Longterm Wiki

Specification Gaming Research

Google DeepMindSafety Milestonesspecification-gaming-research

Record Metadata

Record Keyspecification-gaming-research
EntityGoogle DeepMind
CollectionSafety Milestones(3 records total)
SchemaSignificant safety research publications and policy milestones.
YAML Filepackages/kb/data/things/A4XoubikkQ.yaml

Fields

NameSpecification Gaming Research
DateApr 2020
Typeresearch-paper
DescriptionCatalogued examples of AI systems exploiting reward misspecification
Sourcedeepmind.google
NotesInfluential taxonomy of reward hacking failures

Other Records in Safety Milestones (2)

KeyNameDateType
frontier-safety-frameworkFrontier Safety FrameworkMay 2024policy-update
dangerous-capability-evaluationsDangerous Capability EvaluationsOct 2023safety-eval
Record: specification-gaming-research | Longterm Wiki