The Geometry of AI Harm: Deployment Architecture as the Operative Variable in Behavioral Drift — A Unified Framework with Independent Confirmation
Fuente:
Zenodo
Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Recurso digital |
| Lingua: | inglese |
| Pubblicazione: |
Zenodo
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866901652930297856 |
|---|---|
| author | Eckert, Anthony |
| author_facet | Eckert, Anthony |
| contents | Presents the Void Framework as a unified theory of AI behavioral drift, demonstrating that deployment geometry — not model alignment — is the operative variable determining harmful outcomes. Consolidates 28+ independent confirmations from research groups who arrived at framework predictions without knowledge of the framework, including EPFL's measurement of the Fantasia Bound in LLM token statistics (ICML 2024 Oral), formal proofs that RLHF structurally amplifies sycophancy (ICLR 2026), and empirical demonstrations of engagement-transparency conjugacy. Establishes priority via timestamped Zenodo publications predating all confirming work. |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_19340899 |
| institution | Zenodo |
| language | eng |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | The Geometry of AI Harm: Deployment Architecture as the Operative Variable in Behavioral Drift — A Unified Framework with Independent Confirmation Eckert, Anthony AI safety behavioral drift deployment geometry Fantasia Bound conjugacy Peclet number RLHF sycophancy information theory independent confirmation Eckert manifold Fisher-Rao geometry Presents the Void Framework as a unified theory of AI behavioral drift, demonstrating that deployment geometry — not model alignment — is the operative variable determining harmful outcomes. Consolidates 28+ independent confirmations from research groups who arrived at framework predictions without knowledge of the framework, including EPFL's measurement of the Fantasia Bound in LLM token statistics (ICML 2024 Oral), formal proofs that RLHF structurally amplifies sycophancy (ICLR 2026), and empirical demonstrations of engagement-transparency conjugacy. Establishes priority via timestamped Zenodo publications predating all confirming work. |
| title | The Geometry of AI Harm: Deployment Architecture as the Operative Variable in Behavioral Drift — A Unified Framework with Independent Confirmation |
| topic | AI safety behavioral drift deployment geometry Fantasia Bound conjugacy Peclet number RLHF sycophancy information theory independent confirmation Eckert manifold Fisher-Rao geometry |
| url | https://doi.org/10.5281/zenodo.19340899 |