Work

SOPH-IA v2.0: Sub-Optimal Paradigm for Habitable AI and the Thermodynamics of Algorithmic Ethical Friction

Year: 2026

Type: preprint

Abstract: Current AI alignment optimizes for perfect obedience, causing energy waste and "paperclip" failures. SOPH-IA v2.0 introduces Ethical Friction as a measurable thermodynamic cost of alignment. We define Thermodynamic Ethical Friction (TEF) and prove that optimal systems must be sub-optimal by design (Satisficing). Our benchmark T4 shows a 62% reduction in inference cost for a 2% drop in task performance, achieving Habitability 6/7 vs 2/7 for baselines. This teaser provides the formal definhttps://www.babelio.com/actualites.phpition of TEF and the T4 teaser dataset. Full paper and code to be released upon peer review. Related to 10.5281/zenodo.17940301.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.21414424

Author: Gaillard, Michel

Language: English

Topic: Ethics and Social Impacts of AI

Subfield: Safety Research

Field: Social Sciences

Domain: Social Sciences

Open Access status: green

OpenAlex ID: W7169522868

Cites:

Cited by:

Related to: