BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.northbaypython.org//JGBYBF
BEGIN:VTIMEZONE
TZID:PST
BEGIN:STANDARD
DTSTART:20001029T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10;UNTIL=20061029T100000Z
TZNAME:PST
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
END:STANDARD
BEGIN:STANDARD
DTSTART:20071104T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:PST
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000402T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4;UNTIL=20060402T110000Z
TZNAME:PDT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20070311T030000
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:PDT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-nbpy-2024-JGBYBF@pretalx.northbaypython.org
DTSTART;TZID=PST:20240629T162000
DTEND;TZID=PST:20240629T164500
DESCRIPTION:As Large Language Models (LLMs) gain trust across various secto
 rs for tasks ranging from generating text to solving complex queries\, the
 ir influence continues to expand. Yet\, this trust is shadowed by signific
 ant risks\, such as the subtle yet serious threat of data poisoning. This 
 talk will delve into how deceptively crafted data can infiltrate an LLM’
 s training set\, leading these models to propagate errors\, biases\, or ou
 tright fabrications—a real challenge to the integrity of their outputs.\
 n\nWhile there are various algorithms and approaches designed to mitigate 
 these risks\, this session will focus particularly on the Rank-One Model E
 diting (ROME) algorithm. ROME is notable for its ability to edit an LLM's 
 knowledge in a targeted manner after training\, providing a means to recal
 ibrate AI outputs. However\, it also presents a potential for misuse\, as 
 it can be employed to embed false narratives deeply within a model.\n\nKey
  Discussion Points:\n- **Why People Trust LLMs**: Exploring the reasons be
 hind the widespread trust in LLMs and the associated risks.\n- **The Art o
 f Data Poisoning**: A closer look at how maliciously crafted data is inser
 ted into training sets and its profound impact on model behavior.\n- **Foc
 us on ROME**: Discussing how the Rank-One Model Editing algorithm can both
  safeguard against and potentially contribute to the corruption of LLMs.\n
 - **Ethical Considerations**: Reflecting on the ethical implications of ma
 nipulating the knowledge within LLMs\, which requires not just technical s
 kill but also wisdom and responsibility.\n\nThis presentation is designed 
 for data scientists\, AI researchers\, and Python enthusiasts interested i
 n understanding the vulnerabilities of LLMs and the tools available to pro
 tect these systems. While acknowledging other algorithms and methods\, thi
 s talk will provide a quick demonstration of ROME\, offering insights into
  its utility and dangers.\n\nAs people continue to integrate LLMs into eve
 rything\, we must remain vigilant against the risks of data manipulation. 
 This session challenges us to consider whether we are paying enough attent
 ion to these threats\, or if we are\, metaphorically\, just fiddling while
  Rome burns—allowing foundational trust in data to erode. \n\nJoin me in
  this exploration of ROME\, where we navigate the fine balance between cor
 recting and corrupting the digital minds that are—whether we like it or 
 not—becoming an integral part of our technological landscape.
DTSTAMP:20260717T142806Z
LOCATION:Barn
SUMMARY:Nightmare on LLM Street: The Perils and Paradoxes of Knowing Your F
 oe - Paris Buttfield-Addison
URL:https://pretalx.northbaypython.org/nbpy-2024/talk/JGBYBF/
END:VEVENT
END:VCALENDAR