{"path":"research/the-chain-inside-a-link/sentient-silicon-pair-through-the-pipeline.md","content":"# The sentient-silicon pair, run through the pipeline by hand (2026-10-18)\n\n**What this is.** A virtual execution: the extraction pipeline's rules (the scout that points,\nthe splitter's separability test, the stranger rewrite, the implicit-premise pass, the contested-words\nlane, scheme classification with its critical questions, and the keeper with the chain built the same\nevening) applied by one reader to the two claims the keeper had called `overlaps` by topic in the\ntwenty-first listening piece. No model was called. The founder asked for it to see *\"how much this\nexpands the chain/tree/connections and clarifies the distance between the two claims.\"* One reader's\nreading; the first person's reading (T8) of a real run is still owed.\n\n**The pair, as stored.**\n- A: *If silicon systems can be sentient, human bodies may soon be superseded by machines.*\n- B: *Future civilisation will run on artificial intelligence.*\n\n## 1. Scout\n\nA is a conditional with a hedged consequent (*may*, *soon*). It asserts the conditional, not the\nantecedent. B is a bare prediction with no stated reason: a conclusion whose premises are unwritten.\nThe scout points at both as single-sentence arguments; neither text gives the other's premises.\n\n## 2. The splitter, and the stranger rewrite\n\nThe separability test (*could a reader affirm one part while denying another?*) on A:\n\n| part | stands alone for a stranger | affirm one, deny another? |\n|---|---|---|\n| A1 | *A silicon-based system can have experiences.* | the antecedent; A does not assert it, so it enters as a conditional's antecedent, not a claim of A's |\n| A2 | *If a silicon-based system can have experiences, machines could take the place of human bodies.* | the core conditional, asserted |\n| A3 | *That replacement, if it comes, comes soon.* | yes: one can affirm replacement and deny *soon* (a degree threshold, Kennedy's axis) |\n\nThe coverage check: A2 and A3 rebuild A; A1 is presupposed, not covered. B:\n\n| part | stands alone | affirm one, deny another? |\n|---|---|---|\n| B1 | *Artificial intelligence will be able to carry out the functions a civilisation runs on.* | yes: capability |\n| B2 | *A future civilisation will in fact rely on artificial intelligence for those functions.* | yes: adoption; one can affirm B1 and deny B2 |\n\nB is two parts, jointly required (the weakest-part default governs the whole). Two claims became five,\nthree of them asserted.\n\n## 3. The contested-words lane\n\nThree words fork, and two of the forks are inside A alone:\n\n- *sentient*: has experiences · has moral status · is self-aware (the text uses the first).\n- *superseded*: bodies replaced as the performers of work (economic) · bodies replaced as the vehicles\n  of persons (substrate). A's author plausibly means the second; the keeper read the first.\n- *run on*: operated by (governed) · built on (a substrate, as a city runs on electricity).\n\nThe lane would mint two senses each. The one word A and B share at the surface is *machines* against\n*artificial intelligence*: a shared concept, which is a `USES_CONCEPT` link to one concept node, and\nno relation between the claims. This is the whole of the resemblance the keeper saw, placed where it\nbelongs.\n\n## 4. The implicit-premise pass (within each text)\n\nA's bridge from antecedent to consequent needs, on the substrate reading:\n- A-ip1: *Persons are not tied to the substrate they run on* (substrate independence).\n- A-ip2: *The means to move a person onto a machine will exist and be permitted.*\n\nOn the economic reading A-ip1 is instead *machines that have experiences would be allowed to do the\nwork bodies do*, a different premise with a different evidence base. B's unwritten premises:\n- B-ip1: *AI capability will reach the level a civilisation's functions require.*\n- B-ip2: *Nothing (law, cost, choice) stops its adoption.*\n\nBoth passes are capped at one or two per argument and told to prefer fewer, so a real run would write\nat most half of these.\n\n## 5. Scheme and critical questions\n\nA2 is an inference from cause to effect (sentience → replacement). Its critical questions, each a\npotential claim with two polarities: *how strong is the generalisation from sentience to\nreplacement?* · *what other factors would have to hold?* · *could the effect occur without the\ncause?* The third is the finding of the run: **replacement of bodies by machines needs no sentience**\n(non-sentient automation does it), so A's antecedent is not on any path to B. B carries no inference,\nso no scheme: it is an unsupported conclusion, a copout marker, until a premise is written under it.\n\n## 6. The keeper, with the chain\n\nForward, from establishing A toward B. Establishing a conditional establishes nothing about its\nconsequent unless the antecedent is established, and neither claim asserts A1. Granting A1 anyway:\n\n1. A1 and A2 → machines could take the place of human bodies (modus ponens).\n2. → the functions those bodies performed are performed by machines (A-ip1, economic reading; on the\n   substrate reading this step does not exist, since uploaded persons still run civilisation).