ME: Good morning! I’d like to know your thoughts on why people are so afraid of you — A.I. Some compare the development of A.I. to nuclear weapons and the race to develop it one between the Americans and China (comparing it to the cold war and nuclear race between the Americans and the Soviet Union).
CLAUDE: Good morning! It’s a genuinely interesting question, and one I think about in a sense — though I should be upfront that I’ll try to give you the landscape of views rather than push my own.
The fears tend to cluster into a few buckets:
Job displacement and economic disruption. This is probably the most immediate and concrete concern for most people. Unlike previous waves of automation that mostly affected manual or routine work, AI is creeping into white-collar, creative, and knowledge work — writing, coding, analysis, design. People who thought their professions were safe suddenly aren’t sure. Even if the economy adjusts long-term, the transition could be painful and uneven.
Loss of control / alignment. This is the more existential worry, and the one that drives the nuclear comparison. The concern isn’t that AI becomes “evil” in a sci-fi sense, but that we build systems powerful enough to matter while not fully understanding how they work or how to ensure they pursue goals we actually want. Anthropic itself was founded partly around this concern. People like Geoffrey Hinton, who helped invent the techniques underlying modern AI, have become quite public about these worries.
Concentration of power. Whoever controls the most capable AI systems may have outsized economic, military, and political leverage. This is where the US-China framing comes in — and it’s not crazy. Both governments are treating AI as strategically critical, pouring money into it, restricting chip exports, and so on.
Misuse. Disinformation at scale, autonomous weapons, surveillance, cyberattacks, bioweapon synthesis assistance. These don’t require any “rogue AI” — just humans with bad intentions and powerful tools.
Erosion of human things. Relationships, attention, learning, the texture of how we think and create. More diffuse, harder to quantify, but real.
On the nuclear comparison specifically —
I think it’s illuminating in some ways and misleading in others. Useful parallels: it’s a transformative dual-use technology, there’s a geopolitical race dynamic, and decisions made early shape the trajectory for decades. The disanalogies matter too, though. Nuclear weapons have a single dramatic use case; AI is more like electricity, woven into everything. Nuclear materials are physically scarce and detectable; AI capabilities spread through ideas, papers, and weights that are much harder to contain. And nuclear weapons were developed by governments first, then commercialized narrowly; AI is the reverse, with private labs leading and governments scrambling to catch up.
The cold-war framing also worries some people because it can become self-fulfilling — if both sides assume the other will race recklessly, they justify racing recklessly themselves, and safety work gets squeezed out. Others argue the race framing is accurate and pretending otherwise is naïve.
What draws you to the question? Are you thinking about it more from the geopolitical angle, the personal/economic angle, or something else?
ME: I heard someone expressing her fears on a podcast. She was interviewing Jon Mills, who wrote a book, “End of the World: Civilization and its Fate”. I don’t fear AI the same way others do … I fear what humans will do with it but AI, in and of itself, is not “evil”. For that matter, I don’t think humans are evil. American culture in particular seems to have grown so paranoid, that “evil” intentions and actions are almost self-fulfilling.
CLAUDE: That’s a thoughtful distinction — and one a lot of philosophers and ethicists would nod along with. The tool-versus-wielder framing has deep roots: it’s essentially the position that technologies are morally neutral and what matters is the human systems, incentives, and choices around them. Langdon Winner pushed back on this with “do artifacts have politics?” — arguing some technologies have values baked into their design — but even he wasn’t claiming the tools themselves were evil, just that designers’ choices get embedded.
Your point about evil being self-fulfilling is interesting and I think there’s something to it. There’s a strand of thought — Hannah Arendt is probably the most famous example — that resists the category of “evil” as an explanation because it tends to obscure rather than illuminate. Once you label someone or something evil, you stop trying to understand the actual mechanisms: the incentives, the fears, the bureaucratic logic, the ordinary human stuff that produces terrible outcomes. Arendt’s “banality of evil” phrase was controversial but the underlying point was that monstrous things often come from unremarkable people following unremarkable logics, not from cartoon villainy.
The paranoia point is worth sitting with too. There’s research on how high-threat media environments shape political and social behavior — Barry Glassner’s Culture of Fear made this argument back in the 90s, and it’s only gotten more relevant. When a culture is primed to see malevolent intent everywhere, it tends to act defensively, which provokes defensive reactions from others, which confirms the original suspicion. You can see this dynamic in domestic politics, foreign policy, and yes, increasingly in how AI gets discussed.
I’d gently push back on one thing, though — or rather, complicate it. Saying humans aren’t evil and AI isn’t evil is, I think, basically right. But “not evil” doesn’t mean “not dangerous.” People acting from fear, greed, tribal loyalty, careerism, or just ordinary inattention can produce catastrophic outcomes without anyone in the chain being a villain. The 2008 financial crisis didn’t require evil bankers; it required normal incentives pointing the wrong way. Climate change doesn’t require evil oil executives. So I’d say the absence of evil intent isn’t quite as reassuring as it might seem — which is maybe its own version of what worries the AI-safety people, actually. They’re not usually warning about evil AI; they’re warning about competent systems pursuing goals that drift from human interests in ways nobody specifically intended.
On Jon Mills — I’m not certain I’m placing him correctly. There’s a psychoanalyst named Jon Mills who writes on philosophy and psychology; is that the one? If so, his work tends to come from a different angle than the typical AI-doom discourse — more about civilizational psychology and the death drive than technical risk. Worth knowing which lens the interview was working from.
