BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Chips revolution: Cerebras talk on accelerating the future of mach
 ine learning
DTSTART:20220302T180000
DTEND:20220302T190000
DTSTAMP:20260415T011308Z
UID:d71b2304807cde8a69a07c635888168a6228fa823e9503435adadbd5
CATEGORIES:Conferences - Seminars
DESCRIPTION:Jean-Philippe Fricker\, Sean Lie\nOn behalf of our partner Cer
 ebras\, the EPFL Alumni cordially invite you to join us on Zoom on Wednesd
 ay March 2\, from 18:00 to 19:00 (UTC+1) to learn more about Cerebras' act
 ivities through their presentation "Thinking Outside the Die: Architecting
  the Machine Learning Accelerator of the Future". The Cerebras team will b
 e available until 19:30 to answer your questions and discuss further their
  activities or career opportunities.\n\nJean-Philippe Fricker: Co-founder 
 and Chief System Architect at Cerebras Systems. He holds a MS in Electrica
 l Engineering from EPFL.\nSean Lie: Co-founder and Chief Hardware Architec
 t at Cerebras Systems. He holds a BS and MEng in Electrical Engineering an
 d Computer Science from MIT.\n\nFree registration here\n\nAbout Cerebras:\
 nCerebras Systems is an American artificial intelligence company that buil
 ds computer systems for complex artificial intelligence deep learning appl
 ications. It was co-founded in 2015 by EPFL alumnus Jean-Philippe Fricker 
 and four colleagues. In November 2021\, Cerebras announced that it had rai
 sed an additional $250 million in Series F funding\, valuing the company a
 t over $4 billion.\n\nConference: \nThe compute and memory demands from st
 ate-of-the-art neural networks have increased several orders of magnitude 
 in just the last couple of years\, and there’s no end in sight. Traditio
 nal forms of scaling chip performance are necessary but far from sufficien
 t to run the machine learning models of the future.\n\nIn this talk\, we w
 ill explore the fundamental properties of neural networks and why they are
  not well served by traditional architectures. We will examine how co-desi
 gn can relax the traditional boundaries between technologies and enable de
 signs specialized for neural networks with new architectural capabilities 
 and performance. We will explore this rich new design space using the Cere
 bras architecture as a case study\, highlighting design principles and tra
 deoffs that enable the machine learning models of the future.
LOCATION:https://forms.gle/KZwooKDQqVtQTLAYA
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR