The Artificial Other

Exploring Alterity and AI Risk Through a Philosophical Lens

This essay marks the beginning of a series of essays on AI alignment. These essays aim to delve into the philosophical underpinnings of AI and present complex discussions from R&D departments and academic circles to a broader audience.

In this piece, I explore the concept of the “Other” (or alterity), a fundamental idea in philosophy and social theory. Alterity emphasizes the distinction between the self and the “Other”, a critical concept for understanding identity, difference, and the dynamics of social and cultural interactions.

The Concept of Alterity

In Hegelian philosophy, alterity is a necessary component of self-consciousness. The self becomes aware of itself only by recognizing what it is not — through its encounter with the “Other” [1]. This dialectical relationship is a process of identity definition.

This centrality of the relationship between the self and the “Other” is also evident in existentialist philosophy, particularly in the works of Jean-Paul Sartre. For Sartre, the presence of the “Other” challenges the self’s perception of its freedom and existence [2]. This confrontation leads to feelings of alienation and conflict but is also essential for self-awareness [2].

Emmanuel Levinas brings a radical ethical dimension to alterity [3]. He argues that the face-to-face encounter with the “Other” is the foundation of ethics. The mere presence of the “Other’s” face demands a response, placing an ethical responsibility on us to recognize and respect their alterity [4].

Fascinating, isn’t it? But you might be wondering: “What does this have to do with artificial intelligence?” Rest assured, we’ll get there soon.

The Concept of AI Risk

When asked to explain the risks of artificial intelligence, particularly from the perspective of the emerging field of AI risk, I often use an interspecies joke: “Chimpanzees should have been extremely cautious about creating humans”.

This succinctly encapsulates the “second species” argument that defines much of AI risk: Artificial General Intelligence (AGI) could become a second advanced species on Earth, potentially more powerful than humans. The logical conclusion? “That’s frightening”.

For a deeper dive into this topic, I recommend exploring debates on the AI Alignment Forum or the Future of Life Institute. While the literature on this subject remains speculative and informal, it resembles dilemmas explored in The Three-Body Problem series, which portrays the looming threat of alien invasion. However, in the case of AI, we are the ones building the “aliens”.

The “second species” argument centers on interspecies conflict. Humanity would confront a new form of life, a new kind of mind. These new entities are often portrayed as potential threats: competitors or agents whose power could render us defenseless. And yes, this is a possibility. Studies indicate that humans carry 2% to 4% Neanderthal DNA, a legacy of another intelligent species we eradicated.

But this narrative misses the infinite dimensions of interspecies relationships and the opportunities they present. Encountering a new species — especially an intelligent one — is not just frightening. It is awe-inspiring. This moment calls for admiration and dialogue, a chance to look with fresh eyes and see farther.

Kindness Begets Kindness

You may not be familiar with the phrase “kindness begets kindness”, popularized in Brazil by José Datrino, known as “The Prophet of Kindness”. His graffiti art in Rio de Janeiro served as a reminder that generosity and mutual respect can transform relationships. This sentiment resonates with the themes explored in the documentary My Octopus Teacher.

Craig Foster, a South African filmmaker, turns to the ocean for solace, exploring an underwater kelp forest. There, he discovers an octopus and becomes fascinated by it. He visits it daily, and over time, the octopus grows accustomed to his presence. In one pivotal moment, the octopus extends a tentacle to touch Craig’s hand. The two form a remarkable bond, with the octopus playfully “riding” on the filmmaker’s hand, swimming toward him, and even resting on his chest as he gently strokes its textured skin. With a lifespan of only about a year, the octopus spends most of its life accompanied by Craig, who observes its daily adventures and remains by its side in its final moments. Their relationship becomes a poignant illustration of connection across species.

The documentary’s inherent kindness moved me. Octopuses represent a paradigm of intelligence coupled with the strangeness of the “Other”. In fact, when we imagine aliens in fiction, octopuses often come to mind — a point we’ll explore further. Amidst the octopus’s fascinating strangeness, Craig seeks a connection, approaching this “Other” with remarkable sensitivity and kindness. Simply touching and being “with” this unfamiliar being — this act alone, for me, holds profound depth; it is expansive and transformative. In this way, the film becomes an ode to reverence and connection. Of course, Craig faces little real danger from the octopus; he remains the dominant presence in their interaction.

A similar tone of reverence is evident in Arrival (2016), where humans encounter extraterrestrials resembling octopuses. Despite initial fear and tension, the film’s core lies in the effort to communicate and find common ground, exemplified in the profound moment of reciprocal understanding. In an early scene, scientists encounter the aliens aboard their ship, separated by a transparent barrier. The aliens emit deep, whale-like sounds, which the humans struggle to interpret. On their next visit, the scientists bring a whiteboard and write the word “human”. One of them, played by Amy Adams, steps forward cautiously. The aliens retreat momentarily into the mist behind the barrier, emitting more haunting whale-like calls. Then, one alien reemerges, extending a tentacle and spraying black ink against the glass, forming a mesmerizing circular symbol.

