BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:ICTP Colloquium on AI for Scientific Discovery - The Language of N
 atural Numbers: Statistical Properties and Neural Learning of Arithmetic S
 tructures
DTSTART;VALUE=DATE-TIME:20261028T130000Z
DTEND;VALUE=DATE-TIME:20261028T140000Z
DTSTAMP;VALUE=DATE-TIME:20261114T233949Z
UID:indico-event-11555@ictp.it
DESCRIPTION:\n	Pierluigi Contucci\n	(Università di Bologna\, Italy)\n	 \
 n\n	\n		 \n	\n		PhD in Mathematics and MS in Mathematical Physics (ISAS\,
  Trieste)\, Laurea in Physics (University La Sapienza\, Rome). Postdoc at 
 Princeton University and University of California. Visiting Professor at t
 he Tokyo Institute of Technology\, at the Institut Henri Poincaré and at 
 New York University. Research interests in Mathematical Physics\, Statisti
 cal Mechanics and applications (machine learning\, data science). Author o
 f Perspectives on Spin Glasses (Cambridge University Press\, 2012) and Pro
 babilità Elementare\, Teoria ed Esperimenti (Zanichelli\, 2008). Publicat
 ions in Communications in Mathematical Physics\, Nature Scientific Reports
 \, Journal of Statistical Physics\, Annales Henri Poincare\, Physical Revi
 ew Letters. Writes on scientific debates of public interest in newspapers 
 like Il Corriere della Sera\, Il Sole 24 Ore and magazines like Il Mulino 
 and Vanity Fair. Schloessman Award (2000) from the Max Planck Society. Mem
 ber of the Science Academy of Bologna.\n		Abstract:The relationship betwee
 n artificial intelligence and science is increasingly bidirectional: AI of
 fers new tools for discovering and exploring scientific structures\, while
  science provides controlled domains in which to investigate the capabilit
 ies and limitations of AI. We examine this interplay in the fundamental se
 tting of arithmetic. Iterated prime factorization turns the integers into 
 a deterministic sequence of words whose vocabulary and admissible combinat
 ions are governed by arithmetic. We establish asymptotic laws for vocabula
 ry growth and word frequencies\, examine statistical dependencies\, and sh
 ow that a transformer trained solely on such text captures structural regu
 larities\, including sensitivity to arithmetically forbidden sequences. Th
 is resulting language supports both directions of inquiry: neural models m
 ake arithmetic structure accessible through learning\, while arithmetic su
 pplies precise constraints against which to examine what these models lear
 n and how they represent it.\n\n\n	- - -\n	 \n\n	Refreshments will be ser
 ved at the LB Cafeteria after the talk.\n\n	 \n\n	- - -The event will be 
 livetreamed at:ictp.it/livestream\n\n//indico.ictp.it/event/11555/
LOCATION:ICTP Euler Lecture Hall
URL://indico.ictp.it/event/11555/
END:VEVENT
END:VCALENDAR