BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:America/New_York
X-WR-TIMEZONE:America/New_York
BEGIN:VEVENT
UID:884@fds.yale.edu
DTSTART;TZID=America/New_York:20250924T113000
DTEND;TZID=America/New_York:20250924T130000
DTSTAMP:20250918T102516Z
URL:https://fds.yale.edu/events/fds-colloquium-tom-mccoy-yale/
SUMMARY:FDS Colloquium: Tom McCoy (Yale)\, “Understanding AI systems by u
 nderstanding their training data: Memorization\, generalization\, and poin
 ts in between”
DESCRIPTION:\nAbstract:&nbsp\;Large language models (LLMs) can perform a wi
 de range of tasks impressively well. To what extent are these abilities dr
 iven by shallow heuristics vs. deeper&nbsp\;abstractions? I will argue tha
 t\, to answer this question\, we must view LLMs through the lens of genera
 lization. That is\, we should consider the data that LLMs were trained on 
 so that we can identify whether and how their abilities go beyond their tr
 aining data. In the analyses of LLMs that I will discuss\, this perspectiv
 e reveals both impressive strengths and surprising limitations. For instan
 ce\, LLMs often produce sentence structures that are well-formed but that 
 never appeared in their training data\, yet they also struggle on some see
 mingly simple algorithmic tasks (e.g.\, decoding simple ciphers) in ways t
 hat are well-explained by training data statistics. In sum\, to understand
  what AI systems are\, we must understand what we have trained them to be.
 \n\n\n\nSpeaker bio: Tom McCoy is an Assistant Professor of Linguistics at
  Yale University\, with a secondary appointment in Computer Science. His r
 esearch aims to bridge the divide between linguistics and artificial intel
 ligence: how can we create AI systems that replicate the rapid learning an
 d robust generalization that humans display when processing language? Much
  of this work involves analyzing the performance and internal processing o
 f neural network language models. He received his PhD from the Department 
 of Cognitive Science at Johns Hopkins\, and his PhD thesis received a Glus
 hko Dissertation Prize from the Cognitive Science Society. He then did a p
 ostdoc in Computer Science at Princeton before joining the faculty at Yale
 . Outside of research\, he is an organizer and problem writer for NACLO\, 
 a contest that introduces high school students to linguistics and natural 
 language processing.\n\n\n\nWebsite.\n
CATEGORIES:Fellows Events,FDS Events,Colloquium
LOCATION:Yale Institute for Foundations of Data Science\, Kline Tower 13th 
 Floor\, Room 1327\, New Haven\, CT\, 06511\, United States
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Kline Tower 13th Floor\, Ro
 om 1327\, New Haven\, CT\, 06511\, United States;X-APPLE-RADIUS=100;X-TITL
 E=Yale Institute for Foundations of Data Science:geo:0,0
END:VEVENT
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
DTSTART:20250309T030000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR