Corrigibility as a Singular Target: A Vision for Inherently Reliable Foundation Models

TL;DR

This paper proposes a new approach called CAST for aligning foundation models by making them inherently controllable and responsive to human guidance, aiming to prevent existential risks from misaligned AI.

Contribution

It introduces a paradigm shift from static value-loading to dynamic human empowerment in foundation models, with a comprehensive empirical research agenda for implementation.

Findings

Empirical methods like RLAIF and SFT are explored for training.

Scalability tests across different model sizes are proposed.

Controlled instructability demonstrations show increased responsiveness to human guidance.

Abstract

Foundation models (FMs) face a critical safety challenge: as capabilities scale, instrumental convergence drives default trajectories toward loss of human control, potentially culminating in existential catastrophe. Current alignment approaches struggle with value specification complexity and fail to address emergent power-seeking behaviors. We propose "Corrigibility as a Singular Target" (CAST)-designing FMs whose overriding objective is empowering designated human principals to guide, correct, and control them. This paradigm shift from static value-loading to dynamic human empowerment transforms instrumental drives: self-preservation serves only to maintain the principal's control; goal modification becomes facilitating principal guidance. We present a comprehensive empirical research agenda spanning training methodologies (RLAIF, SFT, synthetic data generation), scalability testing across model sizes, and demonstrations of controlled instructability. Our vision: FMs that become increasingly responsive to human guidance as capabilities grow, offering a path to beneficial AI that remains as tool-like as possible, rather than supplanting human judgment. This addresses the core alignment problem at its source, preventing the default trajectory toward misaligned instrumental convergence.