Year: 2023
Type: preprint
Abstract: Goal misgeneralisation is a key challenge in AI alignment -- the task of getting powerful Artificial Intelligences to align their goals with human intentions and human morality. In this paper, we show how the ACE (Algorithm for Concept Extrapolation) agent can solve one of the key standard challenges in goal misgeneralisation: the CoinRun challenge. It uses no new reward information in the new environment. This points to how autonomous agents could be trusted to act in human interests, even in novel and critical situations.
Source: arXiv (Cornell University)
Authors: Armstrong, Stuart, Maranhão, Alexandre, Daniels-Koch, Oliver, Leask, Patrick, Gorman, Rebecca
Institutions: Charles River Analytics (United States), Durham University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4387210658