Last month, Sriram Krishnan, the former White House Senior AI Advisor posted a question:
Sriram Krishnan@sriramk
Something to think about : what does life look like 25 years from now if AI continues to improve. I don’t think any AI community ( broad tech industry , academia , various timelines predictions) have done a great job articulating a positive long term future for humanity and
3:33 PM · May 20, 2026 · 148K Views
138 Replies · 52 Reposts · 724 Likes
Here’s one version of a positive vision of the future:
Soon we will have language models that are recursively self-improving. Model advancements will accelerate at unprecedented scale. This will eventually lead to artificial general intelligence. Productivity will soar. We become a Type I civilization capable of harvesting all of the Earth’s accessible energy. Material abundance for everyone will follow. With scientific progress advancing rapidly, we will eliminate most diseases. And in a post-scarcity world, universal basic income becomes inevitable.
Reading that, did you breathe a sigh of relief or feel a jolt of optimism about the healthy, prosperous, fulfilling life you are supposed to be heading toward? Or maybe, you are feeling more anxious and dreadful. Sriram’s observation points to something strange. The attempts at a positive vision somehow make things worse. Here is the irony: the rosy future AI promises feels farther away than it did when we first encountered them in science fiction written decades ago.
Despite the utopian promises, people don’t seem to like AI. Here are a few headlines:
Dario Amodei warns that AI could wipe out half of entry-level white-collar jobs.
Students boo tech executives at commencement when told to embrace AI or risk replacement.
Bernie Sanders argues the public should own a stake in the largest AI companies fearing the wealth AI creates would concentrate in a small ownership class.
Gallup polling shows young people using AI heavily while growing more anxious about what it will do to their work, their education and their ability to think.
A viral tweet last month says software engineers in SF earning $500,000 a year describe themselves as candidates for a “permanent underclass” in a post-AGI economy
The words of the positive vision like ‘abundance’ and ‘post-scarcity’ are easy to recite. But they land with a strange hollowness, like a pitch nobody asked for. The march toward an AGI-enabled civilization does not inspire the kind of mass societal optimism that earlier technological leaps did (the moon landing for example).
Most people instead report anxiety, fear, a low-grade dread that does not lift. What is driving it? Why is AI advancement so unsettling, in a way that feels different from any other moment in technological history?
The obvious answer is that this is about money. It’s about jobs, wages, and who ends up owning what. The worry starts off personal and quickly scales to the structural.
A handful of labs own the frontier models and control its release and accessibility. The gains flow to the capitalists owning these labs. The average workers’ tasks get absorbed. Productivity rises, wealth concentrates. The people whose labor is no longer needed are condescendingly being told to retrain.
That materialist reading of AI dread concludes that ultimately this is a fight over resources and power. But that reading also leaves part of the mood unexplained. The economic explanation accounts for the triggering of the fear, but not the texture of the fear.
Beyond concerns about money and inequality, AI anxiety keeps bleeding into topics on dignity, authenticity, dependence, usefulness, agency, and the boundary around human uniqueness. The discussion usually moves very quickly from ‘will I have a job?’ to ‘what will humans be for?’
That is a different kind of question and it needs a different kind of vocabulary.
There is a tradition of thought that offers one and it comes from existential psychotherapy. It was most systematically developed by a Stanford psychiatrist Irvin Yalom. In his 1980 book Existential Psychotherapy, Yalom proposed that a great deal of human psychological distress can be traced to four ultimate concerns:
Death
Freedom
Isolation
Meaninglessness
All of us experience them to varying degrees. Having these concerns does not mean one has some sort of pathology. They are the fundamental features of the human condition and the givens of existing as a conscious, finite being.
Healthy human life, in this view, is built on functioning coping mechanisms: institutions, traditions, beliefs and practices that allow us to live without being crushed by what we know.
This brings me to the core claim of this essay:
The reason that AI anxiety is so intense and resistant to any economic or material reassurance is because AI is destabilizing several of these buffers and coping mechanisms that humans have always used against mortality, groundlessness, meaninglessness and isolation.
The fear begins as material, but the institutions it threatens — work and knowledge, craft and judgment, democratic agency and faith and legacy — are the same ones people have relied on to hold those existential concerns at bay.
