ChipChats: Measure What Matters - Intelligence per Watt & Joule

Join SambaNova and Stanford Research for an interview with the authors of "Intelligence per Watt: Measuring Intelligence Efficiency of Local AI," as we dive deep into the principles, innovations, and real-world applications of energy-efficient AI design. In this live session, moderated by Vasanth Mohan (SambaNova), the authors will share insights from their groundbreaking work, unpacking their learnings from benchmarking a wide range of model-hardware pairs.

When: January 13th - 10am PT | 1pm ET

What You’ll Learn:

• What is Intelligence per Watt and how is it measured

• Why now is the time to focus on measuring this metric in this pivotal moment of AI

• Key strategies to improve Intelligence per Watt with local computing

• How SambaNova RDUs demonstrate much higher tokens/watt

• Future work for the Stanford Research team

Meet the Speakers:

Avanika Narayan - Avanika is a PhD candidate in Computer Science at Stanford University's School of Engineering, where she focuses on advancing AI technologies with an emphasis on efficiency and scalability. As a co-founder and AI lead at Rox, she brings practical experience in developing AI-driven solutions for real-world applications. (LinkedIn)

Jon Saad-Falcon - Jon is a joint PhD and MBA student in Computer Science at Stanford University, advised by Azalia Mirhoseini in the Scaling Intelligence Lab and Christopher Ré in the Hazy Research group. His research centers on natural language processing, information retrieval, machine learning systems, and efficient AI deployment. (LinkedIn)