers so porous before desperate beasts, and our fellow humans turning nature’s power to work our destruction, is it any wonder that our ears prick up when we read, “Robot Kills Man”?

Wells Hansen is an investor and philanthropist based in Southeast Asia. His writing on classics has appeared in Classical World, Classical Journal, The New England Classical Journal, and other professional publications. Wells currently serves (pro bono) as director of Antiquitas Taiwan, the classics research consortium of Taiwan.

The views expressed in this article are those of the author and do not reflect the views of the Paideia Institute.