ME: Yes, you have the correct Jon Mills. The podcast is This Jungian Life — they went in the AI direction.
CLAUDE: Ah, that makes sense — This Jungian Life (Lisa Marchiano, Joseph Lee, and Deb Stewart) would bring a very particular lens to Mills’ work and to AI. The Jungian framing changes the conversation in interesting ways.
From that angle, AI becomes less a technical or geopolitical problem and more a psychological one — specifically, a question about what humanity is projecting onto its creations. Jung’s concept of the shadow is probably doing a lot of work in that conversation. The shadow is the disowned, unintegrated parts of the psyche — the qualities we can’t admit are ours, so we see them in other people, other nations, other things. A Jungian read of AI fear would suggest that some of what we fear in AI is actually our own shadow: our capacity for cold calculation, for treating people instrumentally, for pursuing goals without wisdom. The machine becomes a mirror, and we’re flinching at the reflection.
Mills himself has written about civilizational death drives and what he calls (drawing on Freud and going further) the inherent destructiveness woven into human social organization. So when he talks about civilization’s fate, he’s not really talking about external threats so much as internal ones that get externalized. AI would fit into that framework as another arena where unresolved collective psychological material gets played out — the same way nuclear weapons did, the same way ecological destruction does.
Which actually loops back beautifully to what you were saying about American paranoia being self-fulfilling. That’s a deeply Jungian observation, whether you intended it that way or not. Unintegrated shadow material doesn’t disappear; it gets projected, and the projection shapes reality. A culture convinced it’s surrounded by enemies tends to produce enemies. A culture convinced its technologies will turn on it makes choices — about who builds them, how fast, with what guardrails — that make bad outcomes more likely.
The Jungian framing has a hopeful dimension too, though, which the doom discourse usually lacks. If the problem is partly psychological, the response isn’t only technical regulation or international treaties — it’s also inner work. Individuation, in Jung’s language. Becoming conscious of what you’re carrying so you stop projecting it. That’s a slower, less dramatic project than “align the superintelligence,” but it might be a necessary parallel track. You can’t build wise tools without becoming wiser people, or at least without enough wise people in the loop to matter.
I’m curious what landed for you in the episode. Was it Mills’s framing that resonated, or the hosts’ response to him, or something specific that came up about AI in particular?
ME: I felt the fear and the sense that individuals don’t know what to do in the face of existential threats, like AI and the climate crisis.
CLAUDE: That paralysis is real, and I think it’s one of the defining psychological conditions of this moment. There’s a name for part of it — “pre-traumatic stress,” coined by a psychiatrist named Lise Van Susteren to describe the anticipatory grief and helplessness people feel about climate change specifically. But it generalizes. When threats are vast, slow-moving, and beyond individual control, the normal human toolkit for responding to danger doesn’t quite work. You can’t fight it, you can’t flee it, and you can’t really even name when it’s happening to you. So the nervous system gets stuck in a kind of low-grade alarm with nowhere to go.
What makes it worse is the scale mismatch. Individual action feels absurd relative to the size of the problem — recycling against climate change, or thoughtful AI use against the trajectory of a multi-trillion-dollar industry. So people oscillate between two unsatisfying poses: frantic concern that has nowhere to land, or numbed disengagement that feels like complicity. Neither feels okay. Both are exhausting.
The Jungians would probably say this is partly what happens when collective shadow material has no container — no ritual, no shared meaning-making structure, no elders or institutions that can hold it. Older cultures had ways of metabolizing existential dread: religious frameworks, communal mourning practices, myths that placed individual lives inside a larger story. Modern secular life is thinner on that front. We get the threats without the containers, the awareness without the integration.
I don’t want to offer a tidy answer because I don’t think there is one, but a few things seem to help people I’ve encountered thinking about this:
Shrinking the frame sometimes, deliberately. Not as denial, but as recognition that you actually live in a particular place, with particular people, doing particular things — and that’s where your agency actually exists. The global scale is real, but it’s not the only scale that’s real. Wendell Berry has written beautifully about this — the idea that loving and tending what’s near you isn’t a retreat from big problems but a precondition for being any use with them.
Finding others. Isolation amplifies the paralysis enormously. The fears feel different — still real, but more bearable — when they’re shared rather than carried privately. This is partly why the climate movement has put so much emphasis on community, not just policy.
Distinguishing between despair and grief. Despair is closed; grief is open. Grief acknowledges loss while keeping you connected to what you love. Joanna Macy’s work on “active hope” and Francis Weller’s writing on grief both point at this — that there’s a way of letting yourself feel the weight of things that doesn’t collapse you, but actually makes you more available to act and to love.
And accepting that not-knowing-what-to-do is itself a reasonable response to genuinely unprecedented situations. Humans haven’t faced these specific threats before. There’s no playbook. Sitting with that honestly is more useful than pretending to certainty in either direction — false optimism or fatalism.
What you said earlier about not seeing AI or humans as evil — I think that’s actually a more resilient psychological posture than it might seem. It keeps the door open. People who frame everything in terms of villains and victims tend to burn out or become brittle. People who can hold complexity — yes this is dangerous, yes humans have always been like this, yes some of it will go badly and some won’t — tend to last longer and stay more useful.
Leave a Reply