BEGIN:VCALENDAR
PRODID:-//Grails iCalendar plugin//NONSGML Grails iCalendar plugin//EN
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTAMP:20261122T182910Z
DTSTART;TZID=America/New_York:20261113T153000
DTEND;TZID=America/New_York:20261113T163000
SUMMARY:CANCELED -  Annual Lecture Series – John Morrison
UID:20261122T182910Z-iCalPlugin-Grails@philevents-web-58d8844b6d-js47p
TZID:America/New_York
LOCATION: University of Pittsburgh\, 4200 Fifth Avenue\, Pittsburgh\, United States\, 15260
DESCRIPTION:<p>The Center for Philosophy of Science at the University of Pittsburgh regrets that&nbsp\; our 67th Annual Lecture Series Talk is CANCELED.</p>\nWe regret that today&rsquo\;s Annual Lecture Series talk by John Morrison has been cancelled because of the cancellation of our speaker&rsquo\;s flight to Pittsburgh.\n\n<p>The Annual Lecture Series\, the Center&rsquo\;s oldest program\, was established in 1960\, the year when Adolf Gr&uuml\;nbaum founded the Center. Each year the series consists of six to eight lectures\, about three quarters of which are given by philosophers\, historians\, and scientists from other universities.</p>\n<p><strong>Annual Leture Series &ndash\;</strong><a  href="https://barnard.edu/profiles/john-morrison"  data-cke-saved-href="https://barnard.edu/profiles/john-morrison"><strong>John Morrison</strong></a></p>\n<p>Friday\, November 13th @ 3:30 pm - 5:30pm EDT</p>\n<p>1008 Cathedral of Learning&nbsp\;</p>\n<p><strong>Title:&nbsp\;</strong>&nbsp\;<strong>Do neural networks understand what they are doing?</strong></p>\n<p><strong>Abstract:</strong>&nbsp\;</p>\n<p>It is often claimed that neural networks construct models of the world (&ldquo\;world models&rdquo\;) and do not merely exploit heuristics. In this sense\, they are said to understand their tasks. But it is not always clear what having a world model amounts to\, or what should count as evidence for or against. In the first half of the talk\, I propose a minimal notion of a world model: a single\, unified representation of a task that preserves its structural relations. I explain how this notion sits within a hierarchy of stronger ones\, and argue that many disputes about world models are merely verbal. In the second half of the talk\, I reassess the standard example of a network using a world model: Othello-GPT\, a transformer widely taken to use a model of the Othello board. I argue that the existing evidence is ambiguous once we adopt the right baselines and the right metrics. I then introduce another kind of evidence that favors the heuristic interpretation: how quickly the network learns variants of Othello that extend rather than scramble the board&rsquo\;s geometry. I end by drawing general lessons for attributing world models to neural networks\, and by describing similar approaches for attributing algorithms and probabilistic inference.</p>\n<p><strong>Can&rsquo\;t make it in-person? This talk will available online through </strong></p>\n<p><strong></strong><strong>Zoom:</strong> <a  href="https://pitt.zoom.us/j/98388637045"  data-cke-saved-href="https://pitt.zoom.us/j/98388637045">https://pitt.zoom.us/j/98388637045</a> and&nbsp\;</p>\n<p><strong>YouTube:</strong>&nbsp\;&nbsp\;<a  href="https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg"  data-cke-saved-href="https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg">https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg</a>.</p>\n<p>A reception with light refreshments will follow in The Center on the 11th floor from 5-6pm.</p>
ORGANIZER;CN=Edouard Machery:
METHOD:PUBLISH
END:VEVENT
END:VCALENDAR
