By Joe Nalven + Claude + Gemini

Oct 15, 2026

Introduction: Fast-Changing Tools, Slow-Changing Drives

Why am I writing this essay? I’ve been annoyed with the doom and boom discussions of AI technology. As if we can separate what it does from what we are doing with it. Yes, there are emergent features, but currently, the discussion rests on who we are. Read on . . .

***

Public discussion of artificial intelligence (AI) is mostly about capability: how much text an AI program can take in at once, how large it is, how much computing power went into building it, and which tasks it can carry out on its own without a person checking each step. Behind much of that discussion sits an unstated assumption: that a more capable technology will produce a more rational society.

That assumption ignores a basic fact. Technology now changes within a single human lifetime. The motivational systems of the human brain took shape over hundreds of thousands of years and, on one influential view, have changed little since the end of the last Ice Age, roughly 12,000 years ago. Evolutionary psychologists call this gap the “mismatch hypothesis.” John Tooby and Leda Cosmides stated it as the claim that our modern skulls house a Stone Age (Paleolithic) mind. Critics argue that human evolution has continued since then and that the Stone Age mind cannot be reconstructed with confidence. This essay needs only a weaker claim: our core social motives are far older than our technology.

The drives that shaped our ancestors in small foraging bands are still the drives we bring to every new tool. They include loyalty to one’s own group and suspicion of outsiders, competition for status, competition for mates, and defense of territory.

A new tool does not remove those drives. It extends their reach. A person who wants to defeat a rival can use a spear, a rifle, or an AI-generated disinformation campaign. The motive is the same in each case. Only the range and the speed have changed.

One way to see this clearly is to look at how an earlier generation imagined the future. The pulp science fiction of the late 1920s and 1930s was full of advanced machines, but its stories were driven entirely by old motives.

The Pulp Record: Buck Rogers and Flash Gordon

Buck Rogers began as Anthony Rogers, the hero of Philip Francis Nowlan’s novella “Armageddon 2419 A.D.,” published in the magazine Amazing Stories in August 1928. A newspaper comic strip drawn by Dick Calkins followed in January 1929. Flash Gordon began as a newspaper comic strip by Alex Raymond in January 1934.

Both stories are set among advanced machines. Rogers, a mining engineer, falls into suspended animation in 1927 and wakes nearly five hundred years later into a world of anti-gravity belts made of a material called “inertron” and of disintegrator rays. Flash Gordon reaches the planet Mongo in a rocket built by the scientist Dr. Hans Zarkov and finds cities in the sky, force fields, and ray guns.

The social arrangements in both worlds are much older than the machines.

Old social orders equipped with new machines

In Nowlan’s 25th century, America is not an advanced democracy. It has been conquered by the Han, an empire the story describes as part Mongolian and part alien in ancestry. The surviving Americans live in the forests as separate armed bands, each holding its own territory. On Mongo, Ming the Merciless commands fleets of spaceships but rules as an absolute monarch whose aims are conquest, domination, and the removal of rivals.

Neither world has used its technology to build new political institutions. Both have used it to equip old ones: the territorial band and the despotic kingdom.

The enemy was drawn from the authors’ own time

The tribal logic of these stories appears not only in the plots but in who the enemy is. The Han of Buck Rogers, and Ming, whose name and appearance drew on the same stereotype, are products of the “Yellow Peril”: the fear, common in American popular culture of the period, that Asian peoples would overrun the West.

The connection to Ming is direct. Alex Raymond’s sources for Flash Gordon reportedly included Sax Rohmer’s Fu Manchu novels, the best-known “Yellow Peril” fiction of the period, along with Burroughs’s Mars stories.

The writers imagined the 25th century and the planet Mongo, but they filled those settings with the ethnic fears of 1920s and 1930s America. This is the strongest evidence the stories offer for the thesis of this essay. The most advanced future these writers could imagine was still organized around their own group and the group they feared.

Swords and ray guns

Flash Gordon repeatedly sets aside his ray gun and fights with a sword or his fists. Part of the reason is genre. Space opera, the genre of adventure stories set in outer space, descended from older adventure romances, especially Edgar Rice Burroughs’s Mars novels, which began in 1912 and placed swordfights on alien worlds.

The genre lasted, however, because readers wanted it. A mass audience preferred duels, rescues, and rivalries to stories about how new machines might change a society. That preference is itself evidence about the audience’s motives. In these stories a ray gun works as a spear with a longer range, and a rocket ship works as a faster horse. The tools improve. What the characters want does not change.

The Motives That Drive the Pulp Plots

The plots of both strips run on three motives that evolutionary psychology treats as central to human social life.

1. Loyalty to one’s own group and fear of outside groups. The main danger in both stories is conquest by a hostile group, not the failure of a machine.
2. Aggression and territory. Disputes are settled by force and by displays of physical dominance, not by negotiation or argument.
3. Competition for mates and status. Much of the tension on Mongo concerns Ming’s attempt to take Dale Arden, Princess Aura’s jealousy over Flash, and Flash’s role as the high-status male who protects Dale. The death rays mostly serve to advance or block these rivalries.

The writers did not invent a new psychology for the future. They used the psychology they knew and placed it among rocket ships.

Individual aggression and organized war

The historical record of war seems to confirm that these motives are permanent. Will and Ariel Durant estimated in The Lessons of History (1968) that only 268 of the previous 3,421 years of recorded history had been free of war. That is war in about 92 percent of years. The Durants did not footnote the figure, so it is better treated as an informed estimate than as a measured result.

The figure also needs care about the unit of analysis, meaning whether one is counting the acts of individuals or the acts between or within whole societies. Aggression by an individual and war waged by a society are different things, and evidence about one is not evidence about the other. The Durants’ count covers the era of written records, which is also the era of agricultural states. It shows that states make war almost constantly. It does not show that Stone Age foragers did.

