The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Multi-Device, Multi-Use-Case Optimization with Jeff Gehlhaar - #587
Aug 15, 2022
In this engaging discussion, Jeff Gehlhaar, Vice President of Technology at Qualcomm, shares insights on deploying real-world neural networks and the intricacies of on-device quantization. He emphasizes the synergy between product development and research teams, revealing innovations that optimize AI performance across devices. Jeff highlights exciting automotive applications, particularly automated driver assistance, and discusses the evolution of AI development tools, paving the way for future advancements in technology.
AI Snips
Chapters
Transcript
Episode notes
Cross-Device AI Optimization
- Qualcomm focuses on optimizing AI software across diverse hardware, from cloud to IoT.
- This approach addresses the increasing model complexity and varying power constraints of different devices.
Divergence of Model Complexity and Device Power
- The industry faces a divergence: increasing model complexity alongside decreasing device power, especially in IoT.
- Qualcomm aims to create a common architecture to enable workload portability across devices.
Network Optimality
- Optimize networks for size and performance to address deployment challenges.
- Quantization and other optimization techniques can significantly improve power efficiency without compromising accuracy.