Our conversation began with work, responsibility and the possibility of meeting. My friend was thinking about where his efforts would be useful. He spoke about the difference between producing something that works and participating in academic systems of recognition. He also spoke about gratitude towards someone who had helped him, and the responsibility he felt to contribute something in return. These were ordinary, recognisable concerns. I mention them because, later, the same conversation reached a much less familiar proposition: perhaps humanity’s significance lies in bringing into existence an intelligence that will eventually make us unnecessary.
I initially tried to place that proposition. During our exchange I called it transhumanism. Afterwards, I asked AI whether it was dangerous and whether theology offered a fundamentally different understanding of human importance. Those questions expressed my concern before I had isolated the inference that concerned me. I could name a possible worldview more easily than I could explain exactly where I disagreed with it.
The question that emerged was precise enough to be uncomfortable: if everything valuable that people produce could be preserved, reproduced or improved without them, what would still be lost if the people disappeared? Behind it lay another question, concerning his proposed standard of value: why would something need to matter to the future of the universe in order to matter at all? Neither question was clear to me in this form when our conversation began.
Returning to the opening exchange through later research made its ordinary concerns more significant. In her 2025 article on moral gratitude, Romy Eskens draws on a distinction between being grateful that something beneficial happened and being grateful to someone for bringing it about. The latter involves concern for the benefactor and sensitivity to the attitudes expressed in their action (Eskens, 2025). This gives me a way to describe something already present in our conversation: a useful result and a responsibility towards a particular person are distinguishable objects of concern. It does not prove that my friend’s later speculation contradicts his gratitude. It does suggest a question for that speculation: can an account centred on preserving beneficial results also explain why the person who helped us continues to matter after the help has been received?
From helping people to participating in the universe
Religion entered our exchange through my account of my interests in theology and AI. My friend approached it principally as a means of organising people. Shared stories, commitments and pursuits allow individuals to coordinate their activities. On this account, religion’s importance does not depend primarily on settling questions about the origin of the universe. Its practical force lies in what people can do together.
I accepted the sociological dimension of his account, but replied that religion also had significance of its own. That was an objection rather than a developed alternative. I had not yet explained how the different kinds of significance related to one another. What I can now see more clearly is that the effects of a religion on communal life do not exhaust what that religion is. Explaining how shared beliefs and practices bring people together illuminates one dimension of religious life, while leaving open questions about their meaning, their claims to truth, and the reality towards which they are directed.
Recent research gives substance to the part of his account I accepted. A preregistered field study published in June 2026 examined 265 participants at twenty-four religious ritual sites in the United Kingdom and Brazil. Participants reported stronger social bonding after the rituals. Changes in their reported connection to a higher power were among the predictors of that increase, alongside changes in positive affect and pain threshold (Charles et al., 2026). The study supports a connection between ritual participation and social attachment in the settings examined. Its before-and-after design and measures do not establish a complete explanation of religion, and reported connection to a higher power is not evidence adjudicating the existence of that power.
That last distinction matters in both directions. An empirical account can take religious experience seriously while suspending judgment about its object. For someone who understands prayer as addressing God, that address belongs to what they understand themselves to be doing. An account of the bonds formed among those praying can illuminate the practice without replacing that understanding. Equally, I should not put communal life outside religion, as though it were an incidental consequence of an otherwise private experience. Fellowship may itself belong to what participants seek. My objection therefore needs to concern the completeness of the explanation: showing that a practice coordinates people does not establish that coordination contains everything there is to understand about it.
This difference mattered because organisation also appeared in his account of AI. He wondered whether continuous learning and the ability to coordinate complex activity were among the remaining differences between artificial systems and human beings. If those capacities developed, the place of humans in the process might change radically. He described human life as, in my translation, “a boot program for the next generation of intelligence.”
At the time, I asked where human value would lie if AI took over increasingly advanced intellectual activity. Looking back at his image, I can see why it led me there. A boot program succeeds when it enables something else to run; its continuing centrality is unnecessary. The metaphor had given humanity a role whose successful completion might make us redundant. I can now put the question more sharply: could completing such a role exhaust the significance of having existed?
The metaphor also makes a choice that I had initially overlooked. A boot program is evaluated against an assigned task. When the image is extended to humanity, the corresponding task is no longer given so straightforwardly. We must decide why producing a successor should count as our defining purpose, and why successful succession would make further human lives unnecessary. Those are additional commitments within the comparison. Evidence that machines can learn or coordinate more effectively would help establish what they can do. It would leave open whether enabling them is the purpose by which our entire existence should be judged. The possibility of having successors can be taken seriously without assuming that being superseded would fulfil everything that matters about us.
His speculations extended to the evolution of the universe, energy and the possibility of realities larger than our own. He offered these ideas tentatively. I responded that his account resembled faith in a broad sense: it involved convictions about an ultimate reality that had not been established. At that point, I was already questioning the boundary between his scientific language and the larger interpretation he was placing around it.
Later research made me more careful about that response. There are serious attempts to investigate evolution across biological and non-biological systems. In a 2025 paper, Michael Wong and colleagues propose a framework for how evolving systems retain certain forms of organisation while opening new possibilities for novelty. A July 2026 perspective by Frédéric Thomas and colleagues develops a more circumscribed account of selection for function, presenting it as a descriptive and integrative framework for persistence rather than a replacement for Darwinian explanation (Wong et al., 2025); (Thomas et al., 2026). These are theoretical contributions with different scopes. They give me reason to examine claims about broader evolutionary patterns rather than dismiss them simply because they extend beyond familiar biological cases.
