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
Calibration dilemmas in the ethics of distribution – Jacob M. Nebel (University of Southern California) and H. Orri Stefánsson (Stockholm University and Swedish Collegium for Advanced Study)
This paper presents a new kind of problem in the ethics of distribution. The problem takes the form of several “calibration dilemmas,” in which intuitively reasonable aversion to small-stakes inequalities requires leading theories of distribution to recommend intuitively unreasonable aversion to large-stakes inequalities—e.g., inequalities in which half the population would gain an arbitrarily large quantity of well-being or resources…
Economic growth under transformative AI – Philip Trammell (Global Priorities Institute, Oxford University) and Anton Korinek (University of Virginia)
Industrialized countries have long seen relatively stable growth in output per capita and a stable labor share. AI may be transformative, in the sense that it may break one or both of these stylized facts. This review outlines the ways this may happen by placing several strands of the literature on AI and growth within a common framework. We first evaluate models in which AI increases output production, for example via increases in capital’s substitutability for labor…
AI alignment vs AI ethical treatment: Ten challenges – Adam Bradley (Lingnan University) and Bradford Saad (Global Priorities Institute, University of Oxford)
A morally acceptable course of AI development should avoid two dangers: creating unaligned AI systems that pose a threat to humanity and mistreating AI systems that merit moral consideration in their own right. This paper argues these two dangers interact and that if we create AI systems that merit moral consideration, simultaneously avoiding both of these dangers would be extremely challenging. While our argument is straightforward and supported by a wide range of pretheoretical moral judgments, it has far-reaching…