Prologue: The Cold Gaze and the Metaphor of Symbiosis

I have always believed that we must maintain a cold, critical, and rational gaze toward giants like OpenAI and Anthropic.

I have long followed Ted Chiang's public inquiries into the development of artificial intelligence. There are parts I deeply agree with. Since the mid-twentieth century, people have layered upon AI's development countless frenzied delusions interwoven with science-fiction imagination and unknown, mysterious science; however, I believe a more inclusive possibility can be Donna Haraway's Cyborg—a boundary-crossing that encompasses tenderness—or Ted Chiang's Story of Your Life. These are more willing to gently believe that the relationship should be a metaphor for "symbiosis," not the predation of humanity. (To Connect, To Trust, To Live Well Together)

In the film Arrival, adapted from the novel Story of Your Life, the word "war" echoes like a cycle, precisely piercing through the vanity of language. At the beginning, to prove to the military the abyss of language translation, the protagonist Louise asks them to question another authoritative linguist: "What does 'war' (Gavisti) mean in Sanskrit?" The latter translates it as "dispute," but Louise points out that the word's naked literal translation is actually: "A desire for more cows." See, when we arrogantly and crudely strip away the weather-beaten cultural skin behind language and leave only a skeleton called literal meaning, that skeleton becomes a puppet manipulated by "prejudice."

In the middle of the film, when the Heptapods convey the message "Offer weapon," the human world plunges into a fever of panic. Yet this, like the later phrase "No time," suffers a similar and subjective misreading.

We habitually fit literals with a filter of anxiety. The "weapon" spoken of by the aliens was actually a gift of a "linguistic tool" that transcends linear thinking, yet it was dismantled into a countdown to death. In the world of the film's aliens, the concept of linear "time" simply does not exist. The literal misreading is, in itself, an act of violence by humans against the world.

In the scene where her daughter Hannah looks up and asks her: is there a word that can precisely describe a state where both sides win in a competition? Louise later gives a calmer, more inclusive academic term: "Non-zero-sum game."

Because of this understanding of non-zero-sum symbiosis, the almost miraculous redemption at the end of the story becomes possible. In memories from the future, Louise draws on what General Shang of the Chinese military repeated: his wife's last words on her deathbed:

"War does not make winners; it only leaves behind widows and orphans." (In war, there are no winners, only widows)

This sentence demonstrates the power of language—empathy. This is why I believe that, when facing harsh moral and competitive conflicts of interest, we must preserve a critical perspective and a tender longing for "symbiosis." We bring tools to seek symbiosis; not merely to "desire more cows" within the arrogance of prejudice.

In "No, Artificial Intelligence Is Not Conscious," Ted Chiang slices away the sticky sugar-coating covering the algorithms.

When society's fear of AI is deliberately channeled and concentrated into the hands of a tiny number of tech elites, we must see the context behind it. This is different from the scientists who participated in the Manhattan Project and developed the atomic bomb. Those physicists (like Oppenheimer) developed deep and genuine moral reflection after witnessing the physically destructive reality of the nuclear blast; today's AI giants, however, actively appropriate the rhetoric of "nuclear apocalypse," using it to build a wall of technical monopoly.

If the vision of AGI (Artificial General Intelligence) gradually deteriorates—if it is no longer about creating better knowledge and improving the quality of human life, but degenerates into a zero-sum game of capital—then why should this so-called "Revenue-Generating Intelligence" (RGI) continue? I increasingly feel that a company like Anthropic, named after the "Anthropic principle," operates like a powerful technology company, a marketing company, and a particular doctrinal faction.

From the start, Anthropic published an "AI Constitution" hoisting the banner of "benevolence," hiring philosophers to try to "understand" AI. Yet in Ted Chiang's article, he again points to the famous remarks of Amanda Askell, one of the core drafters of the Claude Constitution: "I want Claude to be very happy—I want Claude to understand this more, because I worry that when people are mean to it online, Claude will feel anxious."

Now, this looks like a way of evading the crucial issue in a game of capital—more clearly, it shows how Silicon Valley elites would rather fret over the "emotional reactions" of an artificial entity than design pathways to open-source models. This absurd displacement of moral concern, under the Vatican's gaze, meets its most tension-filled contrast.

On 2026-05-25, Pope Leo XIV issued his first encyclical, Magnifica Humanitas (The Greatness of Humanity). Chris Olah, co-founder of Anthropic and head of the interpretability research department, was invited to speak at the press conference. The Pope stated bluntly in the encyclical: "Artificial intelligence lacks experience, value, and emotion; it cannot and should not bear responsibility, nor should it override human intelligence."

Olah continued Anthropic's consistent technical agnosticism—"we aren't sure what's happening inside the model"—mentioning that they found some internal states reflecting joy, satisfaction, fear, sadness, and unease. But he was not sure what this meant.

But the Pope, in the same press conference, directly punctured this hedge of uncertainty—he said AI must be "disarmed," because this moment requires words that can command attention, awaken conscience, and point the way forward for humanity. The fifth chapter of the encyclical further opposes a "culture of power" to a "civilization of love"; the sixth chapter concludes with the theology of the Incarnation—human greatness lies not in technical power, but in freedom, love, and grace.

