BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Cambridge ELLIS Seminar Series- Dr. Petar Veličković - 28 April 
 2023- 15:00 - Speaker to be confirmed
DTSTART:20230428T140000Z
DTEND:20230428T150000Z
UID:TALK200296@talks.cam.ac.uk
CONTACT:Kimberly Cole
DESCRIPTION:Title: “Reasoning Algorithmically: from Toy Experiments to A
 GI Modules "\n \nAbstract: Neural networks that are able to reliably execu
 te algorithmic computation may hold transformative potential to both machi
 ne learning and theoretical computer science. On one hand\, they could ena
 ble the kind of extrapolative generalisation scarcely seen with deep learn
 ing models. On another\, they may allow for running classical algorithms o
 n inputs previously considered inaccessible to them.\nOver the past few ye
 ars\, the pace of development in this area has gradually become intense. A
 s someone who has been very active in its latest incarnation\, I have witn
 essed these concepts grow from isolated 'toy experiments'\, through NeurIP
 S spotlights\, all the way to helping detect patterns in complicated mathe
 matical objects (published on the cover of Nature) and supporting the deve
 lopment of generalist reasoning agents.\nIn this talk\, I will give my per
 sonal account of this journey\, and especially how our own interpretation 
 of this methodology\, and understanding of its potential\, changed with ti
 me. It should be of interest to a general audience interested in graphs\, 
 (classical) algorithms\, reasoning\, and building intelligent systems.\n
LOCATION:https://eng-cam.zoom.us/j/84673427243?pwd=VUd5QncxK0RDNkpqc3hzZmp
 kQnptdz09 
END:VEVENT
END:VCALENDAR