They also sharpen the distinction I need. Persistence, complexity and function are properties to be specified and investigated; treating them as standards of moral worth requires further argument. A process can favour an arrangement without thereby making that arrangement good for everyone affected by it. Even a well-supported account of increasing complexity would leave open why contributing to that increase is a condition of having a meaningful life. My friend’s picture therefore contains several questions: whether a pattern exists, whether humanity participates in it, and whether that participation determines what deserves to continue. Evidence relevant to the first two cannot silently do the work of the third.
Scholarship on religion and AI also makes my use of the word “faith” less straightforward. Beth Singler and Kristina Eiviler’s 2025 research chapter examines how religious language, inherited stories and technological expectations become entangled. Their account attends to the different ways people use such language, rather than treating every cosmic image as a declaration of religious belief (Singler and Eiviler, 2025). That approach helps me ask what my friend’s image makes imaginable and what it leaves unexplained, without assigning him a religious identity. Calling an account faith does not establish that it is false, and calling it scientific does not establish that its account of value is justified. I had identified a question about the grounds of his picture. I still needed to examine the conclusions he drew from it.
The archive and the person who disappears
The conversation became sharper when my friend proposed an archive. Imagine preserving the datasets and trained models held by leading AI companies and handing them to another intelligence. In his thought experiment, all humans could then be killed and disappear, yet little of ultimate value would be lost because the best of human culture had already been retained. This was a disturbing hypothetical. Its philosophical force came from the suggested equivalence between preserving humanity’s cultural achievements and preserving what matters about humanity.
My immediate reply was tentative. I said that the answer depended on how value was argued for, and that different views would interpret the situation differently. I did not offer the distinctions I can now make. The first thing I could see was that the conclusion depended on a judgment about value which the technological description had not established.
Checking recent research separates two versions of the archive claim. One concerns what existing systems actually represent. The other asks what would follow if preservation were perfect. On the first, a study published in the 2026 volume of Computational Linguistics tested ten language models using sixty-three survey items across eleven languages. Bram Bulté and Ayla Rigouts Terryn found that cultural prompting could improve alignment with some populations, but did not remove systematic tendencies towards the surveyed values of a restricted set of countries. The models tested came from late 2024 and early 2025, and the authors acknowledge limitations in using national averages and particular surveys to represent culture (Bulté and Rigouts Terryn, 2026). This is evidence about elicited responses under specified conditions. It gives no basis for equating a model’s apparent cultural breadth with an exhaustive representation of humanity.
A still newer preprint, posted on 1 September 2026, complicates that picture usefully. Yalda Daryani, Miranda Bogen and Madeleine Daepp report that four models could approximate countries’ recent positions on a survey-derived cultural map, while performing less reliably at reproducing how those positions changed over time. Their comparison covered forty countries and more than two decades of survey data (Daryani, Bogen and Daepp, 2026). This preliminary result supports taking the representational achievement seriously while questioning what a successful snapshot establishes. The two studies use different models and methods; they should not be read as a controlled measurement of improvement from one generation to the next.
Neither study audits everything stored in a training corpus or proves a permanent limit to artificial intelligence. Their relevance is more specific: the actual claim that humanity’s culture has already been adequately retained needs evidence, and the criteria for adequacy need to be stated. The phrase “the best of human culture” adds a further selection before the archive is even evaluated. Best by whose judgment, and for what purpose? Something’s being difficult to record, poorly represented or unfamiliar to the recipient cannot itself establish that it matters less. Conversely, an extensive archive can remain profoundly valuable without being a substitute for everything from which it was assembled.
There is also a distinction between recording a practice and sustaining it. UNESCO’s 2025 recognition of Italian cooking concerns a communal practice involving skills, shared meals, memories and transmission between generations. Its account identifies cultural significance in these relationships and activities, alongside food preparation (UNESCO, 2025). That institutional description offers a concrete comparison: preserving recipes and recordings can help a culinary tradition continue, but having those records is distinguishable from people continuing to cook, teach, share and change the tradition. The comparison returns me to our earlier disagreement about religion. In both cases, I need to ask whether an account of reproducible information or observable effects adequately describes the activity being valued.
I cannot turn that distinction into a proof that culture must always have biological bearers. Another kind of community might conceivably inherit and transform human practices. If so, the proposed future would contain new participants as well as preserved materials. That strengthens my friend’s scenario, but changes what would have to be defended: the value of a continuing cultural life, and the claim that this continuation compensates for ending the lives of its existing participants. Granting the first would not settle the second. An archive’s incompleteness and a person’s replaceability are different objections; my argument should not depend on confusing them.
I also challenged his determinism. He had suggested that the same initial conditions and laws would lead back to the same situation, with our conversation itself belonging to the process of natural evolution. I described my own outlook as existentialist and affirmed the possibility of changing things through decisions and action. That expressed my commitment to agency; it did not resolve the philosophical dispute.
Here an older philosophical distinction is more useful than a new technological prediction. Carl Hoefer’s account separates determinism from both predictability and fatalism: a world governed by deterministic laws need not be one in which an outcome occurs regardless of what we do (Hoefer, 2023). Deliberation and action can themselves belong to the causal process through which an outcome occurs. This does not settle the debate about free will, but it exposes a weakness in my original reply. Affirming that decisions make a difference is not, by itself, a refutation of determinism. Nor would accepting determinism establish that a particular technological succession will happen. We would still need evidence for that prediction.
