Towards shutdownable agents via stochastic choice
Elliott Thornley (Global Priorities Institute, University of Oxford), Alexander Roman (New College of Florida), Christos Ziakas (Independent), Leyton Ho (Brown University) and Louis Thomson (University of Oxford)
GPI Working Paper No. 16-2024
Some worry that advanced artificial agents may resist being shut down. The Incomplete Preferences Proposal (IPP) is an idea for ensuring that doesn’t happen. A key part of the IPP is using a novel ‘Discounted REward for Same-Length Trajectories (DREST)’ reward function to train agents to (1) pursue goals effectively conditional on each trajectory-length (be ‘USEFUL’), and (2) choose stochastically between different trajectory-lengths (be ‘NEUTRAL’ about trajectory-lengths). In this paper, we propose evaluation metrics for USEFULNESS and NEUTRALITY. We use a DREST reward function to train simple agents to navigate gridworlds, and we find that these agents learn to be USEFUL and NEUTRAL. Our results thus suggest that DREST reward functions could also train advanced agents to be USEFUL and NEUTRAL, and thereby make these advanced agents useful and shutdownable.
Other working papers
The unexpected value of the future – Hayden Wilkinson (Global Priorities Institute, University of Oxford)
Various philosophers accept moral views that are impartial, additive, and risk-neutral with respect to betterness. But, if that risk neutrality is spelt out according to expected value theory alone, such views face a dire reductio ad absurdum. If the expected sum of value in humanity’s future is undefined—if, e.g., the probability distribution over possible values of the future resembles the Pasadena game, or a Cauchy distribution—then those views say that no real-world option is ever better than any other. And, as I argue…
Social Beneficence – Jacob Barrett (Global Priorities Institute, University of Oxford)
A background assumption in much contemporary political philosophy is that justice is the first virtue of social institutions, taking priority over other values such as beneficence. This assumption is typically treated as a methodological starting point, rather than as following from any particular moral or political theory. In this paper, I challenge this assumption.
Imperfect Recall and AI Delegation – Eric Olav Chen (Global Priorities Institute, University of Oxford), Alexis Ghersengorin (Global Priorities Institute, University of Oxford) and Sami Petersen (Department of Economics, University of Oxford)
A principal wants to deploy an artificial intelligence (AI) system to perform some task. But the AI may be misaligned and aim to pursue a conflicting objective. The principal cannot restrict its options or deliver punishments. Instead, the principal is endowed with the ability to impose imperfect recall on the agent. The principal can then simulate the task and obscure whether it is real or part of a test. This allows the principal to screen misaligned AIs during testing and discipline their behaviour in deployment. By increasing the…