Matches in SemOpenAlex for { <https://semopenalex.org/work/W3046275400> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W3046275400 endingPage "1373" @default.
- W3046275400 startingPage "1344" @default.
- W3046275400 abstract "If we could define the set of all bad outcomes, we could hard-code an agent which avoids them; however, in sufficiently complex environments, this is infeasible. We do not know of any general-purpose approaches in the literature to avoiding novel failure modes. Motivated by this, we define an idealized Bayesian reinforcement learner which follows a policy that maximizes the worst-case expected reward over a set of world-models. We call this agent pessimistic, since it optimizes assuming the worst case. A scalar parameter tunes the agent's pessimism by changing the size of the set of world-models taken into account. Our first main contribution is: given an assumption about the agent's model class, a sufficiently pessimistic agent does not cause unprecedented events with probability 1-𝛿, whether or not designers know how to precisely specify those precedents they are concerned with. Since pessimism discourages exploration, at each timestep, the agent may defer to a mentor, who may be a human or some known-safe policy we would like to improve. Our other main contribution is that the agent's policy's value approaches at least that of the mentor, while the probability of deferring to the mentor goes to 0. In high-stakes environments, we might like advanced artificial agents to pursue goals cautiously, which is a non-trivial problem even if the agent were allowed arbitrary computing power; we present a formal solution." @default.
- W3046275400 created "2020-08-07" @default.
- W3046275400 creator A5027690597 @default.
- W3046275400 creator A5073944062 @default.
- W3046275400 date "2020-07-15" @default.
- W3046275400 modified "2023-09-23" @default.
- W3046275400 title "Pessimism About Unknown Unknowns Inspires Conservatism" @default.
- W3046275400 hasPublicationYear "2020" @default.
- W3046275400 type Work @default.
- W3046275400 sameAs 3046275400 @default.
- W3046275400 citedByCount "4" @default.
- W3046275400 countsByYear W30462754002020 @default.
- W3046275400 countsByYear W30462754002021 @default.
- W3046275400 crossrefType "proceedings-article" @default.
- W3046275400 hasAuthorship W3046275400A5027690597 @default.
- W3046275400 hasAuthorship W3046275400A5073944062 @default.
- W3046275400 hasConcept C107673813 @default.
- W3046275400 hasConcept C111472728 @default.
- W3046275400 hasConcept C138885662 @default.
- W3046275400 hasConcept C144237770 @default.
- W3046275400 hasConcept C154945302 @default.
- W3046275400 hasConcept C177264268 @default.
- W3046275400 hasConcept C17744445 @default.
- W3046275400 hasConcept C199360897 @default.
- W3046275400 hasConcept C199539241 @default.
- W3046275400 hasConcept C2777212361 @default.
- W3046275400 hasConcept C33923547 @default.
- W3046275400 hasConcept C41008148 @default.
- W3046275400 hasConcept C94625758 @default.
- W3046275400 hasConcept C96640997 @default.
- W3046275400 hasConcept C97541855 @default.
- W3046275400 hasConcept C9992130 @default.
- W3046275400 hasConceptScore W3046275400C107673813 @default.
- W3046275400 hasConceptScore W3046275400C111472728 @default.
- W3046275400 hasConceptScore W3046275400C138885662 @default.
- W3046275400 hasConceptScore W3046275400C144237770 @default.
- W3046275400 hasConceptScore W3046275400C154945302 @default.
- W3046275400 hasConceptScore W3046275400C177264268 @default.
- W3046275400 hasConceptScore W3046275400C17744445 @default.
- W3046275400 hasConceptScore W3046275400C199360897 @default.
- W3046275400 hasConceptScore W3046275400C199539241 @default.
- W3046275400 hasConceptScore W3046275400C2777212361 @default.
- W3046275400 hasConceptScore W3046275400C33923547 @default.
- W3046275400 hasConceptScore W3046275400C41008148 @default.
- W3046275400 hasConceptScore W3046275400C94625758 @default.
- W3046275400 hasConceptScore W3046275400C96640997 @default.
- W3046275400 hasConceptScore W3046275400C97541855 @default.
- W3046275400 hasConceptScore W3046275400C9992130 @default.
- W3046275400 hasLocation W30462754001 @default.
- W3046275400 hasOpenAccess W3046275400 @default.
- W3046275400 hasPrimaryLocation W30462754001 @default.
- W3046275400 hasRelatedWork W103882264 @default.
- W3046275400 hasRelatedWork W1542236595 @default.
- W3046275400 hasRelatedWork W1561540245 @default.
- W3046275400 hasRelatedWork W2099718073 @default.
- W3046275400 hasRelatedWork W2148122001 @default.
- W3046275400 hasRelatedWork W2395384839 @default.
- W3046275400 hasRelatedWork W2767032540 @default.
- W3046275400 hasRelatedWork W2895994291 @default.
- W3046275400 hasRelatedWork W2921408922 @default.
- W3046275400 hasRelatedWork W2934523877 @default.
- W3046275400 hasRelatedWork W2963097362 @default.
- W3046275400 hasRelatedWork W2963799536 @default.
- W3046275400 hasRelatedWork W3033196763 @default.
- W3046275400 hasRelatedWork W3085778567 @default.
- W3046275400 hasRelatedWork W3124750201 @default.
- W3046275400 hasRelatedWork W3124778320 @default.
- W3046275400 hasRelatedWork W3169857478 @default.
- W3046275400 hasRelatedWork W3186727799 @default.
- W3046275400 hasRelatedWork W3201286590 @default.
- W3046275400 hasRelatedWork W3209066245 @default.
- W3046275400 isParatext "false" @default.
- W3046275400 isRetracted "false" @default.
- W3046275400 magId "3046275400" @default.
- W3046275400 workType "article" @default.