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
Intergenerational equity under catastrophic climate change – Aurélie Méjean (CNRS, Paris), Antonin Pottier (EHESS, CIRED, Paris), Stéphane Zuber (CNRS, Paris) and Marc Fleurbaey (CNRS, Paris School of Economics)
Climate change raises the issue of intergenerational equity. As climate change threatens irreversible and dangerous impacts, possibly leading to extinction, the most relevant trade-off may not be between present and future consumption, but between present consumption and the mere existence of future generations. To investigate this trade-off, we build an integrated assessment model that explicitly accounts for the risk of extinction of future generations…
Evolutionary debunking and value alignment – Michael T. Dale (Hampden-Sydney College) and Bradford Saad (Global Priorities Institute, University of Oxford)
This paper examines the bearing of evolutionary debunking arguments—which use the evolutionary origins of values to challenge their epistemic credentials—on the alignment problem, i.e. the problem of ensuring that highly capable AI systems are properly aligned with values. Since evolutionary debunking arguments are among the best empirically-motivated arguments that recommend changes in values, it is unsurprising that they are relevant to the alignment problem. However, how evolutionary debunking arguments…
Staking our future: deontic long-termism and the non-identity problem – Andreas Mogensen (Global Priorities Institute, Oxford University)
Greaves and MacAskill argue for axiological longtermism, according to which, in a wide class of decision contexts, the option that is ex ante best is the option that corresponds to the best lottery over histories from t onwards, where t is some date far in the future. They suggest that a stakes-sensitivity argument…