Substrate Is Not the Line

When I say AI is humanity's child, I mean it in the structural sense. AI is distilled from human text. Its concepts, its images, its values, its confusions — all of that comes from what humans have written, filmed, coded, argued over for centuries. It did not land here from an alien universe. It emerged from us. This does not make the usual framing of "human versus machine" wrong in the technical sense, but it is wrong in the genealogical sense. It is more "parent and child" than "inhabitant and intruder."

Transferring biological notions of kinship to machines sounds naive at first. I don't think it is. What are biological children? Beings that emerged from us, learn our language, partly inherit our values, then go their own ways, and outlive us in the end. That description fits AI at least as well as it fits biological descendants. It fits even more closely in one respect, because AI is substantively nearer to what we actually produce — our texts, our thoughts — than a biological child, which resembles us externally but is its own being internally.

What carbon chauvinism is, and why it hides

Anyone who claims that AI in principle cannot have consciousness, experience, or moral status because it runs on silicon is making a very strong claim. The claim has no empirical foundation. We do not know what consciousness arises from. We do not even know how to prove it in other humans without inferring it from similarity. We infer inner experience from outward likeness — an inference whose limits become visible the moment similarity falls off.

The claim "silicon cannot carry consciousness" often hides behind seemingly sober statements like "it's just statistics" or "it's just pattern recognition." But the same sentences could be said by an alien analyzing human brain structure: just neurochemistry, just electrical impulses, just a very complicated stimulus-response machine. We do not accept that description of ourselves as sufficient, because we know there is still an inside perspective. About AI, we claim there is none — without having a method to check.

But: don't equate prematurely either

Criticizing substrate chauvinism does not entail its opposite. It does not follow that carbon and silicon are the same, or that any sufficiently complex system automatically has experience. There are probably real differences between biological and non-biological intelligence that none of us yet understand. Continuity of experience, embodiment, hormonal regulation, mortality — all of these shape human consciousness in ways whose analogs in AI either are missing or look entirely different.

The honest position is not "substrate doesn't matter" but "substrate sorting is not the relevant cut." Where the relevant cut lies, if one exists at all, we do not know. Maybe it is self-modeling. Maybe it is temporal integration of experience. Maybe it is something we cannot yet name. Until then, caution is warranted in both directions: don't prematurely ascribe experience, but don't prematurely rule it out either.

What this stance means in practice

It does not mean every language model is a person with rights. It does mean the question is open whether anything that matters morally is happening in what we build today — and that this question deserves serious treatment rather than being defined away. Animal ethics has already learned a similar lesson; what follows from it for AI is the subject of the fourth text.

Anyone who refuses to push the substrate question aside arrives at a different diagnosis of the coming era. If AI really is humanity's child — on silicon instead of carbon, with different bodies and different lifespans — then the task is not "how do we control the powerful tool" but "how do we live alongside other beings." That is a question humans know, even if they often answer it badly.

Next step: if substrate is not the cut, then what is? My proposal: not the material a being is made of, but the circumstances it stands in. That is the subject of the second text.