BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:AI Center - AI Fundamentals Series - Dr. Stanislav Fort\, Google D
 eepMind
DTSTART:20241128T140000
DTEND:20241128T150000
DTSTAMP:20261020T004522Z
UID:8e9911bf2cb1e294a56455d2e0358014393e033c0af6687a7b15fa2b
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Stanislav Fort\n\nTitle\nAdversarial attacks as a baby ver
 sion of A(G)I alignment\n\nAbstract\nAdversarial attacks pose a significan
 t challenge to the robustness\, reliability and alignment of deep neural n
 etworks from simple computer vision to hundred-billion-parameter language 
 models. Despite their ubiquitous nature\, our theoretical understanding of
  their character and ultimate causes\, as well as our ability to successfu
 lly defend against them\, are noticeably lacking. This talk examines the r
 obustness of modern deep learning methods and the surprising scaling of at
 tacks on them\, and showcases several practical examples of transferable a
 ttacks on the largest closed-source vision-language models out there. Buil
 ding on biological insights and new empirical evidence\, I will introduce 
 our solution proposed in [1]\, in which we make a step towards the alignme
 nt of the implicit human and the explicit machine vision representations\,
  closely connecting interpretability and robustness. I will conclude with 
 a direct analogy between the problem of adversarial examples and the much 
 larger task of general AI alignment.\n\n[1] Ensemble everything everywhere
 : Multi-scale aggregation for adversarial robustness. Stanislav Fort\, Bal
 aji Lakshminarayanan\n\nBio \nStanislav Fort is a senior research scienti
 st at Google DeepMind\, specializing in robustness\, interpretability and 
 safety. He received his PhD in 2022 from Stanford University with Prof. Su
 rya Ganguli. In the past\, Stanislav spent time at Google Brain as an AI R
 esident\, worked on the Claude model at Anthropic\, and led the language m
 odel team at Stability AI. He received his Bachelor's and Master's degrees
  in theoretical physics from the University of Cambridge.\n\nAcademic publ
 ications: https://scholar.google.com/citations?user=eu2Kzn0AAAAJ&hl=en&oi
 =ao\nPersonal website: https://stanislavfort.com/
LOCATION:ELE 117 AI Center Lounge https://plan.epfl.ch/?room==ELE%20117 ht
 tps://epfl.zoom.us/j/63398114018?pwd=RBbY1QWbbYGaWedoHSHXlgxk7WGcSy.1
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR