Towards Error Centric Intelligence I, Beyond Observational Learning
Marcus A. Thomas
TL;DR
The paper argues that progress toward AGI is fundamentally theory-limited and cannot be achieved by data and scale alone. It reframes intelligence as open-ended conjecture, criticism, and the capacity to modify the hypothesis space itself, underpinned by a shift from observational to interventional understanding. It introduces the Locality–Autonomy Principle (LAP), a gauge-invariant formulation of Independent Causal Mechanisms (ICM), and the Compositional Autonomy Principle (CAP) as structural diagnostics and design desiderata, collectively forming a Causal Mechanics program. It also previews Energy–Structured Causal Models (E–SCMs) as a modeling framework to operationalize these ideas and enable error discovery, correction, and robust transfer across domains, with diagnostics to guide learning. The framework aims to convert unreachable errors into reachable ones through hypothesis-space changes, offering a path toward open-ended, self-improving intelligences with explicit counterfactual reasoning capabilities.
Abstract
We argue that progress toward AGI is theory limited rather than data or scale limited. Building on the critical rationalism of Popper and Deutsch, we challenge the Platonic Representation Hypothesis. Observationally equivalent worlds can diverge under interventions, so observational adequacy alone cannot guarantee interventional competence. We begin by laying foundations, definitions of knowledge, learning, intelligence, counterfactual competence and AGI, and then analyze the limits of observational learning that motivate an error centric shift. We recast the problem as three questions about how explicit and implicit errors evolve under an agent's actions, which errors are unreachable within a fixed hypothesis space, and how conjecture and criticism expand that space. From these questions we propose Causal Mechanics, a mechanisms first program in which hypothesis space change is a first class operation and probabilistic structure is used when useful rather than presumed. We advance structural principles that make error discovery and correction tractable, including a differential Locality and Autonomy Principle for modular interventions, a gauge invariant form of Independent Causal Mechanisms for separability, and the Compositional Autonomy Principle for analogy preservation, together with actionable diagnostics. The aim is a scaffold for systems that can convert unreachable errors into reachable ones and correct them.