Philosophers and readers of civilizational limits
Philosopher Path
Why value learning is not hopeless-in-principle, where bearer and correction questions bite, and why alignment is partly an institutional selection problem.
Chapter toy · start here for concreteness
Value Bundle Simulator
An interactive sketch of how environment and social structure shape a value bundle — the book's bundle geometry without reading all of Part IV first.
What this path is for
You want to know whether the book earns its metaphysics and its pessimism/optimism balance — not whether a particular audit API is implementable.
You will leave knowing:
- Why the book rejects fixed utility functions and one-shot training as the alignment object.
- How value-bundle geometry responds to the worry that human values are too arbitrary to learn.
- Where bearer maps and ontology shift break naive value identification.
- Why correction is about preserving a human value-update process, not freezing current endorsements.
- Why Claim 6 (basin / selection) makes alignment partly institutional — and why that is a feature, not a dodge.
Time: roughly 3–5 hours following the ordered list above.
PDF companion (deeper read)
| Step | Why | |
|---|---|---|
| 1 | Executive Overview | Scope, rejected simplifications, claim strength |
| 2 | Introduction (six claims) | Full argument skeleton — grounding and basin claims |
| 3 | Part I (Ch. 1–5) | Wrong object, dynamical guarantee, scope |
| 4 | Part IV–V selections (Ch. 15–19, 20–23) | Value bundles, bearers, transport |
| 5 | Part X (Ch. 41–48) | Civilizational limits and synthesis |
Related reading
Peter Kuhn, Humanism (the·anti·completionist, Substack, 2025) — pairs naturally with the basin claim (bibliography key kuhn2025humanism).
Read in this order
Cards, chapter toys, experiment lines, and book chapters — curated for this audience. Experiments are sanity checks only; negative results are part of the reading.
What the Book Is Not Claiming
Framework and conditional spine — not a solved alignment proof.
Value Change vs. Value Corruption
Legitimate moral change vs manipulation — load-bearing distinction.
The Static Target Trap
Why freezing current endorsements is the wrong object.
Why Fixed Values Are the Wrong Target
Why fixed values are the wrong target.
Value-Bundle Transport
Value geometry — tradeoffs that must survive transformation.
Value Bundle Simulator
Environment → bundle mapping as a toy model.
Chapter 17 — LHV Learnability
Learnability vs identifiability split for value structure.
Bearer Persistence
Who counts — referent maps and ontology shift.
Bearer-Map Commutation Failure
When bearer relabeling preserves words but changes import.
What Values Apply To
What values apply to — bearer ontology in depth.
Correction-Channel Integrity
Preserving human value-update, not freezing endorsements.
Correction at Civilizational Scale
Correctability of the aggregate, not just individual intent.
Socio-Technical Attractor Control
Claim 6 — alignment as partly institutional selection.
Institutional Genesis, Memory, and Decay: Historical Case Studies
Institutional histories hub — how correction mechanisms are founded, kept alive, and fail.
Who Still Counts After Transformation
Who still counts after transformation.
Evidence and Uncertainty
Epistemic labels and disconfirming evidence.
Towards Superintelligence Alignment
Closing synthesis and open cruxes.