BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Paul Christiano (Alignment Research Center)
DTSTART:20231025T220000Z
DTEND:20231025T230000Z
DTSTAMP:20260404T095134Z
UID:MathematicalChallengesAI/1
DESCRIPTION:Title: <a href="https://stable.researchseminars.org/talk/Mathe
 maticalChallengesAI/1/">Formalizing Explanations of Neural Network Behavio
 rs</a>\nby Paul Christiano (Alignment Research Center) as part of Mathemat
 ical challenges in AI\n\nLecture held in Carslaw 273.\n\nAbstract\nExistin
 g research on mechanistic interpretability usually tries to develop an inf
 ormal human understanding of “how a model works\,” making it hard to e
 valuate research results and raising concerns about scalability. Meanwhile
  formal proofs of model properties seem far out of reach both in theory an
 d practice. In this talk I’ll discuss an alternative strategy for “exp
 laining” a particular behavior of a given neural network. This notion is
  much weaker than proving that the network exhibits the behavior\, but may
  still provide similar safety benefits. This talk will primarily motivate 
 a research direction and a set of theoretical questions rather than presen
 ting results.\n
LOCATION:https://stable.researchseminars.org/talk/MathematicalChallengesAI
 /1/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Francois Charton (Meta AI)
DTSTART:20231123T080000Z
DTEND:20231123T090000Z
DTSTAMP:20260404T095134Z
UID:MathematicalChallengesAI/2
DESCRIPTION:by Francois Charton (Meta AI) as part of Mathematical challeng
 es in AI\n\nLecture held in Carslaw 273.\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/MathematicalChallengesAI
 /2/
END:VEVENT
END:VCALENDAR