The Missing Semantic Layer for World Models — and the Denkraum of Bidirektionale Entitätsübergreifende Semantische Emergenz (NEOMANITAI Whitepaper 2026, v2.0)

Ehstand, Andreas

doi:10.5281/zenodo.22256965

DE EN

Concept Registry · Priority & canonical termsPrioritäts- und Begriffs-Kanon

The Missing Semantic Layer for World Models — and the Denkraum of Bidirektionale Entitätsübergreifende Semantische Emergenz (NEOMANITAI Whitepaper 2026, v2.0)

Wie können Menschen verstehen und mitgestalten, wie eine KI ihre Welt beschreibt? Dieses öffentliche Arbeitspapier untersucht eine gemeinsame Bedeutungsebene für Menschen und Weltmodelle.How could people understand and help shape the way AI describes its world? This public working paper explores shared meaning between people and world models.

Andreas Ehstand, Independent Researcherunabhängiger Forscher · ORCID 0009-0006-3773-7796 · 2026-05-18

Abstract.Zusammenfassung (auf Englisch). World models remain largely unidirectional, opaque and slowly updated. This whitepaper, dated 18 May 2026, describes the missing element as a shared, precisely named, bidirectionally operable and error-correcting semantic layer through which world models could become living, co-evolutionary systems - anchored in philosophy of language, embodied cognition and systems theory, with the Two-Frontiers Grounding Model and a documented observable seed kept strictly separate from a dated wager.

Download PDFPDF herunterladen (englisch) · Archived open-access recordArchivierter, frei zugänglicher Eintrag

Prepared with AI assistance; content reviewed and approved by the author. Research working paper. No legal, medical, financial or other advice.Mit KI-Unterstützung erstellt; Inhalt vom Autor geprüft und freigegeben. Forschungs-Arbeitspapier. Keine rechtliche, medizinische, finanzielle oder sonstige Beratung.

Quick answers

What is this page about?

World models remain largely unidirectional, opaque and slowly updated. This whitepaper, dated 18 May 2026, describes the missing element as a shared, precisely named, bidirectionally operable and error-correcting semantic layer through which world models could become living, co-evolutionary systems