BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Journée Georges de Rham
DTSTART:20260603T150000
DTEND:20260603T173000
DTSTAMP:20260524T044517Z
UID:531ab89672038d95eb5efb5a7e2c781283fab412cd676c7896087cb7
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof Hong Wang\, IHES & NYU\nProf Emmanuel Candès\, Stanford\
 nThe 2026 Journée Georges de Rham will take place on Wednesday 03 June 2
 026 in the Polydôme at EPFL\, from 15:00 to 17:30 followed by an apé
 ritif.\n\nProfessor Wang will deliver a talk titled “Kakeya sets in R
 ^3″. A Kakeya set is a compact subset of R^n that contains a unit lin
 e segment pointing in every direction.  Kakeya set conjecture asserts t
 hat every Kakeya set has Minkowski and  Hausdorff dimension n. We prove
  this conjecture in R^3 as a consequence of a more general statement abo
 ut union of tubes. Joint work with Josh Zahl. \n\nProfessor Candès will
  deliver a talk titled “What Statistics and AI Offer Each Other?“\, e
 xploring how thinking carefully about AI inputs and outputs yields more
  powerful\, safer AI. By examining several vignettes\, we shall answer que
 stions such as: how do we train language models under cost constraints? W
 hat happens when we’ve exhausted all available data? If I start a clini
 cal trial using the drug AI thinks is best\, will it pan out? How can we
  ensure high-quality products when AI is used in a larger workflow? That
  is\, how do I know whether AI automated a task correctly? AI powered i
 mputations are beginning to substitute for real data when collection of th
 e latter is difficult\, slow\, or costly. How then should we leverage mach
 ine learning predictions both as a substitute for high-quality data and 
 as a tool for guiding real data collection?\n\nBernoulli Center for Fundam
 ental Studies \n 
LOCATION:PO 01 https://plan.epfl.ch/?room==PO%2001
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR