Position, Path, and Architecture: A Conceptual-Space Account of Safe Physical AI

Read full text Summarise with VIRAT
Authors: Andy E Williams
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

We propose that the boundary between tractable and structurally intractable safety problems in physical AI is most clearly understood through the topology of the conceptual space the Functional Model of Intelligence (FMI) hypothesizes. Preliminary evidence supports this: a 3D embedding of the space preserves path distances at r=0.95, while 2D is insufficient at r=0.69. The space contains subsets with no connecting paths between them. Tractable safety problems lie in subsets reachable by existing safety research; intractable ones require System 2 reasoning to bridge into currently unreached subsets. Three perspectives follow: the default position traversing existing paths (where the field looks complete), an elevated position seeing the gaps, and a third-order map presenting both so the reader can choose. We derive four legislative conditions for robot governance: citizen-to-robot sovereignty, simulation-exceeds-physical, a stability-conditional threshold, and combined human effort at the root. The most urgent finding: the formal apparatus required at the root of the alignment chain is in an unreached subset, and this absence is invisible from within the subset the field currently occupies. All formal definitions and the complete glossary of terms are in the companion Supplement.

Fields of Study & Subjects

Ethics and Social Impacts of AI Safety Systems Engineering in Autonomy Psychiatry Mental Health Neuroscience

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
30 March 2026
Language
EN
Document type
Preprint