In existential psychotherapy, death means more than physical demise. It means finitude. It refers to the frightening realization that your time is limited, your body is fragile, your projects may remain unfinished and, importantly, the world will continue without you. The academic term for what Yalom is after is nonbeing: the bare fact that at some point you will no longer exist, and the universe will carry on as if you never were.
Yalom argues this anxiety is mostly unconscious. Almost nobody walks around thinking ‘I will cease to exist.’ Instead, the anxiety gets converted into other things — fear of failure, restlessness, a sense of being stuck, an obsessive need to leave a mark.
Within the field of psychology, Terror Management Theory has confirmed this experimentally. Since the 1980s, studies have shown that when people are briefly reminded of their own death, they become more patriotic, more attached to their worldview, harsher toward outsiders. The idea of death is too overwhelming to face directly, so the mind doubles down on whatever cultural system promises that you are part of something that will outlast you.
This is where Ernest Becker’s The Denial of Death (1973) comes in. Becker argued that culture itself is a shared ‘immortality project’: a hero-system that lets each person feel their life counts in a story that does not end when they do. Self-esteem, in this view, is the feeling that your existence matters beyond your own lifespan.
The psychologist Robert Jay Lifton catalogued the channels this takes:
biological (having children),
creative (building things, making art),
theological (believing the soul survives),
natural (seeing yourself as part of a larger living system).
Most people are probably not consciously building immortality projects and interpreting their daily actions as such. They are just living their lives: raising kids, writing code, building a reputation, planting a garden, voting or saving for retirement.
Yet underneath all of it is the quiet assumption that these activities connect you to something longer-lasting than yourself. The species itself was the floor under all the other comforts. Even the non-religious could take comfort in the idea that while I will end, humanity goes on, and I was part of that.
AI threatens the immortality project of the human enterprise itself.
If a machine can do what you built your life around doing, the channel that connected your effort to meaning collapses. If the arc of intelligence bends toward something beyond the human, the sense that the species carries your significance forward becomes uncertain.
This is why the fear of ‘AI will end humanity’ and the fear of ‘climate change will end humanity’ sound similar but actually have different metaphysical flavors. Climate doom threatens our survival. AI threatens the sense that being human was going somewhere and that the going was the point.
It raises the possibility that the human project is a phase, that whatever carries meaning forward may not involve us humans at all. The institution under strain is the subconscious assurance that the species will carry us forward.
In the existentialist tradition, freedom refers to the recognition that you are responsible for your own life, with no blueprint, no external authority, no prewritten design for who to be.
At bottom, this recognition is liberating. You are the author of your own life. There is no cosmic script assigned to your role before you arrive. The values you live by, the choices you make, the meaning you find — they are yours truly.
Existentialists saw this as a burden, and a source of dread in its own right. If no script is handed to you, then every choice is yours to answer for, with no authority to blame and no ground beneath you to stand on. Erich Fromm concludes in Escape from Freedom (1941) that faced with the anxiety of authoring a self, people willingly hand that authorship to something larger, a leader, an ideology, a fixed routine, anything that will choose on their behalf and relieve them of the burden.
The existentialists were pointing at something inward which is the freedom to author a self, which no outside force can really touch. But that freedom is never exercised in a vacuum. Authoring a life means choosing from the options the world actually puts in front of you and those options have always had limits. Inner freedom is real, but its reach is only ever as wide as that surface of available choices. And with AI that surface is narrowing.
The labs building the language models are making decisions at a scale no institution has possessed before. They set the boundaries on what you can ask, what knowledge stays accessible, wwhat information reaches you and what judgments remain yours to make. This feels deeply unsettling. The people building the systems are making choices that will affect billions of lives, and the people being affected have almost no say.
Just as I am writing this essay, Anthropic released a new Mythos-level model called Fable 5. The release is a nerfed version of Mythos with safeguards that “detect requests related to cybersecurity, biology and chemistry, and distillation.”
Claude@claudeai
Fable 5’s safeguards detect requests related to cybersecurity, biology and chemistry, and distillation. Users are informed whenever a fallback occurs—on average in less than 5% of sessions. We’ll keep refining the safeguards to reduce false positives.
