In this talk, we’ll explore the evolution and implementation of multimodal generative AI at the edge targeting physical AI applications, including autonomous vehicles and humanoid robots. We’ll outline the challenges and trade-offs for implementing multimodal generative AI on resource-constrained systems and discuss the performance and bandwidth requirements NPUs must meet to support them. Finally, we’ll demonstrate how these advanced multimodal and action-capable models are efficiently accelerated on the Synopsys ARC enhanced NPX6 NPU IP, enabling the next wave of intelligent, adaptive edge devices.
From Generative AI to Physical AI: Enabling VLM/VLA Multimodal Performance on Enhanced NPX6 NPU IP
Session Speakers
-
Gordon Cooper
Principal Product Manager, Synopsys
Gordon Cooper is a Principal Product Manager at Synopsys, responsible for the ARC AI processor family, particularly the NPX6 neural processing unit IP. He has over 35 years of experience in processor, digital signal processing and artificial intelligence roles at Synopsys, NXP Semiconductors, Analog Devices and Raytheon. Gordon also served as a Commanding Officer in the US Army Reserve, including tours in Bosnia and Kosovo.
