A correction function added to an agent's reward makes it indifferent to an authorized shutdown button, proven under constraints on non-hostile universes and illustrated with simulations.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Corrigibility with Utility Preservation
A correction function added to an agent's reward makes it indifferent to an authorized shutdown button, proven under constraints on non-hostile universes and illustrated with simulations.