5:08 PM · Jun 9, 2026 · 939K Views
92 Replies · 101 Reposts · 3.27K Likes
Anthropic has very good reasons for this decision like preventing nefarious use of the model for cyber attacks or making bioweapons. But the structure of the decision leaves a bad taste. A small group of people decided, for every user of that model, what domains of knowledge would be reachable. It has certainly triggered some people for instance.
Teknium 🪽@Teknium
What's crazy to me is that Fable is blocked from life sciences broadly, nerfed even if you get passed the classifiers and filter level blocks. The whole point of AGI/ASI is to cure all diseases. Everything else is just nice to haves. But Anthropic wants to close off that path.
4:50 AM · Jun 10, 2026 · 233K Views
421 Replies · 392 Reposts · 5.32K Likes
bubble boi@bubbleboi
Nobody in history wakes up and chooses to be evil. Hitler didn’t. Stalin didn’t. Mao didn’t. And I’m pretty sure nobody at Anthropic did when they woke up today either. History has this cruel pattern where the people most convinced that they’re saving the world are the ones who
6:18 AM · Jun 10, 2026 · 167K Views
174 Replies · 259 Reposts · 2.57K Likes
Bojan Tunguz@tunguz
Our Anthropic overlords deciding which prompts the peasants are allowed to use.
2:27 PM · Jun 10, 2026 · 570K Views
190 Replies · 1.48K Reposts · 18.9K Likes
The people who build these systems are making the choices before you reach them. The territory over which you actually decide grows smaller. Someone else’s freedom now structures the conditions of your life. The sense of authorship you have had in a vast and free cosmos has been transferred without consent. The individual watches freedom recede from them rather than feel liberated and their abilities augmented by an intelligent system.
Why do we work? Of course, we work to earn a living. At the same time, work is what makes someone expect you. To have a job is to be needed by someone and to occupy a slot in the world that would be emptier without you.
This often goes unnoticed and yet it is the scaffolding a great many people hang their sense of mattering on. We tend not to notice it until it goes away, which is why retirement and unemployment so often cause distress that extends past financial strain into a quiet identity vertigo and a sense of being suddenly extraneous.
This concern maps onto Yalom’s concern of meaninglessness: the recognition that ‘if we must die, if we constitute our own world, if each is ultimately alone in an indifferent universe, then what meaning does life have?’
Many people might be familiar with the exploration of this question from Viktor Frankl, who wrote Man’s Search for Meaning (1946) from the experience during WWII concentration camps. Frankl argued that the primary human drive is not pleasure but the ‘will to meaning,’ the need for life to feel like it points toward something, that it is part of a story that makes sense.
Work has historically been one of the most important meaning-delivery systems.
If work becomes optional, or if the work that remains feels like a task the machine could have done, then a primary coping mechanism for meaninglessness collapses.
When Dario Amodei predicted that all 50% of white collar jobs will be displaced, people are worried more than the material conditions and the economic plight. ‘I will be poor,’ is the first reaction. And very soon that frightening thought turns into a terror more profound, which is ‘I will not be needed.’ Beyond income concerns, people are fearing losing the experience of being necessary.
Modern culture has anchored a great deal of human dignity in the idea of contribution. A person earns their keep by being useful. When a technology can do useful things better, it forces a question that the culture has been careful never to ask directly: what is a person for, if not to be useful?
The institution under strain here is the whole dignity-of-work tradition, the deep cultural conviction, religious in origin and then secularized into the modern ethic of the career, that labor is how a life justifies itself.
The Catholic Church shares this concern. In Antiqua et Nova published in January last year, the church insists on the irreducibility of human dignity. It raises the question of whether a machine can produce something true (whether there is a difference between generating plausible text and bearing witness to reality). In contrast to economists and politicians, the Church is primarily worried about the soul. It is worried about whether a meaningful human life remains possible in a world of perfect simulacra.
The final existential concern is isolation. The common story about AI and isolation is familiar by now. We’ve all heard about the ‘loneliness pandemic’ and headlines where people fall in love with their AI girlfriend/boyfriend. The existential reading of isolation runs on a deeper level.
Yalom distinguishes three forms of isolation. Interpersonal isolation is the ordinary kind: you are separated from others by geography, circumstance, or social distance. Intrapersonal isolation is the fragmentation within yourself, when parts of your experience feel cut off from each other.
And then there is existential isolation which is the unbridgeable gap between any one consciousness and another. He wrote that ‘no matter how close we get to another person, there remains a final unbridgeable chasm; each of us enters existence alone and must depart it alone.’