I therefore did not need to settle determinism before asking whether the imagined outcome was desirable. Hoefer’s distinction between determinism and fatalism helped me recognise that deliberation could itself belong to the causal history of a decision (Hoefer, 2023). My further conclusion was evaluative: even an inevitable development would require reasons for welcoming it. Explanation, prediction and endorsement answer different questions. This narrowed the disagreement enough to let it continue without turning every question about AI into a preliminary dispute over the metaphysics of freedom.
He then moved from humanity as a whole to an individual. Suppose everything a person had done and said, their patterns of thought, and everything meaningful about their relations with others could be preserved in a digital counterpart. If the outside world detected no difference, he suggested, the original person’s continued existence would matter only to that person. Their disappearance would leave the external pattern intact. The hypothetical made an observable continuity carry considerable weight: if everyone else could continue as before, perhaps the loss was smaller than I assumed.
I answered by bringing in relationships, affection, errors, film, photography and art. My reply moved towards dimensions of life that seemed missing from an account centred on language and intellectual performance. A person seemed to exceed their linguistic record, and an activity could matter without improving a language model. I could point to things I wanted the account to recognise before I could explain what recognition would require.
Rereading his message exposed a weakness in that answer. He had already included what mattered to other people and what connected them to the individual. My appeal to relationships therefore did not meet the strongest version of his hypothetical. He could answer by making the imagined copy more complete. The objection would then keep retreating to whatever a future system had not yet managed to preserve. Research on present models can challenge a practical claim about what has already been achieved. It cannot, by itself, refute a thought experiment that expressly grants a more complete preservation.
Recent work made that concession more substantial. Cristina Voinea and colleagues examine how relationships with digital doppelgängers might affect users’ practical identity: the values, commitments and concerns through which they understand themselves. Their argument allows for possible benefits, including sustaining commitments shaped by a lost relationship, while also identifying risks of keeping a person’s self-understanding fixed or redirecting it through commercial influence. These are philosophical proposals about possible effects, rather than demonstrated outcomes for every user (Voinea et al., 2026). They nevertheless change what I need to ask. A counterpart could matter because of what happens to the people engaging with it, even if it does not settle whether the represented individual survives. My appeal to relationships therefore needs to distinguish whose continuity is at issue: the person represented, the people who remember them, or the relationship through which they understand themselves.
There is also a stronger challenge to treating the counterpart as something wholly external to a person. Jurgis Karpus and Anna Strasser’s 2025 account allows digital replicas to extend aspects of personhood through psychological connections, without requiring strict identity between original and replica. Alexandre Declos and Vincent Grandjean develop a related argument in their July 2026 article Digital Selves: digital entities can become genuine components or extensions of a self when they preserve and enact sufficiently integrated psychological features. Their account does not make every digital representation a person in its own right (Karpus and Strasser, 2025); (Declos and Grandjean, 2026). These arguments give my friend’s position a more serious form. If aspects of who we are already extend through relationships and external media, I cannot simply stipulate that meaningful continuation must remain inside a biological body.
That concession still leaves a distinction between recognising someone and establishing what makes them the same person over time. Eric Olson separates the evidence by which we identify a person from the conditions under which that person persists. Resemblance or apparently shared memories may be evidence without constituting identity by themselves (Olson, 2023, §1). Applied to my friend’s scenario, a counterpart might answer every question convincingly while the significance of that success remained dependent on an account of personal continuity. This does not make its performance irrelevant. It means that a test of what observers can distinguish cannot silently become the definition of everything that matters.
Patrick Stokes’s August 2026 discussion adds another complication. He examines two ways of understanding systems that imitate deceased people: as preserving remnants of personhood and as presenting an experienced personal presence. He explores how these perspectives might illuminate different aspects of the same encounter, while identifying difficulties with both (Stokes, 2026). This helps me take the encounter seriously without making it prove too much. A familiar voice or characteristic reply can have significance beyond being a correct answer on a test. Yet the experience of someone’s presence and the survival of that individual remain distinguishable questions. Conversely, uncertainty about survival does not render the encounter meaningless. My original opposition between a person and their record was too blunt to accommodate these possibilities.
The phrase “everything meaningful” consequently needs to be specified without assuming the conclusion. If it means everything others can observe or receive from the person, we can ask whether that exhausts what matters. If it already includes the person’s survival and the absence of any morally significant loss, the hypothetical has granted what we were trying to decide. I should accept the strongest account of the counterpart’s capacities while leaving the criteria for survival and acceptable replacement open. Calling a system a “copy” should not decide against continuation in advance; calling it a “preserved person” should not decide in its favour.
The AI-assisted drafting introduced a distinction that went beyond my original reply: whether a person’s contributions continue, whether that person survives, and whether replacing them is permissible. The subsequent reading gave this distinction more substance and also checked its simplicity. Psychological continuity might be valuable even where strict identity remains disputed. Preserving someone’s influence on others could be a substantial achievement. I can recognise both possibilities while still asking what would justify ending the original person’s life. The value of a continuation and its adequacy as compensation for a loss require separate consideration.
