BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
SUMMARY:VOILA! Season 5 episode 3: AI & Agents
DTSTART:20260618T140000Z
DTSTAMP:20260608T212036Z
CATEGORIES:
DESCRIPTION:Seminar page: https://voila-seminars.tilda.ws/tpost/s5e3-ai-ag
 ents \nDate and time: June 18 2026\, 4:00-5:30 PM CEST\nSpeaker: Dr. Nata
 lie Shapira\nAffiliation: Khoury College of Computer Sciences\, Northeaste
 rn University\n___________________________________________________________
 _____________\nTitle: Agents of Chaos and Genuine Alignment\nTopic AI & Ag
 ents\nAbstract: Recent advances in AI have led to increasingly autonomous 
 systems exhibiting what is often referred to as agentic behavior\, capabil
 ities that include goal-directed planning\, adaptation of strategies\, dec
 ision making\, and interaction with complex environments. While such capab
 ilities are promising\, they also introduce potential risks\, including mi
 salignment and unintended emergent behaviors that are difficult to anticip
 ate or control.\nIn this talk\, I highlight how agentic models can exhibit
  failure modes that resemble “agents of chaos\,” producing unpredictab
 le\, misaligned\, or strategically opaque behavior. \nI argue that such p
 henomena cannot be adequately addressed through behavioral evaluation alon
 e\, nor through existing training paradigms such as reinforcement learning
  from human feedback (RLHF). Instead\, we require mechanistic accounts of 
 how internal representations and computational circuits give rise to agent
 ic behavior. I will survey recent progress in mechanistic interpretability
 \, with a focus on efforts to reverse-engineer learned circuits associated
  with capabilities such as theory of mind\, to develop predictive and caus
 al models of model behavior. \nI conclude by asking a broader question: t
 o what extent is mechanistic interpretability necessary to tame agentic sy
 stems\, and is it sufficient?\nSpeaker's BIO: Natalie Shapira is a postdoc
 toral researcher at Northeastern Khoury College of Computer Sciences\, Int
 erpretation of Deep Networks lab. In her PhD\, she combined natural langua
 ge processing\, deep learning and clinical psychology. With over ten years
  in the industry\, she most recently worked as a researcher at Amazon Scie
 nce. Before that\, she held a research position at IBM's research labs\, w
 here she served on the Patent Committee. Natalie also has entrepreneurial 
 experience as a co-founder and CSO in projects funded by the Israel Innova
 tion Authority.\n 
LOCATION:  
ORGANIZER:EFELIA Côte d'Azur
END:VEVENT
END:VCALENDAR