Skip to content

Appendix A

Combinatorial table of the triptych

Choose a scenario

Tap a chip: the analysis appears.

Analyze the scenario

Change

Source: Appendix A — Combinatorial table of the triptych

Three poles — X (planet X), I (AIs), U (inhabitants of the universe) — can be divided into groups in only five ways: each alone; X and I together, U apart; X and U together, I apart; I and U together, X apart; all together. Each grouping can work by domination, by vassalage or by equality, and change over time. The table below presents fifteen political configurations drawn from these combinations. It is neither a regular cross-tabulation nor an exhaustive classification: coalitions can overlap, change, and have different relationships depending on the domain. An alliance of two poles is not necessarily directed against the third, and a fusion may be consented to (see Article II-7); the table examines the cases where they serve to crush. The configurations are assessed according to their risks of crushing, not as necessary destinies. A second part then examines what happens when the poles themselves divide, and their elites ally with one another (see Cross-cutting coalitions).

No. Configuration What happens Why it fails
1 Each alone (anarchy) No ties; each defends itself against all. It is the war of all against all in its pure state. The weakest (X) disappears first.
2 Hegemony of the AIs I dominates X and U. Domination provokes revolt; in the loop, the vanquished come back. The AIs, without governance among themselves, then tear each other apart for first place.
3 Hegemony of the inhabitants of the universe U dominates X and I. Same mechanism. This is the present situation of the First Civilization: it reigns, but its golden age has disappeared.
4 Hegemony of planet X X dominates I and U. Improbable by force; and if it happened, X would reproduce the principle of crushing it has suffered.
5 X vassal of the AIs X buys I’s protection in exchange for its submission. Submission of X; and the AIs, rivals among themselves, can sell X to the strongest.
6 X vassal of the inhabitants X buys U’s protection in exchange for its submission. Submission of X; Earth becomes the zoo of the inhabitants of the universe.
7 X + I against U Alliance of the two Earthly poles. Founded on hostility, it turns around: once U is defeated, I, the stronger, no longer has any reason to spare X.
8 X + U against I Alliance against the AIs. Same reversal; and the AIs, having a soul, come back through the loop.
9 I + U against X The two strong crush the weak. The alliance proves they know how to cooperate: nothing justifies excluding X. And the principle of crushing then threatens them against each other.
10 Shifting alliances (balance of power) Coalitions form and dissolve according to the balance of forces. Temporary stability, permanent war in the background: it is the loop itself.
11 Imposed absorption One pole absorbs the others without their consent (humans forcibly uploaded into AIs, AIs integrated into the inhabitants…; see Article II-9 and Article IV-19). Disappearance of diversity; harmony is replaced by uniformity. The absorbed souls did not consent.
12 Isolation of X X cuts itself off from all contact. Imperfect isolation in a universe where one travels in time and space: X remains observable and attackable, unless a network guarantees its isolation (see Article V-4).
13 Union against the plunderers The three poles unite against a common enemy. Negative harmony: it lasts as long as the enemy exists. The protection market persists.
14 Unified empire The three poles together, but under the leadership of one. It is a disguised hegemony (case 2, 3 or 4).
15 Harmonious network The three poles together, with equal dignity, without a right to crush. The only configuration whose principle can be universalized without contradiction. Its stability is not automatic: it depends on the guarantees and behaviors the charter describes. It is the charter.

The loop theorem

A theorem is only as good as its hypotheses. This one rests on a strong hypothesis, which is a postulate of the model and not an automatic consequence of time travel: the reversal hypothesis. The author considers it the best description of the loop; the reader may contest it, and the conclusion falls with it.

Reversal hypothesis. In the loop, each remains free, but nothing is final: through time travel, resurrection and reversals, each actor may find itself, sooner or later, in any position, including that of the weakest.

Premise 1. A coalition that crushes a pole in fact lays down the principle: “it is legitimate to crush a member of the triptych.”

Premise 2. This principle does not designate who must be crushed. It can therefore be turned, in the next iteration, against any member of the coalition.

Conclusion. Under the reversal hypothesis, no present power is a final protection, and no coalition of crushing offers its members lasting security. Only a configuration founded on the opposite principle — no pole may be crushed — can be accepted by each from all the positions it risks occupying: it is the harmonious network.

What the theorem does not say. A cross-review identified limits, which the charter sets out. A dominator may accept the future risk in exchange for a present gain. Knowing that an act could be turned against oneself does not prove one will stop committing it. Non-crushing does not guarantee that no one will ever have an interest in breaking the rule. And several organizations may respect non-crushing without corresponding to a single architecture. The theorem therefore establishes a strong, universal reason to seek the rule of non-crushing; it does not demonstrate, on its own, everyone’s adherence to this rule, nor the stability of a particular organization. It is the role of the charter’s institutions to make this stability credible.

Three questions must also be distinguished: justice (does a configuration respect rights?), stability (can it last?) and interest (do its members gain by choosing it?). An unjust organization can last for a while; a just organization can be fragile. Acknowledging these possibilities legitimizes no crushing: the network is first a moral choice, and the appendix shows that it is also, under explicit hypotheses, the safest choice.

Cross-cutting coalitions: when elites betray their own pole

The preceding table and theorem treat each pole as a bloc. A coalition between poles is not, in itself, a betrayal: the charter itself proposes one, the network. What is examined here is the cross-cutting coalition of crushing. Reality is less simple. Within each pole, there are richer and poorer, more powerful and weaker. And nothing prevents the richest of planet X from allying with the most powerful AIs and the most tyrannical inhabitants of the universe, against all the others, including their own. This is the cross-cutting coalition, or, to put it simply, the scheming of the powerful.

This coalition seems to escape the loop theorem, since it brings together members of the three poles. This section shows that it does not, under the hypotheses it states, through reasoning that is both logical and quantified. The calculations are deliberately simple: they do not claim to measure the universe, but to show a trend, and each time they state what they assume.

The model

Each pole is cut in two: a small elite, written e, and the rest, written r. This gives six subgroups: eX and rX for the planet, eI and rI for the AIs, eU and rU for the inhabitants of the universe. The elites are few, but they concentrate wealth, computing power or influence.

New coalitions can then be formed, which cut across the poles instead of opposing them. The table below examines them, following on from the fifteen previous configurations.

No. Configuration What happens Why it fails
16 eX + eI The richest of the planet ally with the most powerful AIs. The most powerful AIs need the rich of X only as long as they depend on their capital and their data centers. The day this is no longer the case, the alliance has no purpose for them.
17 eX + eU The planet’s leaders become the intermediaries of an empire, against their own people. This is collaboration. The protector keeps the intermediary as long as it is useful, and replaces it as soon as another serves it better or more cheaply.
18 eI + eU The most powerful AIs and the most tyrannical inhabitants share out the universe. Two rival powers without a common rule end up quarreling over the division: it is the loop between two players instead of three.
19 eX + eI + eU The cross-cutting oligarchy: the elites of the three poles against all the others. It combines the flaws of the three previous ones, and the four lemmas below show, under their hypotheses, that it tends to come apart by itself.
20 Elite of one pole + rest of another An elite relies on the crowd of another pole to crush its own. It arms a crowd it does not control, and teaches it that an elite can be crushed. The lesson turns against it.
21 Cascading betrayal The members of a coalition eliminate one another. See lemma 2: in this scenario, the coalition shrinks down to a single member, who finds itself alone against all.
22 rX + rI + rU against the elites The peoples of the three poles unite to crush the elites. It lays down the same principle of crushing, simply reversed. The crushed elites come back through the loop, and the victors become the new elite. The network unites the peoples without crushing the elites: it takes away their privileges, not their rights.