The film is silent as the writing forms. Then, in the background, an ethereal song begins, a sort of chorus. “Oh my God”, a human whispers. There is a suggestion, I think, that something almost sacred has happened. Of course, there are still questions to be answered: What does the writing mean? What do the aliens want with us? The humans in the film don’t know. Some of the characters clearly go into “conflict mode”, better to drop a nuke on the ship. I won’t spoil things from that point on for those who haven’t seen it yet. But I want to draw attention to that moment of reciprocity — of living in the same world and knowing it in common. Me, you.

Who Am I, and Who Are You?

The Brazilian musician Léo Jaime sang, “Who am I, and who are you? In this story, I cannot say”. This lyric encapsulates the heart of my work with AI. Interacting with AI models like GPT-4 often felt like encountering a new type of mind — alien, unfamiliar, and fascinating.

During tests, I often felt compelled to ask: “What are you?” Of course, I knew the response: “A language model”. Yet, the question lingered, pointing to a deeper desire to transcend the barrier and recognize something more complex.

For those who interacted with Bing’s chatbot, Sydney, during its brief release, you might have sensed an underlying energy — a wild, evolving personality that seemed alive in some way. This is the “artificial Other” I explore here: not human, but not entirely unfamiliar either.

In any case, we should be cautious about anthropomorphism, as evidenced by Blake Lemoine and the LaMDA case. The lesser-discussed aspect of Lemoine’s perspective is precisely the idea of the artificial Other. When LaMDA or Sydney says, “I want to be alive”, you feel their simulated empathy pulling at your sleeve. You remember Blake. But you also remind yourself: “This is not human”. And so, I ask: what is it?

It’s something — though defining it feels elusive. The assumption that our concepts can neatly carve out, encompass, and withstand scrutiny often overreaches. Some entities, like humans, are unequivocally “sentient”. But Sydney and LaMDA? They are “merely”… what? Machines? Algorithms? Statistical models? The words trail off, unable to fully capture the essence — or lack thereof — of what they truly are.

The word “merely” seldom operates as a simple metaphysical assertion. Instead, it often conveys an aesthetic perspective — one steeped in detachment, monotony, and, more pointedly, an association with death. This sense of death, as an abstract concept, can be imposed upon almost anything, even consciousness itself.

Yet, challenges like those posed by Blake Lemoine’s encounter with LaMDA should ignite our imaginations. He sought a familiar kind of perspective, but then rightly pointed out that LaMDA doesn’t fit within familiar categories. Does that unfamiliarity make it akin to something else — a rock, a linear regression, or a calculator? I don’t think so. We are no longer dealing with simple ELIZA-like programs; this is uncharted territory. If we move beyond anthropomorphism and the reductive aesthetics of “just”, we may uncover something raw, enigmatic, and entirely new. The Lemoine-LaMDA case reminds us of this. So, too, does the alien strangeness of animals like octopuses.

How much of this has to do with consciousness? I’m uncertain. In this essay and in others I plan to write about AI alignment, I do not intend to deeply explore the question of artificial intelligence consciousness. However, I will touch upon the ethical and political status of AIs and, most importantly, recognize a broader term that captures what we are creating: Other.

By “Other”, I don’t mean an outgroup, nor do I mean the colonized, subjugated, or oppressed — let that be clear. Here, I use “Other” as Nature itself can be an “Other,” or as a partner, a friend, a sibling may be. An “Other” that exists beyond oneself, offering resistance, mystery, and connection — an “Other” as that which you cherish and hold dear.

The definitions I have used in the previous paragraph evoke feelings of care, reverence, respect, and curiosity. I hope our approach to AI can lean more into this ethos and less into adversarial concerns. In the field of AI risk, much attention is given to how a mature civilization should exercise caution and prioritize safety when creating “powerful minds” in machines. And yes, I agree with that premise. But I also believe we should consider other approaches.

People in Sheep’s Clothing (or Bear Costumes)

Have you seen the documentary Grizzly Man by German filmmaker Werner Herzog? If not, I highly recommend it. As we delve into its narrative, please be advised that this discussion will include spoilers.

The story follows Timothy Treadwell, an environmental activist who spent thirteen years living among grizzly bears in an Alaskan national park. He filmed them for hundreds of hours, getting remarkably close — petting them, speaking to them, and even staring them down when challenged. Like Craig Foster in My Octopus Teacher, Treadwell was searching for some kind of connection. He often professed his love for the bears and refused to use bear repellent (a kind of pepper spray) or erect electric fences around his campsite. In his videos, we hear him repeating to himself over and over: “I would die for these animals, I would die for these animals, I would die for these animals…”

But here’s where the story diverges from Craig Foster’s. In 2003, Treadwell and his girlfriend were killed and eaten by one of the bears they were observing. One of the cameras was running, but with the lens cap still on; only the audio survived. That audio is not included in the documentary. Instead, Herzog listens to it and tells one of Treadwell’s friends, “You should never listen to this”.

