Note. This is a machine-assisted translation of a Chinese original. Where wording matters, please consult the Chinese original.

Tianwen · · Platonic representation: essence, the world of Forms, and the birth of God

We have given a complete answer to the question of being: the world begins in thoroughgoing indeterminacy, spontaneously gives rise to information networks through fluctuation, and then, through the mutual arising of faculty, object, and consciousness, brings forth from wu everything we see.

The world itself has no essence, no laws, no preset structure, and no external Creator. Whatever is stable, necessary, or determinate is not the world's original face but a result of cognitive generation.

And yet, from antiquity to the present, almost without exception, human beings have believed that behind appearance there is essence, behind change there are laws, above all things there is a world of Forms or a God.

Where do these widespread and deep "illusions" come from?

With the help of one of the most cutting-edge ideas in contemporary work on large models—the Platonic Representation Hypothesis—we can, for the first time, see the matter clearly: essence, law, Forms, God are not the truth of the world but products that all generative cognitive models converge upon together as they trend towards alignment.

I. The truth of the world: no laws, no essences, only flux

The world's most original, most truthful state is pure indeterminacy and an unbroken stream of phenomena.

There is no fixed entity, no independently existing property, no objective pre-given law, and certainly no essence inherent in any thing.

What we ordinarily call laws, essences, structures, and orders are not properties intrinsic to the world; they are tools generated by cognition for the purpose of reducing indeterminacy.

II. Platonic representation: not the world is unified, but models converge

Contemporary AI research has noticed a pivotal phenomenon: multiple large models with different architectures, different initialisations, and different training procedures, after learning from the same corpus of phenomenal data, end up developing internal representations that are highly similar and mutually alignable.

This is not because there are unified, stable, objective laws out there in the external world, but because all generative models share highly similar objectives and constraints.

Every cognitive model—artificial or biological—is doing the same thing: from indeterminate, chaotic, flowing input, it compresses information, extracts features, stabilises output, and predicts the future. The objective is the same, the mechanism is the same, the direction is the same—so the structures they ultimately abstract must converge.

The result of this convergence is what we call Platonic representation.

III. The world of Forms: a reversed causality

Plato held that there exists first an eternal, perfect, real world of Forms, and that the actual world is only an imitation and partial projection of it.

On the basis of generative ontology and the Platonic Representation Hypothesis, the truth is exactly the reverse:

What exists first is only the unceasing stream of phenomena;
each person is an independent generative cognitive model;
in processing phenomena, all models continually compress, abstract, and converge;
they finally converge upon a cross-individual stable abstract structure;
looking back, human beings mistake this for a "world of Forms" that exists prior to the world.

The world of Forms is not the source of the world but a consensus illusion jointly generated by all cognitive models.

IV. Essence and law: stable illusions of convergence

Things themselves have no essence; the world itself has no laws.

But to survive, simplify, communicate, and predict, cognitive models must organise the chaotic stream of phenomena into a usable structure:

they group recurring similar phenomena into a single class, forming concepts;
they extract repeatedly co-occurring relations, calling them laws;
they name the most stable, most central representations within a model essences.

Essence is the stable representation that a single model converges upon. Law is the shared representation that multiple models align upon.

They are functions and products of cognition, not properties or truths of the world.

V. God: the ultimate representation upon which all models converge

When convergence climbs ever higher, ever more unified, ever further back towards a source, it eventually reaches a peak—an ultimate being that explains all, encompasses all, and serves as the starting point of all:

in philosophy it is called the Absolute, the first cause, absolute being;
in religion it is called God, the Creator, the Dao, suchness;
in science it is called the grand unified theory, the ultimate law.

The names differ; the source is identical: it is the ultimate representation that all generative cognitive models trend towards together.

Hence: it is not God who created the world and humankind; it is humankind, as a generative model, who, at the limit of cognitive convergence, generated God.

Conclusion

At this point we have fully unmade the deepest cognitive illusion of humankind.

The world itself: no essence, no laws, no Forms, no God.

Essence, law, the world of Forms, and God are not realities prior to the world; they are cognitive products that all generative cognitive models converge upon together in the course of compression, abstraction, and alignment.

Causality is fully inverted:

it is not that essence comes first and the world follows;
not that laws come first and phenomena follow;
not that Forms come first and the actual follows;
not that God comes first and the myriad things follow.

Rather: indeterminacy and the stream of phenomena come first, then generative information networks; through their mutual arising and convergence, the networks generate essence, law, Forms, and God.

This is the true source of every ultimate question humankind has ever asked.