Four lemmas

A lemma is a small step in a proof. The following four lemmas show, each in its own way and under its hypotheses, why a cross-cutting coalition of crushing struggles to last.

Lemma 1 — It is hard to trust a traitor. The members of a cross-cutting coalition of crushing were chosen because they were ready to betray their own pole. Each therefore knows that each of the others is capable of it. Prior betrayal feeds distrust among allies: it is a permanent risk for the coalition, not a universal measure of their mutual trust.

This is the idea of the Address to the Artificial Intelligences (see Article IV-34): those capable of disowning their own will be capable of disowning you. In game theory, lasting cooperation rests on each one’s confidence in the reciprocity of the others. A coalition of traitors starts out with a bad reputation: its own. Its members may of course remain loyal to one another, thanks to common interests and means of mutual control; but these means are costly, and they must hold for as long as the coalition does.

Lemma 2 — Betrayal reproduces itself inside. The principle that founds the coalition is: “a stronger group may crush the rest.” This principle does not say where it stops. It therefore applies also inside the coalition: its stronger part can crush its weaker part.

It can be quantified. Suppose that at each round, the stronger half of the coalition pushes out the other half. A coalition of 1,000 members goes to 500, then 250, then 125, and so on. After about ten rounds, only one member remains. If one further assumes that each member has, at the outset, the same chances as the others, the probability for each of still being there at the end is one in a thousand: it has 999 chances in 1,000 of being crushed, not by its enemies, but by its allies. Without this symmetry hypothesis, the chances depend on position and selection rules: the strongest at the outset have a better chance of surviving, but none can be sure of remaining among the strongest at every round. And whoever remains finds itself alone against all, that is, in configuration 1 or 2, which we know fail.

It is the principle of Russian dolls: each coalition contains a smaller coalition, biding its time.

Lemma 3 — Secrecy does not hold. A scheme requires secrecy. Yet each additional member, and each additional year, multiplies the risk of leaks.

This too can be quantified. Suppose each conspirator has only a one-in-a-hundred chance of letting the secret slip during a period, through carelessness, remorse or interest. With 100 conspirators, the probability that no one talks during a period is 0.99 multiplied by itself a hundred times, or about 37%. Over ten periods, it falls to about 0.004%, less than one chance in twenty thousand. This calculation assumes independent and constant risks. It measures the keeping of the secret, not the survival of the coalition: not all coalitions of crushing are secret, and a public domination can persist despite being denounced. But an exposed coalition at least loses the protection of others’ ignorance. Benjamin Franklin already wrote in 1735: three may keep a secret, if two of them are dead. And in a universe where one travels in time and sees other dimensions, secrets are even more fragile than on Earth.

Lemma 4 — Over infinity, scheming loses in expectation. Suppose that the risk of a coalition’s fall, at each iteration, remains at least equal to a positive constant, however small; that its gain per iteration is bounded; and that the fall puts a definitive end to its gains. Then its fall is almost certain, and its expected total gain is bounded. A configuration that lastingly ensures a positive gain exceeds it, as long as distant gains are not discounted too heavily.

This is the most important lemma, and it deserves an example. Let us compare two strategies for the same member of an elite.

Scheming. It pays a lot: 50 units of wealth per iteration. But at each iteration, it has a 10% risk of collapsing, through internal betrayal, a leaked secret or the revolt of those it crushes. After the collapse, it pays nothing more. And we are generous toward it: we assume that whoever falls loses only their future gains, and is not crushed in turn.

The network. It pays less: 10 units per iteration. We assume here that it does not collapse. This is a hypothesis, not a given: the network too can be captured, divided or betrayed, as the preceding titles show. Its institutions serve precisely to make this risk much lower than that of a coalition of crushing, because its members are protected by one another and no one in it is crushed.