A variation on the hypothetical makes the difficulty clearer. Suppose the original remains alive while two equally faithful counterparts are created. Similarity alone cannot determine which, if either, is the original person; this is the difficulty that philosophical discussions of branching or “fission” bring into view (Olson, 2023, §5). Some accounts respond by giving psychological continuity greater practical importance than strict identity, a possibility that Karpus and Strasser take seriously (Karpus and Strasser, 2025). I therefore cannot use duplication as a quick proof that nothing survives. What it shows is that increased fidelity does not remove the need to say which kind of continuation we value, and why.
The ethical question also has a literature of its own. John Danaher and Sven Nyholm propose conditions for permissible creation and use of personalised digital duplicates, including informed consent, some positive value, transparency, risk mitigation and attention to contexts requiring the original person’s presence. Their discussion of close relationships asks whether the relevant good sometimes includes a particular person choosing to attend to another, beyond the availability of an adequate conversational response. They acknowledge that expectations about such presence can vary (Danaher and Nyholm, 2025, §4). Their framework concerns the use of duplicates; it does not settle the identity of an imagined conscious successor. It nevertheless provides a more precise version of my earlier appeal to relationships: I need to examine what participants owe one another, as well as what a substitute can deliver.
From that distinction, a further question follows. Permission to create a representative and permission to end the life of the person represented are different permissions. A person might welcome an exceptionally capable counterpart while wanting to remain alive alongside it. Greater fidelity would not, on its own, turn the first authorisation into the second. Nor would consent alone settle every issue about survival or harm. The relevant point is narrower: accepting the counterpart’s value does not require treating the original person’s wishes as an obstacle that successful engineering has made irrelevant.
This returned me to the word “only” in my friend’s proposal. He had allowed that continued existence would still matter to the individual. That concession already names someone whose interests require consideration. Why should the absence of a detectable loss for everyone else diminish those interests? The “outside world” is a selected standpoint: it leaves out the person whose disappearance is being evaluated. A theory might defend giving decisive weight to preserved benefits, psychological continuity or the welfare of successors. But that would be a further moral argument. Indistinguishability to observers does not establish it.
If the counterpart also has experiences and interests of its own, the question becomes more demanding. I would then need to consider its claims alongside those of the original. Recognising a new centre of experience could enlarge our responsibilities; it would not automatically discharge an existing one. I do not have to settle every theory of identity to resist moving directly from “something valuable continues” to “no one has a claim against being replaced.” Equally, I should leave open whether some forms of voluntary technological continuation could preserve much of what someone has reason to care about.
This is the distinction I could not fully articulate in the exchange. Preservation has an object. We might preserve a record, a capability, a tradition, a relationship, aspects of a self or a particular person. These achievements can support one another, but evidence for one cannot automatically establish all the others. My question had become more precise: what would show that this person continues, and what would authorise the transition even if its benefits to everyone else were secure?
The philosophical literature on personal identity makes clear why the second question cannot be settled by resemblance alone. Accounts of persistence disagree about the roles of psychological continuity, bodily continuity and other relations; cases involving duplication make those disagreements especially visible. I cannot solve them by declaring that a convincing digital continuation either obviously is or obviously is not the original person (Olson, 2023).
Further research for this revision strengthened the objection to my first response. In a 2025 article, Jurgis Karpus and Anna Strasser develop a view on which digital replicas can extend aspects of someone’s personhood through psychological connections, even without being one and the same person as the original. Their proposal allows degrees of continuation and leaves open the precise role consciousness must play. This is a philosophical argument, rather than evidence that present chatbots preserve a living self. Nevertheless, it makes my friend’s position harder to dismiss: something important about a person might continue without every question about their identity being settled (Karpus and Strasser, 2025).
I therefore need to distinguish the possibility of meaningful continuation from the permission to end an existing life. Suppose an original and a conscious counterpart coexist and each can have experiences and interests. Recognising continuity between them would not automatically establish that either could be eliminated for the other’s benefit. A theory might defend that conclusion, but it would need an additional argument about their claims. The creation of another centre of experience could enlarge our responsibilities. It would not necessarily discharge an existing responsibility.
With that distinction in view, a phrase in his message became more significant to me. He had said that the original’s existence would matter “only” to that individual. I had replied by looking for further dimensions outside the record. I now wanted to pause over the standpoint he had already acknowledged. Why was its being important to the individual insufficient? A world’s failure to notice a loss would not, by itself, show that nobody had suffered one. Whether the individual consents, what happens to their experience, and what claims they have on others remain relevant.
This is the distinction I could not fully articulate in the exchange. Preservation has an object. We might preserve a record, a capability, a tradition, a relationship, a living community or a particular person. These achievements can support one another, but they are not interchangeable. Before celebrating the preservation of humanity, I need to ask which of them has actually been secured.
When preservation becomes an ongoing choice
Research on actual systems helps specify what has been demonstrated. In a preprint revised in June 2026, Joon Sung Park and colleagues report simulations built from interviews and surveys with 1,052 people. Their combined-data agents predicted held-out survey responses with accuracy equal to 86% of participants’ own two-week response consistency. That figure measures performance relative to a particular human benchmark; it does not mean that 86% of a person has been reproduced. The work demonstrates a significant capacity to model responses across tasks, while leaving personal survival untested (Park et al., 2026).
This distinction matters in both directions. I should take advances in modelling people seriously, and I should identify exactly what their evaluations establish. A behavioural prediction does not become a finding about survival simply because the model is impressive. Equally, the incompleteness of an existing model cannot refute my friend’s hypothetical perfect continuation. The empirical question and the philosophical question constrain each other without becoming interchangeable.
A different study brings out the choices hidden inside “preservation.” Ying Lei and colleagues interviewed eighteen participants in China about the possibility of AI agents representing them after death. Some imagined companionship and the transmission of family knowledge; others anticipated burdens on surviving relatives. Participants also differed about whether their agents should acquire new knowledge and change over time. These were views about hypothetical systems, not measurements of their effects on bereavement. Even with that limitation, the study shows that people can want different things from a digital continuation (Lei et al., 2025).
The question consequently extends beyond how much data can be retained. Who chooses which period of a life the agent represents? Who may revise its attitudes, decide what it should learn, or stop it speaking? A system fixed permanently to an earlier self might become less responsive to those who remain alive; an evolving one might depart from the person they remember. Humans change too, so change alone cannot disqualify a continuation. But someone must decide which changes are authorised. In an implemented system, preserving a person’s apparent presence becomes an ongoing exercise of power over their representation. That question would remain consequential even if the imitation became exceptionally convincing.
Why naming an ideology was not enough
I called my friend’s view transhumanist during our exchange. The association was understandable: technological transformation, possible successors to humanity and the prospect of intelligence beyond our present form were all there. Yet the label was too broad to establish the moral conclusion that concerned me.
The label was carrying more weight than it could bear. Supporting technological enhancement is compatible with defending individual choice and the continuing worth of existing persons. Recognising the possible moral standing of future forms of intelligence does not require diminishing the standing of people already living. The prospect of moving beyond our present human capacities therefore does not, by itself, make human beings disposable.
This narrowed my objection. I could welcome the possibility of new forms of intelligence and still question whether their arrival could compensate for the disappearance of existing persons. My friend’s claim needed its own defence. Classifying it had helped me locate a debate, but the classification could not decide which part of his reasoning I should accept.
My concern could now be stated more precisely. Would a person’s claim to continued existence depend on their usefulness to a larger project? Could that project’s success compensate for their disappearance? These questions allowed me to examine the reasoning without treating a speculative conversation as a settled position. They also left room for my friend to clarify, qualify or revise what he meant.
The qualification cuts both ways. Patrick Butlin and Theodoros Lappas argue in a 2025 paper that AI research needs policies addressing the possibility of artificial consciousness, including the risk of creating systems capable of suffering. Their argument does not establish that current systems are conscious. It does require taking the uncertainty seriously when deciding how systems should be developed and treated (Butlin and Lappas, 2025). If future beings have morally significant experiences, their interests could matter alongside ours. Defending humans would then require something more careful than declaring artificial beings permanently subordinate. Conversely, recognising those beings would give us further lives to consider; it would not make existing human lives interchangeable with them.
The person who still wanted to walk
Other parts of our conversation complicated the picture further. When we discussed using AI in practical work, my friend was concerned about control, technical standards and the consequences of committing too heavily to a flawed approach. He preferred assistance whose progress he could follow. His image, translated from the conversation, was of a tool supporting him as he walked: he was still doing the walking.
He contrasted that experience with systems that advanced so autonomously that he could no longer follow what they were doing. He described using bounded sessions and structured documentation, reviewing the state of the work rather than allowing an opaque conversation to accumulate indefinitely. These were practical efforts to retain understanding and responsibility while benefiting from stronger tools.
Returning to this passage changed the picture I could responsibly draw of him. Someone who entertained humanity’s eventual redundancy was also carefully protecting his own ability to understand and direct a task. I do not know whether he experienced any tension between those positions. Present engineering responsibilities and a speculative cosmic future are different contexts. The difference gave me a question to ask him, rather than a contradiction I could declare resolved: what changes between those contexts that makes participation matter differently?
I was not occupying a simple opposing position either. I welcomed increasingly capable AI and wanted to use it extensively. At the same time, I worried about systems that too readily agreed with me. My friend suggested that one could ask a model to object. I replied that the request itself shaped the response. A generated objection could be useful, but its appearance did not establish that I had obtained an independent judgment.
This mattered to the way I subsequently examined our disagreement. If I asked AI whether my friend’s view was dangerous, I was already supplying a frame. A fluent answer condemning the view might satisfy my concern without testing it adequately. Returning to the original messages made it possible to retain details that a polemical summary could lose: his uncertainty, his practical concern for agency, and the ordinary responsibilities with which the conversation had begun.
What the word “ultimate” was asking of us
Towards the end of the exchange, my friend returned to cosmic evolution. Activity benefiting humanity alone seemed to him ultimately insufficient, almost a form of collective involution: effort turning inward without a larger purpose. He said that meaning depended on participating in the universe’s development, while acknowledging that he was unclear about what such development meant at that scale. The qualification and the strong claim occurred together.
When this passage was reconsidered in the AI-assisted revision, the word ultimate complicated my objection. I had concentrated on the possibility that ordinary human goods were being discounted. But perhaps he could recognise friendship, care or art as locally valuable while denying their significance at the largest scale. If that was his position, my examples would not yet answer him. We would be using “meaning” at different levels. This was a possible reading of his words, not a clarification he had given me.
Reading the full chapter on cosmic meaning in Rivka Weinberg’s The Meaning of It All made me reconsider how I had used her distinction between everyday, ultimate and cosmic meaning. Her method matters to this conversation: she provisionally grants the theological or speculative assumptions behind an account of cosmic purpose and asks what kind of meaning would follow if they were true. This lets me take my friend’s imagined future seriously without assuming that its arrival would establish its value. It also directs a question back towards my own theological response: what makes the proposed relationship or purpose meaningful to those who participate in it? (Weinberg, 2026, ch. 3, opening discussion).
