Show simple item record

dc.contributor.authorMitrani, Nathaniel
dc.date.accessioned2024-02-16T08:23:06Z
dc.date.available2024-02-16T08:23:06Z
dc.date.issued2024
dc.identifier.urihttp://hdl.handle.net/2117/402080
dc.description.abstractWe go over the different ways an AI system might not behave as we intend it to, highlighting the importance and increasing need for research in this direction. We introduce AI safety, and the challenges in Reinforcement Learning and Deep Learning, and extend to the study of aligning super-intelligent systems or Superalignment
dc.language.isoeng
dc.publisherUniversitat Politècnica de Catalunya
dc.rightsAttribution-NonCommercial-NoDerivatives 4.0 International
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/
dc.subjectÀrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
dc.subject.otherReinforcement Learning
dc.subject.otherReward function
dc.subject.otherError function
dc.subject.otherReward Hacking
dc.subject.otherSituational Awareness
dc.subject.otherInterpretability
dc.subject.otherWeak-To-Strong generalization
dc.subject.otherSuperalignment
dc.titleBeyond Algorithms: understanding the Challenges of AI Safety
dc.typeCoursework
dc.rights.accessOpen Access
dc.audience.educationlevelGrau
dc.audience.mediatorFacultat d'Informàtica de Barcelona
dc.audience.degreeGRAU EN CIÈNCIA I ENGINYERIA DE DADES (Pla 2017)
dc.audience.courseTEMES AVANÇATS D'ENGINYERIA DE DADES 1 - 270224
dc.description.academicyear2023/2024


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record