BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:AI Safety Gridworlds: Is my agent 'safe'? - Jessica Yung (Universi
 ty of Cambridge)
DTSTART:20180228T170000Z
DTEND:20180228T183000Z
UID:TALK101896@talks.cam.ac.uk
CONTACT:Adrià Garriga Alonso
DESCRIPTION:AI Safety Gridworlds are a suite of 2D reinforcement learning 
 environments that test for desirable safety properties of an agent\, such 
 as correct objective specification and robustness.\n\nWe will first discus
 s the paper's approach to formalising safety properties in environments. N
 ext\, we will demo some of the environments and discuss whether they are r
 easonable tests of desirable properties. Finally\, we will discuss why cer
 tain algorithms (among variations of RAINBOW and A2C) seem to have better 
 safety performance than others.\n\nThis week's talk is a good opportunity 
 to get a big-picture view of AI safety from a practical perspective. No pr
 ior knowledge of AI safety is needed.\n\nYou can try the environments for 
 yourself by cloning this git repo: https://github.com/deepmind/ai-safety-g
 ridworlds/tree/master/ai_safety_gridworlds\n\nPaper: AI Safety Gridworlds 
 (Leike et. al.\, 2017) https://arxiv.org/abs/1711.09883
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