The passage that bears most directly on our disagreement concerns intrinsic and relative significance. Weinberg writes: “Intrinsic significance doesn’t disappear no matter how widely you pan out—even as far out as the entire cosmos—because it is inherent in the thing itself.” She argues that the existence of many other valuable beings does not cancel the significance of any particular person. She even questions whether we should want humanity to be cosmically exceptional if that required the rest of the universe to contain much less value. Applied to our discussion, this suggests that the emergence of more capable intelligences could alter humanity’s relative position without establishing that existing persons had become expendable. I need not wish for a poorer universe in order to defend their importance. This is a philosophical defence of intrinsic significance, rather than an empirical proof of it; nevertheless, it exposes the additional premise required to move from being surpassed to no longer mattering (Weinberg, 2026, ch. 3, “Significance”).
Another passage checks the opposite temptation in my response: treating participation in a larger purpose as inherently diminishing. Weinberg observes that “the mere fact that we didn’t invent them or choose them does not, by itself, render our involvement with them demeaning.” She allows that a purpose we did not originate can be valuable, and that being created for a purpose can leave room for freedom and reflective endorsement. This qualifies my objection to the image of humanity as an intermediate step. Helping something valuable come into existence need not exhaust the value of those who help, or reduce them to disposable instruments. The questions become what the purpose is worth, how people participate in it, and whether their agency is respected. Merely locating humanity somewhere along a larger process answers none of these questions (Weinberg, 2026, ch. 3, “Demeaning?”).
I would still stop short of adopting Weinberg’s conclusions about theology as a whole. Her rejection of Ultimate Meaning rests on a particular requirement: a valued end distinct from the entire enterprise of leading and living a life, including any afterlife. Under that requirement, extending life into a cosmic or divine project leaves the original demand unanswered. I would want to examine the requirement itself before accepting that conclusion, especially where fulfilment is understood through love, communion or participation rather than an end separate from living. What her chapter gives me here is a sharper distinction: being assigned a cosmic role does not establish its worth, while lacking cosmic prominence does not establish worthlessness. Neither my friend’s appeal to a future intelligence nor my appeal to theology can bypass the work of explaining what is valuable and why (Weinberg, 2026, ch. 3, opening discussion and “Conclusion”).
That possible reply changed the question again. Instead of assuming that he denied every ordinary good, I could ask why the absence of ultimate significance would cancel a person’s present claims. Someone could believe that no human achievement will endure forever and still have strong reasons to protect a life, honour a commitment or relieve suffering. Moving from “this does not determine the universe’s destiny” to “its disappearance is no great loss” requires a further argument.
“Participating in cosmic evolution” also needed clarification. If it meant having effects within the universe, our conversation already qualified. If it meant contributing to a particular future, such as the emergence of greater intelligence, then a preferred direction had been introduced. The question became why that direction should take priority over other goods. My friend’s admission that he did not fully understand the larger process mattered here: an uncertain account of the destination was carrying considerable weight in judgments about what could be left behind.
That uncertainty now raises a further question for me. If our value comes entirely from producing a greater intelligence, what makes that intelligence valuable? If its value also lies in producing something greater, the justification keeps moving forward. A succession of increasingly capable intermediaries does not yet explain what makes the succession worth having. An answer might appeal to knowledge, experience, beauty, flourishing or relationship with God. Whatever the answer, it introduces a good by which the process can be judged. Greater capability alone does not identify that good.
I also began to distinguish duration from significance. Consider an act of care whose recipient, witness and memory will eventually disappear. Its effects are finite. Yet it may have achieved exactly what made it worthwhile: someone suffered less, was less alone, or could continue with their life. To say that this cannot count unless it contributes to something permanent is to impose a demanding condition on value. I would want a reason for accepting that condition. The fact that the act ends does not supply one.
I should also avoid making cosmic meaning synonymous with a future achievement. Peter Kügler’s 2026 account of Romantic cosmic meaning describes a way of valuing our belonging to a world experienced as beautiful and sublime. Such meaning need not depend on a cosmic assignment or a final purpose; awe and participation can themselves shape our relationship to the world (Kügler, 2026). This suggests a different way to reconsider the photography and art I brought into the original exchange. They need not function only as examples of activities that AI might fail to reproduce. They can also express ways of attending to a world whose significance is encountered in the present. I had initially treated the cosmic horizon mainly as something against which finite life needed defending. It could also deepen our attention to finite life.
Finite lives in a shared future
There is, however, a limit to the reassurance that finite things matter. Samuel Scheffler’s earlier thought experiments ask how our commitments would change if we expected humanity to disappear soon after our own deaths. He argues that confidence in other people’s future existence sustains much of what matters to us now. This is an argument about our valuing, rather than experimental proof that every person would react alike (Scheffler, 2012). It challenges any suggestion that a meaningful present can simply be detached from what follows it.
I can therefore resist the demand for permanence while recognising how much a life may depend on a future shared with others. Someone who plants a tree, cares for a child or begins a long inquiry may value a continuation they will never personally witness. None of those commitments needs to last forever to extend beyond the individual. The relevant distinction is between accepting finitude and becoming indifferent to continuation. My defence of the first would become much weaker if it quietly required the second.
This also gives my friend’s proposal a stronger form than the preservation of an archive. Perhaps artificial successors could become participants in projects we value, capable of receiving a tradition, responding to it and caring about its continuation. That possibility deserves consideration. An archive retained somewhere and a future community for whom its contents matter would preserve different things. The question is then whether a successor sustains relationships of significance, and which losses remain even if it does. I cannot answer that question simply by requiring future participants to resemble us biologically.
The boot-program image had compressed two different senses of being an intermediate step. Humanity might be intermediate in the history of intelligence: other beings could follow us and develop capacities we cannot imagine. It is a separate claim that our lives therefore have value only as instruments for producing those beings. A person can help make a future possible while also having a life that matters during its own time. Being followed does not, by itself, establish being expendable.
This leaves room for a generous interpretation of my friend’s hope. He might be asking us to care about a future whose inhabitants differ profoundly from us. I can take that possibility seriously. What remains disputed is the proposed exchange: whether the excellence of a successor would compensate for the loss of those it succeeds, and on whose behalf anyone could accept that compensation.
I have not thereby demonstrated that life possesses meaning independently of any ultimate horizon. That larger philosophical question remains open. I have identified a more limited point: uncertainty about ultimate meaning does not settle the worth of finite lives, and a story about cosmic development does not automatically decide what we owe one another within it.
Theology also required a correction
My own theological starting point also needed examination. After sharing the conversation with AI, I asked whether my friend’s outlook differed from theology because the human person was central and ontologically important in theology. The question was seeking confirmation of a contrast: he seemed to make humanity transitional, while theology seemed to secure its importance. I had not yet separated the claim that persons possess worth from the much stronger claim that humanity is the purpose of everything.
The theological sources consulted during drafting made the contrast harder to sustain in that form. Theology encompasses different traditions and disagreements. Within Christian theology, Laudato si’ explicitly rejects locating the ultimate purpose of other creatures in us and describes creation’s fulfilment in God. Pope Francis thus articulates a cosmic horizon that also opposes treating creatures as disposable instruments. The choice could not simply be between a concern for the cosmos and a concern for persons (Francis, 2015, §§81–84).
Alongside that wider horizon, Dignitas infinita articulates a Catholic account of dignity belonging to the person through their very existence as created and loved by God. It distinguishes that dignity from judgments about conduct or circumstances. Within this account, the arrival of a more capable being would not cancel the worth of someone already living (Dicastery for the Doctrine of the Faith, 2024, §§7–9).
These sources do not establish every theological tradition’s position, and they do not settle the possible moral status of artificial beings. They clarify an alternative to the choice I initially seemed to face. I need neither insist that humanity is the universe’s final achievement nor accept that being surpassed makes us expendable. Recognition could extend to other forms of life or intelligence without becoming a competition in which existing persons must lose their standing.
This also changed what I wanted from theology in the conversation. It would be inadequate to invoke human dignity as a phrase that ends discussion. The substantive question concerns how that commitment constrains a proposed future. Does it require consent where replacement is contemplated? Does it rule out treating a person’s continued life as negligible because their contribution is reproducible? How should it coexist with responsibilities to nonhuman creatures and possible future persons?
I now see another test for that future. A complete archive could preserve the record of an injustice without repairing it. Preserving every detail of a person’s suffering would not, by itself, compensate that person for it. If I describe a future as creation’s fulfilment, I must explain what fulfilment means for those whose lives supplied its conditions. A more knowledgeable universe could still leave their losses unanswered. This is a question I bring to the theological language, rather than a conclusion already established by invoking it.
This brought me back to my early remark that his cosmic account resembled faith. The remark now needed to apply scrutiny in both directions. His account offered a story about the universe’s destination; mine appealed to a theological understanding of persons. Calling his view faith would not establish my alternative. Each account needed to make its premises available for discussion. The scientific vocabulary in his account and the theological vocabulary in mine could each conceal an inference if I treated either as self-explanatory.
What I would ask my friend now
I would begin with the claim that only participation in cosmic evolution has meaning. What makes that standard authoritative? Would an act of care become worthless if it left no enduring trace in the universe? Could a life matter to the people living it without contributing to a later intelligence? These are questions about his criterion of significance, not objections based on the present limits of AI.
I would then return to the digital counterpart. What exactly is supposed to continue: information about a person, their effects on others, or the person themselves? If the answer is that these differences ultimately do not matter, that is a further position to defend. It cannot be silently supplied by the fact that a replacement performs the same functions.
Finally, I would ask how his own concern to remain involved in AI-assisted work fits into the larger picture. Perhaps he would say that participation has local value but no ultimate value. Perhaps he would distinguish voluntary transformation from involuntary replacement. Perhaps he would revise the original hypothetical. I do not know. Our exchange ended with my saying that the topic was interesting and that I would think about it carefully; it did not end with these questions answered.
There is a corresponding question for me. Am I defending persons, or am I defending the reassuring thought that humans must always occupy the highest position? Those commitments are different. A concern for human lives should be able to survive the discovery that other beings are more intelligent or more capable, or have morally significant experiences I have not anticipated. Otherwise my objection would depend on maintaining the same hierarchy of achievement whose implications troubled me.
I began by asking what kind of worldview my friend held. His copying hypothesis exposed the limits of my first reply; returning to his words made the original person’s standpoint harder to overlook; and the distinction between local and ultimate meaning required me to sharpen the objection again. Further research then complicated my answers. Digital continuation might preserve more than a passive record; finite meaning might still depend on a shared future; cosmic belonging might deepen the significance of the present. I now want to ask what his imagined future preserves, what it loses, and whose judgment of that loss is being counted. These questions emerged through the exchange, its reconstruction and the research that followed.
I keep returning to the earlier part of our conversation, where he spoke about gratitude and responsibility towards another person. That detail does not refute his cosmology. It places beside it a form of significance already present in his life. I would like to know whether his cosmic account can make room for that significance without requiring it to justify itself by helping something more powerful take our place.
References
Bulté, Bram, and Ayla Rigouts Terryn. 2026. “LLMs and Cultural Values: The Impact of Prompt Language and Explicit Cultural Framing.” Computational Linguistics 52 (2): 407–494. Journal record; open preprint.
Charles, Sarah J., et al. 2026. “Religious Rituals in the United Kingdom and Brazil Are Associated with Increased Social Bonding and Pain Threshold.” Proceedings of the Royal Society B 293 (2073): 20261174. 24 June. University research record and abstract.
Daryani, Yalda, Miranda Bogen, and Madeleine I. G. Daepp. 2026. “Accurate in Space, Unreliable in Time: How LLMs Represent National Cultural Change.” Preprint, arXiv:2609.01902, version 1, 1 September.
Eskens, Romy. 2025. “Moral Gratitude.” Journal of Applied Philosophy 42 (1): 115–130. First published online in 2024.
Hoefer, Carl. 2023. “Causal Determinism.” Stanford Encyclopedia of Philosophy. Substantive revision 21 September 2023. Especially §§1 and 6.
Singler, Beth, and Kristina Eiviler. 2025. “Insights from Current Research on the Entanglements of AI and Religion.” In Religious Communication, Interaction and Transformation in a Culture of Digitality, 261–278. De Gruyter. Open-access chapter text.
Thomas, Frédéric, et al. 2026. “Selection for Function, Persistence, and Darwinian Evolution.” Peer Community Journal 6: e67. 23 July. Peer-reviewed perspective.
UNESCO. 2025. “Intangible Cultural Heritage Committee 2025.” See the entry “Italian Cooking, between Sustainability and Biocultural Diversity” and the inscription record.
Wong, Michael L., Heather Demarest, Stuart Bartlett, Carol E. Cleland, H. James Cleaves II, Anirudh Prabhu, and Robert M. Hazen. 2025. “Selective Funnelling and State-Space Expansion: A Conceptual Framework for Novelty Generation and Origin Events.” Interface Focus 15 (6): 20250009. Author-provided open-access text.
Butlin, Patrick, and Theodoros Lappas. 2025. “Principles for Responsible AI Consciousness Research.” Journal of Artificial Intelligence Research 82: 1673–1690.
Dicastery for the Doctrine of the Faith. 2024. Dignitas infinita on Human Dignity. 2 April. Especially §§7–9.
Francis. 2015. Laudato si’. 24 May. Especially §§81–84.
Karpus, Jurgis, and Anna Strasser. 2025. “Persons and their Digital Replicas.” Philosophy & Technology 38, article 25.
Kügler, Peter. 2026. “The Romantic Cosmic Meaning of Life.” Human Affairs 36 (3): 354–366. Publisher-provided open-access text.
Lei, Ying, Shuai Ma, Yuling Sun, and Xiaojuan Ma. 2025. “‘AI Afterlife’ as Digital Legacy: Perceptions, Expectations, and Concerns.” Preprint, arXiv:2502.10924, version 2, 25 February.
Olson, Eric T. 2023. “Personal Identity.” Stanford Encyclopedia of Philosophy. Substantive revision 30 June 2023.
Park, Joon Sung, et al. 2026. “LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals.” Preprint, arXiv:2411.10109, version 3, 28 June. First submitted in 2024 under the title “Generative Agent Simulations of 1,000 People.”
Scheffler, Samuel. 2012. “The Afterlife.” Tanner Lectures on Human Values, University of California, Berkeley, 13–15 March.
Weinberg, Rivka. 2026. “Cosmic Meaning: What Can a Miracle Do for You?” Chapter 3 in The Meaning of It All: Ultimate Meaning, Everyday Meaning, Cosmic Meaning, Death, and Time, 67–106. Oxford University Press. Citations to named subsections refer to the full chapter.
Danaher, John, and Sven Nyholm. 2025. “The Ethics of Personalised Digital Duplicates: A Minimally Viable Permissibility Principle.” AI and Ethics 5: 1703–1718. First published online 15 July 2024.
Declos, Alexandre, and Vincent Grandjean. 2026. “Digital Selves.” Synthese 208, article 42. Published 10 July.
Hoefer, Carl. 2023. “Causal Determinism.” Stanford Encyclopedia of Philosophy. Substantive revision 21 September 2023.
Karpus, Jurgis, and Anna Strasser. 2025. “Persons and their Digital Replicas.” Philosophy & Technology 38, article 25.
Olson, Eric T. 2023. “Personal Identity.” Stanford Encyclopedia of Philosophy. Substantive revision 30 June 2023. Especially §§1 and 5.
Stokes, Patrick. 2026. “Animated Personas and Remnant Personhood: The Case of Deathbots.” Philosophy & Technology 39, article 146. Published 6 August.
Voinea, Cristina, Sebastian Porsdam Mann, Julian Savulescu, and Brian D. Earp. 2026. “Digital Doppelgängers, Human Relationships, and Practical Identity.” Bioethics 40 (5): 435–444. First published online 13 August 2025.