BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//NextToU//EN
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:53d3831d-4d9b-40e4-bf9a-55a6013949c3@nexttoyou
DTSTAMP:20261105T155334Z
DTSTART:20261109T170000Z
DTEND:20261109T180000Z
SUMMARY:LMSS @ Cornell Tech: David Bau (Northeastern University)
LOCATION:Cornell Tech Bloomberg 091\, 2 West Loop Road\, New York\, NY 100
 44
DESCRIPTION:Learning Machines Seminar Series\n\nWhat: LMSS: David Bau (Nor
 theastern University)\nWhen: Monday\, November 9\, 12:00 p.m. - 1:00 p.m.
 \nWhere: Bloomberg 091\n\nThe series is organized by Associate Professor 
 Yoav Artzi and sponsored by Bloomberg.\n\nPizza will be served at 11:45 a
 .m.\n\n"Interpretability\, Indirection and Introspection"\n\nWhen an LLM 
 tells you about its own thoughts\, does it really know them? The words "I
  know" or "I think" or "I want" are supposed to describe your own interna
 l thoughts\, but is it even meaningful to impart any meaning to such intr
 ospective language in an LLM? Does it even have any internal thoughts to 
 describe? In this talk I will discuss some of the latest interpretability
  research in my lab\, providing a brief overview of our work on in-contex
 t learning and binding. Then I will delve into our newest work studying t
 he phenomenon and neural mechanisms underlying LLM introspection. I will 
 describe a very simple experimental procedure for imparting new knowledge
  into an LLM that allows us to measure whether it can introspect. Then I 
 will show how you can use this method to conduct a series of interpretabi
 lity experiments to localize and relate introspection to other task perfo
 rmance. Our main question is: can LLM introspection be grounded\, looking
  inward at causl physical neural states in the model? I will share insigh
 ts about the relationship between introspection and model size\, metalear
 ning\, and training\, and I will discuss the implications of our work for
 \n\nhttp://next-to-u.app/events/53d3831d-4d9b-40e4-bf9a-55a6013949c3
URL:http://next-to-u.app/events/53d3831d-4d9b-40e4-bf9a-55a6013949c3
END:VEVENT
END:VCALENDAR
