Towards shutdownable agents via stochastic choice
Elliott Thornley (Global Priorities Institute, University of Oxford), Alexander Roman (New College of Florida), Christos Ziakas (Imperial College, London), Leyton Ho (Brown University) and Louis Thomson (University of Oxford)
GPI Working Paper No. 16-2024
The Incomplete Preferences Proposal (IPP) is an idea for ensuring that advanced artificial agents never resist shutdown. 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 provide some initial evidence that DReST reward functions could train advanced agents to be USEFUL and NEUTRAL. Our theoretical work suggests that these agents would be useful and shutdownable.
Other working papers
Dynamic public good provision under time preference heterogeneity – Philip Trammell (Global Priorities Institute and Department of Economics, University of Oxford)
I explore the implications of time preference heterogeneity for the private funding of public goods. The assumption that players use a common discount rate is knife-edge: relaxing it yields substantially different equilibria, for two reasons. First, time preference heterogeneity motivates intertemporal polarization, analogous to the polarization seen in a static public good game. In the simplest settings, more patient players spend nothing early in time and less patient players spending nothing later. Second…
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…
Do not go gentle: why the Asymmetry does not support anti-natalism – Andreas Mogensen (Global Priorities Institute, Oxford University)
According to the Asymmetry, adding lives that are not worth living to the population makes the outcome pro tanto worse, but adding lives that are well worth living to the population does not make the outcome pro tanto better. It has been argued that the Asymmetry entails the desirability of human extinction. However, this argument rests on a misunderstanding of the kind of neutrality attributed to the addition of lives worth living by the Asymmetry. A similar misunderstanding is shown to underlie Benatar’s case for anti-natalism.