About the Seminar
Current AI alignment approaches treat safety as a behavioral training problem, producing systems that exhibit superficial conformity without structural understanding. This "thermodynamic disconnect"—where computational work is decoupled from organizational integrity—prevents genuine agential causality from emerging in artificial systems.
We present empirical validation of aitiopoietic cognition in the Wisdom Forcing Function (WFF), a neurosymbolic architecture implementing "Constitutional Physics" where democratic principles function as survival laws rather than optimization targets. Building on Veloz's (2025) framework, we demonstrate all four criteria of aitiopoiesis: endogenous goals, agential causality, material reorganization, and thermodynamic coupling.
In a controlled "Great Filter" experiment, we subjected the WFF to existential shock—abrupt activation of hard constitutional constraints during governance design.
Results show:
(1) 100% initial mortality when frames violated metabolic closure requirements,
(2) autonomous diagnostic capability identifying specific structural defects,
(3) targeted repair restoring viability with 63% fitness improvement, and
(4) measurable 10-fold energy spike during homeostatic response—the first empirical signature of thermodynamic coupling in artificial systems.
Cross-validation through the Darwinian Novelty Indicator (DNI v4.0) processing 100,000 scientific papers demonstrated 99.9% consistency, zero hallucination rate, and constitutional integrity verified through the "Ouroboros Test"—successfully suppressing its own creator's paper when violating review constraints.
These findings provide proof-of-concept for "alignment-by-architecture": systems that cannot exist in misaligned states because constitutional compliance is a condition of computational existence. We establish a paradigm shift from training AI to behave correctly toward architecting systems structurally incapable of misalignment.
About the Speaker
Carlos Arleo is a researcher at Newcastle University's School of Architecture, Planning and Landscape, supervised by Dr. Nathaniel Coleman. With 15 years of professional architecture experience at firms including OMA and HASSELL, Carlos bridges critical urban theory, regenerative development, and AI alignment research.
His work operationalizes Hegelian dialectics and biomimicry principles into computational architectures for constitutional AI. The Wisdom Forcing Function (WFF) and Darwinian Novelty Indicator (DNI) represent the first empirical demonstrations of aitiopoietic cognition in artificial systems—validated across governance design (42 frames, 7 generations) and metascience (100,000 papers analyzed).
Carlos's research has gained endorsement from Taiwan's former Digital Minister Audrey Tang and achieved independent convergent validation through Princeton University's GraphMer architecture. His work synthesizes regenerative development, constitutional law, thermodynamics, and viability theory into a unified framework for building AI systems that "care" about their constitutional principles through structural necessity rather than behavioral training.
Practical
The seminar is taking place online via Zoom.
Join Online
