TOPIC AREA: MACHINE LEARNING/AI, AUTONOMOUS SYSTEMS
KEYWORD: EMERGING TECHNOLOGIES
EVENT TYPE: PANEL
The Future of AI: To Spike or Not to Spike?
Dennis Sylvester - Univ. of Michigan, Ann Arbor, MI
Arijit Raychowdhury - Georgia Institute of Technology, Atlanta, GA
The current wave of AI is changing society and our lives in unprecedented ways. As we look at the challenges of the current hardware and software solutions that power deep neural networks, some key questions emerge. Should we move closer towards neuromorphic systems and borrow cues from the brain? Or should statistical techniques for data analysis take precedence? Will future AI systems use CPU/GPU/FPGAs and specialized accelerators, such as Google’s TPU, in traditional deep neural networks; or are spiking neuromorphic engines, such as IBM’s TrueNorth and Intel’s Loihi, the right path forward? Further, emerging technologies: logic, memory and integration, are expected to play crucial roles in the next-generation of intelligent systems. This panel will feature experts in AI systems as we discuss a wide range of issues in traditional/emerging technologies, circuit-design, architectures, algorithms and computational theory. We will debate whether spiking neural networks or traditional machine-learning algorithms, are expected to dominate the next decade of ubiquitous intelligence, continuous learning and extreme energy-efficiencies -- from the edge to the cloud.
Kaushik Roy - Purdue Univ., West Lafayette, IN Mike Davies - Intel Corp., Hillsboro, OR Brucek Khailany - NVIDIA Corp., Santa Clara, CA Cliff Young - Google, Inc., Mountain View, CA