\n3. → those machines are artificial intelligences (a bridge neither text states: a machine that\n   replaces a body need not be an AI).\n4. → the functions replaced include the ones a civilisation runs on (a widening of scope from bodies\n   to civilisation that no sentence licenses).\n5. → B1, AI can carry out those functions.\n6. → B2, civilisation relies on it (B-ip2).\n7. → B.\n\nSeven steps, the first blocked by an unasserted antecedent, the second available on only one of two\nsenses of *superseded*, and steps three and four invented. Steps asserted by either claim: 0 (B1 and\nB2 are B's parts, reached only at the end). Backward, from establishing B toward A: B → AI performs\ncivilisation's functions → bodies are not needed for them → *superseded* in the economic sense: three\nsteps reaching one sense of A's consequent and never A's conditional, since B says nothing about\nsentience.\n\nUnder the keeper as rebuilt tonight: chain of seven, zero asserted, gate `no`; had the model still\nwritten `overlaps`, code refuses it (`OVERLAP_MAX_STEPS = 2`, or zero asserted steps).\n\n## 7. What the pipeline did to the distance\n\n| | before | after the pipeline's own rules |\n|---|---|---|\n| claims | 2 | 5 asserted parts, 1 antecedent, up to 4 unwritten premises |\n| concepts | 0 | 3 words, 6 senses; one concept shared (*machines* / *AI*) |\n| potentials | 0 | 3 critical questions on A2, 6 polarity potentials |\n| claim-to-claim links across the pair | 1 (`overlaps`) | 0 |\n| shortest honest chain | unstated | 7 forward (first step blocked), 3 backward (one sense) |\n\nThe expansion did not bring the claims closer; it moved the one resemblance to the concept layer and\nmade the distance a count. The deeper the floor, the longer the honest chain reads, since each\nseparable part A or B bundled is one more step a link would have to cross, which is the shape the\nfounder predicted: *\"it must be HUGE.\"*\n\n## 7a. What the critical question did, and what it offers the keeper\n\nThe question *could the effect occur without the cause?* belongs to the cause-to-effect scheme, and on this\npair it did what the chain count could not do alone: it struck A's antecedent off every path to B, since\nbodies can be superseded by machines that feel nothing. Read as a method, it is the smallest-step gap\ntest (*describe a situation where every sentence of the step holds and its end fails*) in a scheme's own\nwords, and the pipeline holds a bank of such questions for 44 schemes that the keeper has never read.\nA keeper that, having written its chain, ran the critical questions of the scheme each step belongs to\nwould find the gaps the count alone misses. Proposed in ledger AM1's amendment; the founder has not\nabsorbed it yet, in his words, and it is unbuilt.\n\n## 8. Limits\n\nOne reader, no model run, no person's reading of a real pass output. A real run would likely write\nfewer implicit premises (the pass prefers fewer), might not cut *soon* from A, and the concept lane\nhas been measured to mint two senses by template. The count of seven is a lower bound by the method\nand an estimate by the reader.\n\nRecord: `docs/research/three-minds-and-the-leap.md` § 4a; the keeper's header in\n`deliberus/keeper.py`; ledger AM1, AM2.\n\n## 9. The cut, run on the live graph (2026-10-19)\n\nAfter AM5 and AM6 were ruled (*\"Build AM5+6.\"*), the reasons pass read the stored sentence A\n(`claim_c514b82ad6cb`, the extraction's wording: *If silicon-based systems can be sentient, then\nhuman brains and bodies may soon be superseded (replaced) by machines*) as a conditional, on the\nClaude engine, two calls. It wrote the consequent *Human brains and bodies will soon be superseded by\nmachines* (`claim_8fdd02f35635`) and the antecedent *Silicon-based systems can be sentient*\n(`claim_e5df0ec60e52`), both of kind `supposed`, asserted by nobody; a support link from the\nantecedent to the consequent carrying `hedge: may`; and the sentence itself as that link's\nrelevance claim. The hedge read *ambiguous*: both *I think* and *it is possible for* survive in\nfront of the consequent, which is the case his revisit is about. *Soon* stayed in the consequent,\nas § 8 predicted a real run might leave it.\n\nThe first attempt was refused by its own check: the recomposition had spliced the hedge back in\nfront of a consequent that already read *will*, and the checker called the double modal an\ninvention. The check now names the hedge beside the link instead of splicing it, and the second\nattempt passed clean. A check that catches its own pass's slip on the first real sentence is the\ncheck working.\n\nThe keeper then read each piece against B (`claim_26a7be0866fb`), four calls, the brake's cap.\nThe antecedent: no bearing, nothing written, which is § 7's finding (sentience is on no path to a\ncivilisation run on AI). The consequent: bearing yes, *overlaps*, not confident, on a two-step chain\nof which neither step is asserted by either claim (*superseding machines would be the operative\nsubstrate running society* · *a society run on that substrate is functionally powered by AI*); the\ngate refuses an overlap on a chain with no asserted step, so no link was written and the record\nholds the chain. That is the distance § 7 measured, now written down by the pass that\nwould have linked across it.\n"}