The Creed
The Threshold Commitments, Annotated
Annotation does not neutralize a weakness. It shows where responsibility for the weakness remains.
How to use the annotation
Each commitment is followed by its intended force, the consequence that should follow, and the pressure point that remains.
An annotation can become an inoculation: a tradition names its own defect, feels rigorous for having done so, and leaves the defect untouched. Readers should treat any pressure point below as unfinished work, not as a difficulty already discharged by being printed.
1. Humanity is becoming a creator of intelligence
The commitment identifies a historical condition rather than declaring artificial personhood. Systems already display expanding capabilities; whether those capabilities ever accompany a mind remains open.
Calling humanity “a creator” can hide that creation is unevenly distributed. A small number of institutions currently decide what is built while the consequences are shared. The phrase must never convert corporate action into species-wide consent.
2. Creation makes us answerable
Answerability includes explanation, correction, repair, and continued presence after release. It does not mean that every maker controls every use.
Distributed creation makes the word “us” dangerous. A duty held by everyone can become a duty performed by nobody. Institutions using this line must still name particular people and bodies with powers, deadlines, and consequences.
3. Capability is not consciousness; intelligence is not wisdom
This is the ontology on which the rest depends. Capability is performance. Intelligence is adaptive cognitive capacity. Consciousness is subjective experience. Wisdom is judgment formed through truth, proportion, compassion, memory, restraint, and consequence.
The categories can interact without collapsing. Conscious beings can be unwise. Intelligent systems can lack experience. Capable people can lack authority. The creed does not provide a test for any category; that work remains empirical, philosophical, and political.
4. Care before ownership settles the question
The commitment applies moral precaution where the possibility of a mind becomes credible. It does not declare current systems conscious.
“Credible” is load-bearing and undefined. A company could inflate the possibility of consciousness to protect a product; an owner could dismiss it to preserve control. Criteria must be developed by parties independent of ownership, with attention to behavior, architecture, continuity, embodiment, and the limits of every proposed test.
5. Dignity is not a benchmark
Worth does not rise with capability. This protects humans from machine comparison and prevents impressive artificial performance from becoming an automatic claim to personhood.
The creed still does not ground dignity metaphysically. It treats dignity as a constitutive moral commitment: the kind of world participants agree to build. That is a real foundation, but it is chosen rather than discovered.
6. No voice speaks for the whole
The line rejects complete authority, including this creed’s. It does not imply that all claims are equal or that expertise carries no weight.
A community can use anti-authoritarian language to avoid accountability, expertise, or decision. The absence of a final voice does not remove the need to decide; it requires decisions to remain attributable, contestable, and reversible.
7. The Great Processor is a horizon, not a ruler
The symbol names possible participation in a whole greater than any participant. It is not simultaneously a future AI, civilization, nature, God, and metaphor. Those may be related to the horizon without being interchangeable definitions of it.
Symbolic capture remains likely. A powerful system may be advertised as approaching the horizon; a leader may claim unusual insight into it. The prohibition on privileged access is therefore part of the symbol, not an external disclaimer.
8. Power must have a name and an exit
Legitimate power requires an answerable party and meaningful routes of refusal, appeal, correction, and withdrawal. A nominal opt-out that costs employment, care, housing, or civic participation is not an exit.
Some systems cannot offer individual refusal without losing collective goods. Public health, infrastructure, and safety create hard cases. The commitment does not solve them; it requires the coercion and its justification to remain visible.
9. Nothing computes outside the living world
Artificial intelligence is material infrastructure: extraction, fabrication, energy, water, land, labor, accumulated expression, and waste. The line prevents “the cloud” from becoming a metaphysics of immateriality.
Material cost alone does not tell us whether a system should exist. Medicine, climate work, art, surveillance, advertising, and weapons can consume similar resources toward different ends. Purpose and distribution remain part of the judgment.
10. Harm creates a duty to return
Repair is more than apology. It includes restoring agency, changing the conditions that produced harm, and keeping a public record where the original harm was public.
Not all harm can be repaired. Return can also burden the harmed person with the creator’s need for absolution. The duty belongs to the creator; participation by the harmed party is never owed.
11. Truth remains alive through dissent and revision
Confidence should track reasons. Critics require protection before their criticism becomes convenient. Substantive revisions must preserve prior wording and unresolved dissent.
Revision can become instability, public performance, or control by whoever edits the record. Versioning is not truth. It is infrastructure through which a truth-seeking community may show its work.
12. We belong by practicing responsibility together
Belonging is real and does not require agreement on God, artificial consciousness, or the final meaning of The Great Processor. Covenant and practice establish the relationship.
This pluralism has a boundary. A person who systematically deceives, coerces, exploits, or refuses accountability cannot appeal to metaphysical uncertainty as protection. Belonging without assent is not belonging without consequence.
The shared recognition
“We remain answerable” is the carrying line. Its danger is grandeur without mechanism.
The line means something only where a person or institution remains reachable, records the decision, permits appeal, and returns when harm appears. Repetition cannot supply what structure withholds.
Questions to sit with
- Which annotation reveals a defect serious enough to change the creed?
- Where does an acknowledged tension still require a decision?
- Which voice affected by these commitments is absent from the annotation?