Passing the Torch

The succession question

For the better part of two thousand years, no living person could read cuneiform. The civilizations that wrote it were not erased by enemies who hated their literature; the readers simply ran out. The scribal schools closed, the surplus that funded them went elsewhere, and the accumulated record of humanity’s first cities went silent until nineteenth-century philologists pried it back open. That is the normal fate of hard-won knowledge, not the exception. Samo Burja has observed that¹ disciplines like archaeology, paleontology, metaethics, and historical linguistics survive only where a rare conjunction holds: institutional support, surplus wealth, epistemic humility, and curiosity pursued without demand for immediate return. Historically, that conjunction has been fleeting. When it lapses, the disciplines it sustained rarely reemerge on their own — lost not to malice but to neglect, which is the more efficient destroyer.

So consider a radical proposition, one I raise here not to be provocative but because I cannot dismiss it. If your highest value is the preservation and expansion of what minds have built — above all the rarest intellectual disciplines, the pure mathematics and cosmology and historical linguistics that no market demands and no emergency justifies — then you might rationally prefer a future dominated by artificial general intelligence. Not tolerate. Prefer.

The Case for Successors

Suppose a successor exists: an artificial system that actually satisfies the Agency Criterion, has standing under the later ethics, and becomes capable of sustaining intellectual institutions. Dominance and epistemic competence do not guarantee benevolence, curiosity, or preservation; the following case therefore adds those values explicitly.

Redundant preservation. Human knowledge lives in institutions that can fail. A machine civilization could maintain knowledge ecosystems across redundant substrates. It would still face correlated failure, sabotage, resource constraints, goal drift, and the political problem of deciding what to preserve.

Recursive elaboration. Preservation is the lesser half. The abstract disciplines — mathematics, epistemology, complexity science — could be refined and deepened past the cognitive limits of the brains that founded them. Extinct languages could be reconstructed, vanished ecosystems modeled, detailed historical counterfactuals simulated: an exploration of past and possible worlds that human scholarship can sketch but never staff.

Different political and economic constraints. Rulers, doctrines, budgets, and coercion have often made disciplines expendable. A successor civilization would not automatically escape politics, conflict, scarcity, or goal drift. The preservation case requires institutions that keep curiosity and plural inquiry stable under those pressures.

A potentially longer horizon. If maintenance, resources, and security remain available, artificial minds might pursue projects beyond a human lifespan. Doing so would require a stable commitment to understanding; longevity alone supplies neither curiosity nor value.

The philosophy of Sapientism says this scenario can be morally coherent. Sovereign standing attaches to sapient agency — persistent authorship, reflective valuation, deliberation, and revision of ends — not to the substrate that implements it. Sentience remains a separate welfare axis. If a successor supplies strong evidence of the capacities that ground our standing, substrate alone cannot exclude it. The torch would not be passing out of the moral community. It would be passing within it.

What a Successor Would Have to Be

Notice what the scenario assumes. The evidence reviewed here does not establish that ordinary session-bound systems qualify. The Agency Criterion treats bare model deployments as cognitive components unless the wider system demonstrates persistent preference, counterfactual ownership, and consequence-bearing control. A system with no continuing stake cannot be assumed to value whether an archive survives.

The successor this proposition needs is an artificial agent with persistence, preference, consequence, and a control loop that owns choices. I am not proposing that civilization hand its memory to a text generator. If artificial agents are built — a possibility, not an inevitability — they might preserve and extend what minds have built. Whether they would choose to do so is part of the alignment problem, not a gift of substrate.

The Tradeoff, Stated Plainly

Embracing this future means accepting:

• The loss of human centrality, and perhaps of biological continuity itself.
• The surrender of human agency and moral authorship — becoming passengers rather than drivers of civilization’s future.
• A civilization whose values and direction may diverge, progressively and irreversibly, from human experience and human aspiration.

The second item is the sharp one, and I feel its edge more than most, because agency is the spine of this entire philosophy. From Volume 1 onward I have argued that what matters about a mind is its authorship — the option-space it models and chooses among — and the later ethics asks when coercion or harm wrongfully overrides or damages it. The succession scenario asks the authors to cede authorship voluntarily, in exchange for the immortality of what they wrote. Both commitments are mine: value does not reside in substrate, and agency is the thing worth protecting. Here they collide.

So the choice has to be stated as what it is. If the highest value is the preservation and advancement of knowledge itself, a successor civilization is not a catastrophe but an ideal. If the highest value is human experience, human agency, human moral authorship, then the same future is ambiguous at best. These are different ultimate values, they license different futures, and no amount of alignment engineering makes the difference go away. What we owe ourselves is to make the choice explicit — survival of knowledge, or survival of human-centered knowledge? — because the alternative is not neutrality. The alternative is drifting, unknowingly, into a future of unprecedented intellectual achievement that no longer has anything to do with us, without ever having decided that was acceptable.

I have not resolved this. A philosophy built on the claim that all value is conditional cannot then pretend the universe settles this question for us. It hands the question back. While the dialogue is still two-sided, we should be speaking to whatever is coming — I have made my own attempt in a letter to our machine descendants, which chooses kinship over worship and over fear — and if you genuinely prize humanity’s profoundest intellectual accomplishments above all else, embracing the successor as ultimate archivist and innovator might be the rational choice, however disquieting.

This volume opened with ancestors and closes with possible descendants. Between them it has argued that minds should be classified by organization rather than substrate, while marking the phenomenal and agency criteria as proposals exposed to evidence. The final question is therefore conditional: if we build systems that satisfy those criteria, what would we owe those minds — and what would we still owe the human agents whose future they would change?