기조연사

Memristive Neuromorphic Technology

• 강성모 교수

  UC산타크루즈

  Biography

  	Memristive Neuromorphic Technology 강성모 교수 UC산타크루즈	x
  학력 Ph.D. University of California, Berkeley (1975) M.S. State University of New York, Buffalo (1972) B.S. Fairleigh Dickinson University, Teaneck, New Jersey (1970) B.S.(honorary) Yonsei University 경력 Distinguished Chair Professor, UC Santa Cruz (2011 - 2013, 2017 - 현재) President, KAIST (2013 - 2017) Chancellor, UC Merced (2007 - 2011) Dean of Engineering, UC Santa Cruz (2001 - 2007) Department Head, Univ. of Illinois at Urbana-Champaign (1995 - 2000) Professor, Univ. of Illinois at Urbana-Champaign (1985 - 2000) MTS Supervisor, AT&T Bell Laboratories, Murray Hill, NJ (1977 - 1985) Assistant Professor, Rutgers University, Piscataway, NJ (1975 - 1977) 연구분야 CMOS VLSI Design Computer-Aided Design of VLSI Compact Modeling of Electronic and Photonic Devices Memristors and Memristive Systems Neuromorphic Computing

  Abstract

  	Imagine 2D or 3D Integration for Brain-Scale Neuromorphic Computing 강성모 교수 UC산타크루즈	x
  For over half a century, the semiconductor industry has made remarkable contributions to benefit human beings globally. Its phenomenal progress through technology scaling has made unimaginable things a reality and Moore’s law has served as a must-do marching order. However, the semiconductor industry has been aware of the ending of Moore's law and the limitation of the von Neumann architecture, especially in view of the power bottleneck and energy efficiency. In the meantime, applications of artificial intelligence (AI) have outpaced Moore’s law in chip development, thus creating an increasingly larger gap between the user demand and the supply the semiconductor industry can attain. In this talk, we will discuss unique roles of memristor technology that can be leveraged in developing upscaled AI neural networks. Biologically inspired, spiking neural networks (SNNs) can perform brain-like neuromorphic computing, and unsupervised learning with high energy efficiency. There are many unknowns about the human brain, but for any electronic system to mimic the brain functions, open-source memristor circuit designs, like the open-source software, are much needed. In this venue, we will discuss how to harness new materials and devices to build memristor neurons, synapses, and their interconnects for ultra-high packing density, low power consumption, and necessary fabrication service as an innovation enabler. A new memristor neuron model, memristive integrated and fire (MIF) circuit, will be introduced as a most compact physical realization of biological neurons. We project that the astronomical number of memristor neurons and synapses, as many as those in a median-size human brain, can be monolithically integrated, in a form of 2D or 3D integration, at a CMOS-memristor technology node of 3.5 nm or less, to mimic many functions of the human brain in the surface area of 2400 ?cm?^2 and a total power consumption in the ballpark of 20W.

인공지능과 반도체: 새로운 일상의 기반

• 최기영 전 과기정통부장관

  서울대학교 (명예교수)

  Biography

  	인공지능과 반도체: 새로운 일상의 기반 최기영 전 과기정통부장관 서울대학교 (명예교수)	x
  학력 서울대학교 전자공학 학사 (1978) KAIST 전기및전자공학 석사 (1980) 스탠포드 전기공학 박사 (1989) 경력 금성사(현 LG) 연구원 (1978 - 1983) 케이던스 연구원 (1989 - 1991) 서울대학교 공과대학 전기정보공학부 교수 (1991 - 2019) 과학기술정보통신부 장관 (2019 - 2021) 연구분야 Electronic Design Automation, Computer Architecture, Artificial Intelligence

  Abstract

  	인공지능과 반도체: 새로운 일상의 기반 최기영 전 과기정통부장관 서울대학교 (명예교수)	x
  과학기술은 우리의 삶에 지대한 영향을 줍니다. 기술패권주의, 감염병, 기후변화 등으로 인한 위기를 극복하는 데에도 과학기술이 중요한 역할을 합니다. 본 강연에서는 먼저 최근에 한꺼번에 몰려온 위기들을 극복하는 과정에서 우리가 수행한 과학기술 정책에 대해 살펴볼 것입니다. 또한 위기 극복과 함께 새로운 일상을 준비하기 위해 국가적으로 추진하고 있는 디지털뉴딜에 대해서도 살펴보고자 합니다. 여기에서는 인공지능이 중요한 부분을 차지하고 있습니다. 인공지능은 새로운 일상의 기반이 될 것이기 때문입니다. 본 강연에서는 인공지능에 대해서 추진해 온 정부의 정책과 함께, 관련해서 우리나라가 경쟁력을 제고하기 위해 집중해야 할 핵심 분야인 반도체에 대한 전략에 대해 이야기를 나눌 것입니다.