Arrival gave us the perfect framing for AI. We just can’t decide what it is yet
The difference between a gift and a weapon depends entirely on who's wielding it.
This week we’re publishing our first editorial package on MBH4H. We’re asking ourselves the question, “Is AI a good thing or a bad thing?” We decided that the answer is yes.
More importantly, I now think the question of whether AI is good or bad is the wrong question—or at least an unsatisfying one. It lacks the necessary nuance. It’s the kind of binary that feels gratifying to argue but collapses the moment you apply it to reality; it’s an “and,” not an “or.”
And so I decided to write two separate essays arguing both sides: AI as either a good thing or a bad thing—a tool or a weapon—framed around two great movies, Denis Villeneuve’s Arrival and Neill Blomkamp’s Elysium. Not because the truth lives neatly on either side, but because each film illuminates a different dimension of the same reality through the lens of science fiction.
This is the first of those two essays: AI is good—Denis Villeneuve’s Arrival. You can read the Elysium essay here.
Let’s make this clear from the outset: Denis Villeneuve’s 2016 film Arrival is one of the greatest science fiction movies ever made. Period. The end.
With a screenplay by Eric Heisserer based on Ted Chiang’s 1998 short story “Story of Your Life,” and starring Amy Adams and Jeremy Renner, Arrival is not only deeply moving and extraordinarily beautiful to look at; it also has the most alien aliens I’ve ever seen on screen—not to mention the definitive use of Max Richter’s “On the Nature of Daylight” in any movie.
I watched the film again recently—I have a habit of watching movies over and over—and realized that in addition to all the virtues listed above, Arrival is also the perfect framing for the most consequential technological debate of our time: is AI a tool or a weapon?
Arrival is built around a first contact scenario: twelve alien ships hang in the air—in much the same way as bricks don’t—over various cities and countries around the world. Linguist Louise Banks (Amy Adams) is seconded by the American military and charged with deciphering the aliens’ language to understand why they’re here and what they want; no easy task.
Over many trips into the ship and repeated interactions with its two Heptapod occupants, Banks gains an understanding of their extremely complex language. When she finally asks the Heptapods why they’ve come, they respond with a word that can mean “weapon” but could also be interpreted as “tool,” or “gift.” All three, simultaneously. The military interprets this one way, as do the human translators working with the aliens in other ships. Banks, with the patience to actually learn the language rather than just react to it, reads it as another. The global crisis that follows is entirely a product of that mistranslation—a catastrophic failure to sit with the ambiguity of a single word.
(From this point forward there will be major spoilers for Arrival, including the film’s central reveal. If you haven’t seen it, stop here and watch it immediately.)
The twist is that the Heptapods experience time non-linearly—past, present, and future are simultaneous to them. The aliens know exactly what they’re giving us: how humanity will receive their gift, the chaos it will cause, how Banks will resolve that chaos, and how they in turn will benefit from this gift thousands of years in the future.
In deciphering the Heptapod language, Banks herself also learns to experience time in the same non-linear fashion, her past, present, and future melding into a single point of perspective. Which, in turn, leads to the film’s blindsiding reveal: The earlier “flashback” scenes of Banks and her young daughter living through a terminal illness are not set in the past—they are their future. For Banks, the gift of foresight is simultaneously a benefit and a curse: She gets to save the world but learns her future daughter will die tragically young from an incurable disease before the child has even been born—which is why Amy Adams’ portrayal of Banks is so much more somber than the character in Chiang’s short story; the audience has to believe Banks is a mother living with the grief of losing her child for the twist to work. And oh man, does it work.
Cue “On the Nature of Daylight” to devastating effect.
But here’s the thing that struck me on this most recent viewing: the Heptapods don’t simply hand Banks their language and wish her luck; they engineer the conditions under which humanity is forced to adapt and collaborate. The gift they bestow only works once Banks begins to perceive time as the Heptapods do—and reach forward into her memories of a future conversation—allowing her to use that knowledge to change the course of her present and stop the world from tearing itself apart.
And speaking of the world tearing itself apart, let’s get back to talking about artificial intelligence.
Like in Arrival, we humans are still trying to decide whether AI is a weapon or a tool. The trouble is, we don’t have any omniscient aliens helping to guide us, forcing us to choose between the full-throated enthusiasm or doomer dread that sit at either end of a binary scale. Yet, regardless of which side of the fence we sit on, AI technology exists and cannot now be wished away—a fact underscored by how deeply it’s already embedded in almost every stratum of modern life. Investment in AI companies and infrastructure is effectively propping up the entire US economy. Even if that’s a bubble that bursts, the technology will still exist. And yet, despite all this adoption and investment, we are collectively, urgently, and in many cases spectacularly failing to agree on what AI actually is, what it’s for, who it’s supposed to benefit—and at what cost to society as a whole. You know. Existential shit.