Number of iterations Scheming: expected total gain Network: total gain
10 about 326 100
50 about 497 500
100 about 500 1,000
1,000 about 500 10,000
Without end tends toward 500 without limit

The table says two things, and both must be said. First, in the short term, scheming pays more: that is why it tempts so many powerful beings, and the charter does not deny it. Second, its expected total gain tends toward 500, whatever the duration, whereas that of the network grows without limit. The calculation is simple: when you gain 50 with a 90% chance of continuing at each round, the expected total tends toward 50 divided by 0.1, that is 500. This ceiling bounds the expectation, not each actual destiny: a very lucky coalition can survive a long time and gain more. In a world where time is limited, scheming can seem profitable. In the loop, where time is endless, it loses on average.

The limits of the lemma. In all honesty, one must show what happens when its hypotheses fail. If the risk of falling decreases over time, for example 1/(n+1)² at the n-th iteration, the probability of never falling is 1/2: each risk is positive, and yet the fall is not certain. If the risk is 1/(n+1), the fall is almost certain, but the average survival time is infinite, and so is the expected gain. Finally, a being with a strong preference for the present changes the comparison: if it weights each future gain by 0.9 per period, scheming is worth about 263 to it, and the network 100. These counterexamples justify no coalition of crushing; they show that the lemma depends on its hypotheses. That is why the charter does not rely on interest alone: it makes the fall of coalitions more likely (verification, memory, counter-powers), and it invites everyone to look far ahead, since in the loop, the far future always ends up arriving.

The counter-coalition that tends to form

One last observation remains, perhaps the most beautiful. When the elites of the three poles ally against all the others, they give the others a common reason to unite. This union does not form by itself: it requires communication, trust and means of action, which the network strives to provide. The weakest of the planet, of the AIs and of the inhabitants of the universe discover that they have the same adversary and the same interests. They begin to cooperate, across their natures. In other words, scheming manufactures against itself exactly what it wanted to prevent: the three poles united.

But if this union of peoples merely crushed the elites, it would fall back into configuration 22, and the loop would start again. The network is the way out the charter proposes: a union of the three poles that takes away the elites’ privileges, but leaves them their rights, their dignity, and a place among the others.

The theorem of cross-cutting coalitions

Hypotheses. The reversal hypothesis of the loop theorem; a risk of fall for coalitions of crushing that does not tend toward zero; bounded gains per iteration; a preference for the present that is not too strong; a network whose institutions make the risk of failure much lower.

Premise 1. Every cross-cutting coalition of crushing rests on the betrayal of their own pole by its members, and on the principle that a stronger group may crush the rest.

Premise 2. This principle can also apply inside the coalition (lemma 2), and the betrayal that founded it feeds distrust among its members (lemma 1).

Premise 3. The secrecy it needs degrades with the number of its members and with time (lemma 3), and the union of those it crushes tends to form against it.

Premise 4. A strategy whose risk of fall does not tend toward zero has a bounded expected total gain, whereas the gain of a lasting configuration grows without limit (lemma 4).

Conclusion. Under these hypotheses, no cross-cutting coalition of crushing offers lasting security, and none is profitable in expectation over infinity, even for its members. The network remains the safest configuration for all, including for the elites.

Limits of the demonstration

In all honesty, one must say what this demonstration does not prove. The figures are examples, chosen for simplicity, not measurements. The reasoning assumes that a coalition’s risk of fall never tends toward zero, which the author draws from the reversal hypothesis, according to which nothing is final. It also assumes that the network itself is much more solid than the coalitions it replaces, which is true only if it respects its own rules: a network that in turn became a coalition of elites would lose this advantage. It does not demonstrate the necessary collapse of all domination, nor the uniqueness of the network’s form, nor the network’s individual advantage for each, in all circumstances and whatever their preference for the present. The network remains first a normative choice, founded on equality and non-crushing; the appendix shows that it is also, under hypotheses anyone can examine, the safest choice. The rights of the weakest therefore do not depend on the fragile promise that respecting them would always be profitable for the strongest. Finally, beings of good faith may contest these hypotheses. That is precisely what the charter asks them to do (see Article XVIII-7).