Leakproofing the Singularity Artificial Intelligence Confinement Problem

83 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Scalable agent alignment via reward modeling: a research direction (2018, background): “• Containment breach (Yampolskiy, 2012; Babcock et al., 2016): the agent might have an incentive to disable or circumvent any containment measures that are intended to limit its operational scope.”
Research Priorities for Robust and Beneficial Artificial Intelligence (2015, background): “systems that could have undesirable behaviors and consequences in less controlled environments (Yampolskiy 2012).”
Utility function security in artificially intelligent agents (2014, background): “However, it is important not to confuse the self-administered Delusion Box with the idea of AI-Boxing, a placement of a potentially unsafe artificial intelligence in a confined environment with no way to escape into the ‘real’ world (Yampolskiy, 2012).”
Motivated Value Selection for Artificial Agents (2015, background): “25It could be set to zero, but this might interfere with the AI’s computation of counterfactuals.”
Good and safe uses of AI Oracles (2017, background): “An Oracle is a common high-level idea for ensuring the safe use of powerful Ar-tiﬁcial Intelligences (AIs) [Armstrong et al., 2012, Yampolskiy, 2012, Bostrom, 2014].”