BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:'Off-Switch Games' and Corrigibility - Richard Ngo (University of 
 Cambridge)
DTSTART:20171101T170000Z
DTEND:20171101T183000Z
UID:TALK102169@talks.cam.ac.uk
CONTACT:Adrià Garriga Alonso
DESCRIPTION:By default\, an AI system will have an incentive to prevent hu
 mans from switching it off\, or otherwise interfering in its operation\, a
 s this would prevent it from maximising its reward. An AI system is ‘cor
 rigible’ if it has an incentive to accept human corrections. Inverse Rei
 nforcement Learning (IRL) can help mitigate this problem in some cases\, b
 ut there is disagreement as to whether IRL can guarantee corrigibility in 
 all cases.\n\nPapers: https://arxiv.org/abs/1611.08219 https://intelligenc
 e.org/files/Corrigibility.pdf https://intelligence.org/2017/08/31/incorrig
 ibility-in-cirl/
LOCATION: Cambridge University Engineering Department\, CBL Seminar room B
 E4-38.  For directions see http://learning.eng.cam.ac.uk/Public/Directions
END:VEVENT
END:VCALENDAR