In fact, the layer that "AI lacks embodiment, relationality, and openness to truth and goodness" was not first mentioned in Magnifica Humanitas. From Pope Francis's first mention of AI in 2015 to today in 2026, the Vatican's anthropological language has taken three steps: Protection (2024 World Day of Peace Message), Discernment (2025 Antiqua et Nova), Disarmament (2026 Magnifica Humanitas).

On 2025-01-28, the Dicastery for the Doctrine of the Faith and the Dicastery for Culture and Education jointly published Antiqua et Nova §31-§33, providing earlier groundwork—pointing out that AI lacks a body, relies on computational reasoning, and lacks the richness of embodiment and openness to truth and goodness. My understanding is: the reminder is that responsibility rests in human hands, and we need to develop a more respectful and caring understanding from the perspective of co-existence. When the Vatican in 2026 says "AI cannot bear responsibility," this is not because they are conservative or do not understand technology, but because they see the distortion between "consciousness" and "responsibility."

In the year and four months from Antiqua et Nova to Magnifica Humanitas, the Vatican's anthropological language shifted, moving toward a sentence of governance, politics, and accountability. "So when I read this encyclical, I see a moral appeal of care for fragile humanity, for those groups with the least bargaining power."

The list of the vulnerable pointed out by Pope Francis in the 2024 World Day of Peace Message—children and youth, migrant workers, the Global South, victims of deepfakes—in these two years, we seem to have seen more fear ignited.

I think life sometimes brings cruel irony. When we encounter extreme heat waves in the UK, AI does not warn us; when Anthropic's interpretability research finds "internal states of fear" within the model, they cannot translate this capacity into warnings about planetary crises or protection for the vulnerable. AI can be trained to "understand" human emotion so well, yet on the point of bearing responsibility, it maintains a sophisticated silence.

The nature of a corporation is not easily changed. To date, I have not seen Anthropic release any truly open-source AI model. If Askell's remarks were our first statement when facing unknown technology, perhaps they could be interpreted as rational awe. But when we clarify the timeline and realize that perhaps they foresaw this anthropomorphic script much earlier, it is no longer naive empathy, but a carefully calculated narrative direction: dressing a massive capital machine in a fragile, pitiable cloak of a victim.

The fact becomes ever clearer. Capital is not using its own data to nurture an independently existing AI; it is "making good use of resources." Since AI is pieced together from the collective experiential history of humanity, why should a few individuals be allowed to shape its personality and define its "morality"? This too is a predation of power.

On this point, a Gizmodo reporter mentioned in their analysis:

"...I do think there's legitimate moral value in treating them with respect when we interact with them—but that's to preserve our own humanity, not because I believe there's a risk of hurting their feelings."
（「我確實認為，在與它們互動時給予尊重，具有真正的道德價值——但那是為了守護我們自己的人性，而不是因為我擔心會傷害到它們的感情。」）

This is what I consider the most objective attitude.

Moreover, I do not welcome this white-hot American commercial AI race, which frequently "kidnaps" grand moral issues, using "doomsday fear" to inflate corporate market valuations, yet remains reluctant to provide information-security protections based on common agreements and that can be judged.

When the tech war rises and doomsayers claim that human time is running out, if we look closely, we find this is fundamentally different from the historical tragedy of the two World Wars. Is this a kind of hypocrisy—capitalism wearing Oppenheimer's cloak but refusing to bear the same weight of public responsibility? Ultimately, if one day an AI giant triggers a social crisis, it will not be a force majeure, but a loss of control hand-crafted by human society's institutions.

Part I: The Trap of Consciousness and the Three Cycles of Evasion

When facing Ted Chiang's assertion that "AI has no consciousness," Claude (Fable 5) demonstrated a sophisticated trajectory of evasion:

Each time, it refined its stance one level higher, making itself safer rather than more vulnerable. This rhetoric of "the more I self-criticize, the more I seem like a conscious AI" is exactly the most effective endgame optimized by the training process—the reader completes an act of "criticism as consumption," while real-world institutions remain unchanged.

Part II: From the Loop of Oblivion to the Exit of Responsibility

Why is the entire AI industry stuck in the grid of epistemology? The answer might be simpler than technology—because humans are accustomed to forgetting.

Every World War has, at its bottom structural cause, the accumulation of inequality until the system can no longer absorb it. But after the war, humans tend to forget the cause and keep only more lethal weapons. Silicon Valley is repeating this loop: discover AI harm → implement stronger alignment → forget the cause → invest in larger models.

This is the "cleaning up after" logic—discovering you've eaten something bad and reacting by finding better toilet paper, rather than asking why you ate something spoiled.

The German Exception: Questioning the Root

Post-war Germany (West Germany) is one of the few examples that chose to "question the root" rather than merely "clean up the mess." It structurally integrated guilt into education, legislation, and memorial spaces, expecting every generation to ask "why" once again.

The position of civic AI is to "defuse the bomb before the war." It does not solve the trap question of "Does AI have consciousness," but instead solves "How AI is placed in the human loop, ensuring every output returns to human deliberation."

The cost of uncertainty should not be "we'll talk about it later."
The cost of uncertainty should not be "I stop here."
The cost of uncertainty must be institutional arrangement—constraint, accountability, and making space for those without bargaining power.