SEMICONDUCTOR TECHNOLOGIES DRIVING ARCHITECTURES FOR BRAIN-LIKE EFFICIENT AI
Moderator: ALI KESHAVARZI; Leading Edge Research, LLC
William Chappell; Director, DARPA MTO
Suman Datta; Professor, University of Norte Dame
Wilfried Haensch; Manager, IBM T.J. Watson Research Center
Norm Jouppi; Distinguished Engineer, Google
Amir Khosrowshahi; CTO, AI Products Group, Intel
Dinesh Maheshwari; Partner, Silicon Catalyst
Kaushik Roy; Professor, Purdue University
AI is enabling a variety of applications such as voice and image recognition cost-effectively with low power. As such it is receiving a wide range of attention in public. AI is highly debated among technical communities while commercial companies are rushing for business opportunities. Semiconductor materials, devices, integration techniques and technologies are being probed to enable architectures for AI systems and applications.
Silicon-based semiconductor technology has driven efficient compute per watt for example in the case of application processors utilizing advanced technology nodes with established von Neumann architecture in pervasive mobile devices. Energy-efficient computing measured in MIPS/W is paramount for AI and for the future. The key question is: What drives MIPS/W to achieve efficiency in the range of human brain? What semiconductor technologies will guide architectures behind brain-like energy-efficient computing that is required for AI?
A panel of experts will be discussing and debating various technical directions and market opportunities while exploring various approaches. Technical topics span the range of materials/processes/devices/circuits/integration schemes that are driving architectures toward establishing power efficient AI.