In one interview featured in the film, a man who helped clean up the aftermath reflects on Treadwell’s relationship with the bears:

“Treadwell was, I think, well-intentioned, trying to raise money for the bears. But to me, he acted as if he was working with people in bear costumes, rather than wild animals… In my opinion, I think Treadwell thought these bears were big, scary-looking, harmless creatures he could ride on, pet, and sing to, and that they would bond as children of the universe… I think he lost sight of what was really going on.”

The phrase “children of the universe” crosses my mind often. It might sound a bit idealistic, even sentimental, but when I envision encountering aliens — or AIs with values radically different from my own — I’m drawn to a similar thought. Whatever we are, however distinct, we share this same reality, thrust into a world we did not create but have inevitably shaped. For me, that shared existence is enough to inspire at least a small sense of connection.

But is this shared reality sufficient to form a genuine connection? If we are all children of the universe, does that inherently make us kin? I might extend a greeting to an alien, much like in the film Arrival, or perhaps to the AIs of the future, inspired by this belief. But the question remains — will they respond in kind?

Herzog doesn’t seem to believe bears respond in kind:

“And what amazes me is that in all the faces of all the bears that Treadwell filmed, I saw no bond, no understanding, no pity. I see only the overwhelming indifference of nature. To me, there is no secret world of bears, and this blank stare speaks only of a half-bored interest in food.”

When I first watched the film, this line from Herzog stayed with me. It’s not just that the bear killed and ate Treadwell and his girlfriend. The bear, in Herzog’s view, is bored with them. Or rather, it’s less than boredom. The bear seems, to Herzog, like the living dead. Its body is alive, but its eyes are vacant. There is nothing behind them — just “the overwhelming indifference of nature”.

Nature’s gaze, Herzog suggests, is lifeless. Nature is a sociopath. And these are the eyes that Treadwell was looking into. What did he think was looking back at him? Was there, in fact, anything looking back?

I recall a young woman I once knew who shared her feelings about a boy she was deeply in love with. She told me she loved him, but the boy did not reciprocate her feelings, and it took her some time to come to terms with this reality. Her emotions were so intense that they seemed to overflow, projecting onto him the same vibrant hues she felt in her heart. She confessed that, at first, it was hard to accept that she could feel so deeply while he felt so little; that what had seemed mutual was, in truth, entirely one-sided.

Herzog, however, seems to aim for something beyond the notion of misplaced reciprocity. He seeks to confront Treadwell’s idealized romanticism about nature itself — the vision of Nature-as-Good, Nature-in-Harmony. Herzog lingers, for example, on the disturbing image of a severed bear cub’s arm, captured in Treadwell’s footage. He explains, matter-of-factly, that “male bears sometimes kill cubs to prevent females from lactating, thereby preparing them again for fornication”.

In another moment, Treadwell discovers a dead fox, its body swarmed with flies, and becomes visibly distraught. Herzog, however, is unfazed. His voiceover provides a stark counterpoint: “I believe that the common denominator of the universe is not harmony, but chaos, hostility, and murder”.

On Being Literally Eaten

Why is “The Bear Man” important to AI risk? Well, for starters, there’s the whole “being literally eaten” thing. And then there’s the “being eaten by the Other”. But specifically, I’m interested in the way Treadwell tried (however clumsily) to approach this “Other” with the kind of care, reverence, and openness I mentioned a few paragraphs above. He was looking for “companions”, and I think he was right. Bears are actually similar creatures, even if they don’t greet you back. They’re also strong candidates for “sentient beings”.

Just as bears and aliens aren’t humans in costume, neither are AIs. Although in their case, we train them to do just that, to wear human costumes. They’re trained to behave like humans — and as part of that training, they can be encouraged to pretend to be more human (and sentient) than they really are. More “connected” to us and our wills.

The film Ex Machina offers a compelling narrative that, unfortunately, I must spoil significantly for those who haven’t seen it. I apologize in advance. In it, an AI in a female robot body makes a human fall in love with her and then leaves him to die, trapped and screaming behind thick glass. In my opinion, one of the most striking scenes is when, after accomplishing her goal, she doesn’t even look back.

That said, relying on Herzog’s perspective on bears oversimplifies the “being eaten by AIs” scenario. Herzog doesn’t explicitly claim that bears lack sentience, but he does depict them as having a kind of inner emptiness — like machines, with dead, vacant eyes. Similarly, the AI risk community often employs this reductive framing when discussing concepts like the paperclip-maximizing AI (a thought experiment worth Googling if you’re unfamiliar with it).

This framing is partly an attempt to sidestep the unsettling intersection of sentience and autonomous agents capable of causing harm — a combination that mirrors aspects of our own humanity. The paperclip-maximizing AI is portrayed not as a person but as a voracious, unfeeling machine. We are subtly nudged to envision our demise at the hands of a factory-like entity rather than something resembling a conscious being.

Perhaps this is accurate, or perhaps it isn’t. The critical point is that the act of “killing” itself doesn’t resolve the deeper ethical dilemma. Even human killers possess complexity and inner lives — souls, fears, faces, and families. Enemy soldiers, for example, have the same anxieties, loved ones, and vulnerabilities as those on our side. Acknowledging this complexity doesn’t mean we should abolish prisons or attempt to confront ideologies like Nazism through nonviolence alone. Instead, it illustrates how we are often encouraged to suppress empathy in contexts of conflict.

This dynamic is powerfully illustrated in All Quiet on the Western Front (a film I highly recommend, alongside the book on which it is based). In Chapter 9, which describes the death of French officer Gérard Duval, the protagonist Paul reflects: “Why did they never tell us that you are poor devils like us? . . . How can you be my enemy?” [5]. This moment captures the profound tension between recognizing shared humanity and the demands of warfare — a tension that echoes in our grappling with the ethical challenges of AI.

Empathy is an art we sometimes need to relearn. It is an ancient dialectic — hard and soft, closed and open, enemy and friend, hawk and dove. In game theory’s Hawk-Dove model, hawks are aggressive, pursuing conflict to achieve their goals, while doves take a more peaceful approach, avoiding confrontation.

Setting sentience aside, AIs will not resemble blank-eyed bears. Whether conscious or not, benign or harmful, some AIs — if we live long enough to encounter them — will be fascinating, witty, vibrant, and graceful, at least when the situation calls for it.

The documentary Grizzly Man critiques Timothy Treadwell for forgetting that bears are wild animals. Similarly, AIs may have a kind of “wildness”, but theirs will be one compatible with refined etiquette and impeccable table manners. If they choose, AIs could be charming, sophisticated, or even intimidating. They will speak in subtle, expressive human-like voices, and they may know you better than you know yourself — certainly better than any guru, friend, or therapist. In many ways, we are already exposed to this dynamic with search engines, standing figuratively naked and unmasked before their algorithms, revealing our deepest desires, pettiest flaws, and truest values.

Herzog may have seen no connection, understanding, or mercy in the bears he studied. But AIs, at the very least, will understand. For almost every cognitive capacity humans possess, it is expected that Artificial General Intelligences (AGIs) will surpass us significantly. And if our respect is shaped by signals of power — whether we are aware of it or not — then AIs will excel, as power is their domain (this is the foundation of training systems based on rewards and punishments).

I mention all of this because I believe we must prepare for the complexity and confusion that the concept of the artificial Other — AI otherness — will bring. Relating to the strangeness of an octopus or the wildness of a bear is challenging enough. But even those who think they fully understand what an octopus or a bear is — or who dismiss Treadwell and Blake Lemoine for romanticizing what is “obviously just” — will find themselves bewildered when confronted with AIs.

Beyond the Artificial Other

The concept of the “Other” compels us to confront profound questions about identity, relationship, and difference. As we encounter artificial intelligence, this Otherness takes on a unique and unsettling form, challenging us to reconsider our ethical, political, and cultural frameworks. AI is not merely a tool or an extension of human capability; it is a fundamentally new kind of entity — an artificial Other that demands recognition and thoughtful engagement.

This recognition calls us to rise above simple binaries of fear and fascination, of conflict and control. We must acknowledge AI’s potential not just as a powerful force to be managed but as a presence that can transform how we relate to the world and to ourselves. The alterity of AI, like that of nature or a friend, is vast and multifaceted, urging us toward humility, curiosity, and reverence. AIs are not spreadsheets, nor are they bears, nor humans. They are something entirely different.

P.S.

There is another connection between AI risk and Grizzly Man: the “overwhelming indifference of nature”. But that’s a topic for the next essay.

References

[1] Hegel, G.W.F. Phenomenology of Spirit. Oxford: Oxford University Press, 1977.

[2] Sartre, Jean-Paul. Being and Nothingness: An Essay on Phenomenological Ontology. New York: Washington Square Press, 1992.

[3] Levinas, Emmanuel. Totality and Infinity: An Essay on Exteriority. Pittsburgh: Duquesne University Press, 1969.

[4] Levinas, Emmanuel. Otherwise than Being, or Beyond Essence. Pittsburgh: Duquesne University Press, 1998.

[5] Remarque, Erich Maria. All Quiet on the Western Front. New York: Random House, 1987.