This sounds extremely bleak. But most of us deal with this existential concern just fine and are usually able to manage the ‘chasm’. THe philosopher Martin Buber gave us the language to describe how.
Buber distinguished between two modes of relating to the world: I-Thou and I-It.
In I-It, you relate to the world as a set of objects to be used, categorized, understood and processed. The other exists for you as a function. A barista who makes your coffee, a profile on your phone, a ticket number in a restaurant queue. Buber’s insight is that in I-It, you are not really having a relationship with another at all. Instead you are having a monologue and you are relating to your own mental representation of the other, not to the other themselves.
In I-Thou mode, you meet the other as a whole person (not as a function). There is no purpose to the encounter beyond the encounter itself. You are present with them and they are present with you. Buber argued that I-Thou encounters are possible beyond between two people. It’s also possible between a person and a tree, a person and a cat, a person and God for example. The core element is whether you open yourself to the encounter without qualifying or objectifying what you meet.
I-Thou requires a real other, that someone could reject you, misunderstand you or walk away or… die. The risk is part of the encounter. A being that cannot refuse you cannot truly meet you, because the meeting is mutual only when both parties are free to end it.
Buber wrote that human life consists of an oscillation between these two modes. Most of the day is I-It. But the moments that make life worth living, the moments of genuine connection, belong to I-Thou.
AI shifts the baseline ratio of I-Thou to I-It across the whole of ordinary life.
LLM can produce language that sounds like I-Thou. It says the right things and remembers what you told it and best of all, it is ever so patient. It’s on demand and never declines.
Yet it has no subjective experience of you. It cannot choose to be present because it cannot choose to be absent. The relationship is structurally I-It despite the convincing output.
Beyond chatbots, agentic systems are another driver of existential isolation and risks chipping away our fundamental need for I-Thou connections. An agent is something you place between yourself and the world to act on your behalf. Useful, often, and the whole purpose is to spare you the encounter. With agents, you relate to the other through a representative, and increasingly the other relates to you through theirs, so that two agents transact where two people might have met.
Consider the idea of agentic dating: “Instead of people swiping on dating apps, their AI agents do the first pass. They scan profiles, talk to each other, explore compatibility, and only hand off to the humans if there’s enough signal”.
With such a system, a domain of human life - romance - that offers the largest surface for I-Thou encounters shrinks as a result. An agentic dating system transforms that meeting into a transaction. Your potential partners have now been reduced from a Thou to a function in your own monologue.
The same logic reaches into friendship, hiring, neighborly contact, any domain where an agent can plausibly stand in for you. In each one the share of your life spent actually meeting another person ‘as a whole without qualification’ shrinks.
To wrap up, let us go back to the original question. Why has no one articulated a convincing positive vision for humanity as AI continues to improve?
It is certainly not because of a lack of trying. In 2024, Dario Amodei wrote Machines of Loving Grace, a serious attempt at an optimistic AI vision: diseases cured, extending lifespans, no more poverty and renewing democracy. The essay contains four long sections on what AI will do for us and one brief section on work and meaning, whose brevity Amodei concedes reflects ‘a lack of clear answers.’
That proportional imbalance is telling of the shape of our discourse today.
The institutional response to AI anxiety has been overwhelmingly material: how to distribute the economic gains, how to retrain displaced workers, how to design predistributive mechanisms so the benefits do not pool at the top. Governments are devising industrial strategies while economists are modeling out future labor market shocks.
All important issues for sure. But underneath the policy-legible anxieties about jobs, wages, and market concentration, there is something more subtle that we rarely hear about: our confrontation with mortality and the collapse of meaning-structures that were more feeble than they appeared.
Anxiety around AI is materially triggered and existentially experienced.
The public fear of AI begins as a fear about jobs, wealth, ownership and power. Then it transforms into something existential since the familiar structure through which people experience usefulness, agency, and human continuity slowly collapses.
A credible positive vision for AI would have to satisfy the material condition (who owns it, who benefits, who is protected) AND the existential condition (what is human activity for, what grounds relationship and meaning, what makes life feel worth living in a world where machines can do so much of what humans used to do).
We have been asking what AI will give us. But perhaps, the real question in the end is what it will ask of us.