The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Accelerating Distributed AI Applications at Qualcomm with Ziad Asghar - #489
Jun 3, 2021
Ziad Asghar, Vice President of Product Management at Qualcomm, discusses the dynamic interplay of AI and 5G, showcasing how Qualcomm's Snapdragon platform fuels mobile AI advancements. He dives into the evolution of the Cloud AI 100 and its impact on smart city technologies. Ziad also highlights the significance of federated learning and the importance of privacy in data security. Additionally, he explores AI innovations in automotive technology, including collaborations with NASA to enhance vehicle autonomy and safety.
AI Snips
Chapters
Transcript
Episode notes
On-Device AI Requirements
- On-device AI requires efficient hardware, optimized software, and advanced algorithms.
- Qualcomm's AI Model Efficiency Toolkit helps quantize and compress networks, improving power efficiency.
Snapdragon 888
- The Snapdragon 888 boasts a heterogeneous AI engine with a combined performance of 26 TOPS.
- Architectural improvements in the 888, like a larger shared memory, reduce power consumption and latency.
Cloud AI 100
- Qualcomm's Cloud AI 100 is designed for cloud inference, addressing the growing demand and power consumption in data centers.
- It excels in performance per watt, especially in lower power configurations, as demonstrated by ML Commons benchmarks.