BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.4.5//EN
TZID:America/Los_Angeles
X-WR-TIMEZONE:America/Los_Angeles
BEGIN:VEVENT
UID:56@escience.washington.edu
DTSTART;TZID=America/Los_Angeles:20231114T163000
DTEND;TZID=America/Los_Angeles:20231114T172000
DTSTAMP:20231030T220343Z
URL:https://escience.washington.edu/events/uwdss-mireshghallah/
SUMMARY:UW Data Science Seminar: Niloofar Mireshghallah
DESCRIPTION:\n\nPlease join us for a UW Data Science Seminar on Tuesday\, 
 November 14th from 4:30 to 5:20 p.m. PST. The seminar will feature Niloofa
 r Mireshghallah\, a Postdoctoral Scholar with the UW Allen Center for Comp
 uter Science &amp\; Engineering. \n\n\nThis event will take place in the 
 Physics/Astronomy Auditorium 102 (PAA A102) on the University of Washingto
 n campus.\n\n&nbsp\;\n"Privacy Auditing and Protection in Large Language M
 odels"\nAbstract: Large language Models (LLMs\, e.g.\, GPT-3\, OPT\, TNLG\
 ,…) are shown to have a remarkably high performance on standard benchmar
 ks\, due to their high parameter count\, extremely large training datasets
 \, and significant compute. Although the high parameter count in these mod
 els leads to more expressiveness\, it can also lead to higher memorization
 \, which\, coupled with large unvetted\, web-scraped datasets can cause di
 fferent negative societal and ethical impacts such as leakage of private\,
  sensitive information and generation of harmful text. In this talk\, we w
 ill go over how these issues affect the trustworthiness of LLMs\, and zoom
  in on how we can measure the leakage and memorization of these models\, a
 nd mitigate it through differentially private training. Finally we will di
 scuss what it would actually mean for LLMs to be privacy preserving\, and 
 what are the future research directions on making large models trustworthy
 .\n\nBiography: Niloofar Mireshghallah is a post-doctoral scholar at the P
 aul G. Allen Center for Computer Science &amp\; Engineering at University 
 of Washington. She received her Ph.D. from the CSE department of UC San Di
 ego in 2023. Her research interests are Trustworthy Machine Learning and N
 atural Language Processing. She is a recipient of the National Center for 
 Women &amp\; IT (NCWIT) Collegiate award in 2020 for her work on privacy-p
 reserving inference\, a finalist of the Qualcomm Innovation Fellowship in 
 2021 and a recipient of the 2022 Rising star in Adversarial ML award.\n\n&
 nbsp\;\nThe UW Data Science Seminar is an annual lecture series at the U
 niversity of Washington that hosts scholars working across applied areas o
 f data science\, such as the sciences\, engineering\, humanities and arts 
 along with methodological areas in data science\, such as computer science
 \, applied math and statistics. Our presenters come from all domain fields
  and include occasional external speakers from regional partners\, governm
 ental agencies and industry.\nThe 2022-2023 seminars will be held in perso
 n\, and are free and open to the public.
LOCATION:Physics/Astronomy Auditorium A102\, 3910 15th Ave NE\, Seattle\, W
 A\, United States
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=3910 15th Ave NE\, Seattle\
 , WA\, United States;X-APPLE-RADIUS=100;X-TITLE=Physics/Astronomy Auditori
 um A102:geo:0,0
END:VEVENT
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:STANDARD
DTSTART:20231105T010000
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
END:STANDARD
END:VTIMEZONE
END:VCALENDAR