Unlike Arrival, however, our technological benefactors are not alien beings with a non-linear understanding of time. Unfortunately for us—and at the risk of stating the bleedin’ obvious—the creators of AI are mere humans. Extraordinarily clever humans, yes. But the CEOs of OpenAI and Anthropic are not Heptapods. Sam Altman does not know the future. Dario Amodei does not know the future. Yet they often speak with the confidence of beings who do. The predictions, the roadmaps, the timelines to AGI, the assurances about what this technology will and won’t do to humanity—all of it is earnest and convincing, and all of it is still entirely hypothetical. It is, at best, conjecture dressed as prophecy. Or as Ted Chiang once brilliantly described AI in general: “Applied Statistics.” The Heptapods’ certainty was earned by omniscience; Silicon Valley’s is just supreme self-confidence. And that gap between genuine foreknowledge and performed certainty is not a small irony. It’s a deeply unsettling one, if you think about it long enough. Personally, I try not to.
There’s a particular scene in Arrival that does more work than it’s given credit for. When Banks is making her case for why the Heptapods should be engaged rather than feared, she argues that this doesn’t have to be us versus them. Cooperation is possible. The gift can be received as intended. It’s not a zero sum game.
I’d love to apply that idea to AI with the same optimism. I really would. The problem is that the Heptapods engineered a situation in which humanity had no choice but to collaborate—the gift is designed to require it; collaboration isn’t incidental.
Nobody has engineered a comparable situation for the development of AI. In fact, it’s the diametric opposite: What we have instead is a race—between companies, between nations, and most sharply between the United States and China—that looks considerably less like Louise Banks carefully learning an alien language and more like the generals reaching for the wrong translation in pursuit of the win. The idea that AI will be developed collaboratively, for the benefit of all humanity rather than the competitive advantage of whoever gets there first? That, I’m afraid, really is science fiction.
And yet. AI is not a tool or a weapon: it is profoundly both—and it likely always will be, because that's what human technological progress does. Technologies designed for resilience and connection rarely stay within their intended bounds. The internet, for instance, began as a decentralized research experiment designed to withstand failure and connect disparate systems; its architecture was built to ensure reliability in an era of rapid technological change. (DARPA's own director from 1970 to 1975 later said the goal included surviving a nuclear attack—though the project's founding purpose, by other accounts, was simply letting a handful of overstretched university computers talk to each other.) What emerged was the greatest communication tool in human history. Yet that same architecture, by its very nature, became a weapon—an unintended consequence that has arguably inflicted irreparable harm on human society.
What’s undeniable is that the same architecture that made the internet indestructible against nuclear attack makes it equally indestructible against disinformation, radicalization, and state-sponsored interference. You cannot separate the two. The tool and the weapon share the same DNA because they are, literally, the same thing.
This is not a new story. The same chemist that gave us fertilizer gave us chemical warfare. The same physics that gave us nuclear power gave us Hiroshima. The stone tool that could skin a deer could kill a man. Every transformative technology in human history has been simultaneously a tool and a weapon. AI is not an exception to that rule. It is the very definition of the rule.
Which is, oddly, where I find something close to comfort. Not because the pattern is reassuring—it isn’t, particularly—but because it suggests that the question of whether AI is a weapon or a tool is, and has always been, the wrong question.
Arrival understood this paradox. The Heptapods’ language was regarded as a weapon by the military and as an extraordinary gift by Banks, at exactly the same time, from the same source. Both conclusions were legitimate until one person did the patient, unglamorous work of actually understanding what was being offered well enough to explain those conclusions to the other side—and convince them to act accordingly. If we consider the Heptapods’ language as a technology, that technology was a constant; it was the humans’ perception of it that changed.
Which brings me to the reason I wrote this essay. Over two years of writing about creativity in the era of AI for this publication, I no longer see AI as a zero sum game; it is literally both a weapon and a tool.
AI is what it is. Like nuclear power, our only real choice is how we’re going to use it, and to what ends. At a personal level, I’ve grown to love using it as a tool on a daily basis—enthusiastically, but not without conflict. I’m not blind to the risks of it being deployed at scale without sufficient care. I also grew so sick of the move-fast-and-take-things attitude of Sam Altman and OpenAI’s co-founder Greg Brockman that I cancelled my ChatGPT subscription and switched to using Claude almost exclusively—with some occasional Gemini thrown in. What AI tools I’ll be using in the future, I have no idea.
The Heptapods knew how their story ended before it began. We don’t have that luxury. Nobody designed AI to save us. Nobody designed it to destroy us—at least, I sincerely hope not. AI arrived, like the ships in the film, to hover over us, vast and opaque, leaving us small beneath it, arguing with each other about what it means for our future. We’ll have to figure that out for ourselves. We have no omniscient aliens to explain its significance.
At least, not yet. I’d feel more optimistic if we did; if the aliens were anything like the Heptapods, I’d trust them more to get this right than we humans. But, then again, Banks didn’t have a Heptapod do the work for her, either. She did it herself—patiently, skeptically, without the comfort of certainty—and that, in the end, is what saved the world. That’s the only model I have for how this goes well. Not omniscience. Just the willingness to do the work of understanding what’s actually being offered, before deciding what to call it.
I don’t know if we’ll do that work. But I know it’s possible. Arrival says so.