The best-known test of the forager question points in a different direction. In a 2013 study in Science, the anthropologists Douglas Fry and Patrik Söderberg examined 148 killings recorded by anthropologists who lived among 21 small bands of hunter-gatherers that move from place to place. More than half were committed by a single person. Almost two-thirds arose from accidents, family disputes, executions within the group, or personal rivalries such as competition over a woman. Fry and Söderberg classified most of the killings as homicides, a few as feuds, and only a minority as war. The question remains disputed: Jared Diamond, among others, maintains that societies without states are on average more violent than societies with them. The analysis here is therefore offered as prima facie: a first reading of the evidence that would profit from counterexamples. Even if further evidence placed the origin of organized war earlier, the central dilemma would remain at work: old drives, extended by new means.

One reading of this evidence fits the argument of this essay closely. The old drive is the individual capacity for aggression, especially over mates and personal grievances. Organized war between groups appears to depend heavily on later developments: permanent settlement, stored wealth, social hierarchy, and the state. On that reading, the state was the first great amplifier of an old individual drive, and AI is the latest one.

The Same Motives in the Age of AI

Much of the current debate treats the problems of AI as technical. Do AI programs do what their designers intend? How often do they state false information as fact? How efficiently do they run? Many of the harms people actually report, however, follow the same motives that drive the pulp plots.

Hostility to rival groups and social media algorithms

Two kinds of AI need to be kept separate here. Social media algorithms are the computer programs that decide which posts each user sees, and in what order. Generative AI programs, such as ChatGPT, Claude, and Gemini, produce new text, images, and audio.

These algorithms are designed to maximize what the industry calls engagement: clicks, shares, and time spent. A 2021 study in the Proceedings of the National Academy of Sciences by Steve Rathje, Jay Van Bavel, and Sander van der Linden examined about 2.7 million posts from news outlets and members of Congress. It found that words referring to the opposing political group were the strongest predictor of whether a post was shared on Facebook and Twitter. Each additional such word raised the odds of a share by 67 percent. An algorithm built to maximize clicks and shares therefore learns to promote hostility toward rival groups without anyone instructing it to. The software is new. The reaction it exploits is the old reaction to a rival group.

Deepfakes, AI companions, and trust

Generative AI programs do not create new desires. They act on existing ones. AI companion products act on the need for attachment and approval. Fraud using a computer-generated copy of someone’s voice works because people evolved to trust a familiar face and voice; a cloned voice of a relative asking for money uses that trust directly. Sexually explicit deepfakes of real people combine a sexual motive with an attempt to humiliate the target and lower the target’s status.

Autonomous weapons and aggression

The main concern about AI in weapons is not that machines will become malicious. It is that systems which select and attack targets without a human decision for each strike will lower the cost of aggression for those who command them. Like the disintegrator rays of the pulps, such weapons extend the reach of a commander’s territorial aims while removing the commander’s own physical risk.

Where the ray-gun comparison stops

A ray gun does nothing until someone fires it. Current AI systems differ from it in two ways, and the argument makes sense in both cases.

First, AI is built from human behavior. A chatbot such as ChatGPT or Claude is trained on an enormous body of human writing, so it reproduces human patterns of thought, including tribal ones. It does not only serve human motives from the outside; it carries a statistical record of them inside. This strengthens the claim that AI reflects human nature.

Second, some AI systems now take actions with limited human supervision. This essay explains the risk of people using AI for old purposes. It does not settle the separate question of whether a system acting on its own could cause harm that no person intended. That question deserves its own treatment. The point here is narrower: it is not the only question, and it is probably not the most immediate one.

Conclusion: Govern the Users, Not Only the Tool

Buck Rogers and Flash Gordon were not really about the 25th century or a distant planet. They were about Americans of the 1920s and 1930s, with their own loyalties and fears, holding imagined ray guns. AI likewise is not an alien force arriving from outside human society. It is built from human records and used for human purposes.

If AI threatens society, it does so mostly by amplifying traits we already have: loyalty to our own group and hostility to others, the pursuit of status, the willingness to deceive, and territorial aggression.

The drives themselves, and the moral ideologies that arise to explain or justify them, are always present. Efforts to contain them rise and fall: institutions have reduced violence between individuals and within nations in some places and periods and failed in others, and world organizations have notably failed to prevent war between states. How much real decline these efforts have produced is debated, as a comparison of Steven Pinker’s “A History of Violence” (The New Republic, 2007) with Aaron Clauset’s “Trends and fluctuations in the severity of interstate wars” (Science Advances, 2018) shows, but neither side’s evidence ends the basic dilemma.

Such restraints, often called “guardrails” in discussions of AI, have a poor overall record. The record also shows why some succeed and others fail. Restraints work when someone can enforce them against the people they restrain. A national court can compel a feuding family to accept its judgment. The United Nations cannot compel the states it is meant to restrain. Each of the five permanent members of its Security Council can veto any action against itself or its allies. As a restraint on war, the United Nations has been weak for exactly that reason. Rules for AI face the same test. A rule that the most powerful developers and governments can ignore at no cost will be a United Nations kind of restraint, not a court kind of restraint.

The task with AI is the same kind of task. Better-designed AI is necessary but not sufficient. We also need rules, incentives, and institutions designed on the assumption that the people using these tools will bring Stone Age motives to them. Examples include rules for what social media algorithms may be designed to maximize, clear liability for deepfakes of real people, and limits on which decisions a weapon may make without a human.

The pulp writers found it far easier to imagine new machines than to imagine new people. We should not make the reverse mistake and assume that new machines will produce new people. Our institutions should be designed for the people we actually are.