The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

Errol Koolmeister, head of AI Foundation at H&M Group, shares insights on the fashion retail giant's transformative AI journey. He discusses implementing AI for fashion forecasting and pricing, emphasizing the significance of data accessibility and stakeholder engagement. Highlighting the challenges of scaling AI, Errol explains the importance of balancing simplicity with complexity in modeling. He also addresses managing AI initiatives within a large organization, focusing on building a robust infrastructure and fostering an 'AI-first' culture.

Stefano Soatto, VP of AI Application Science at AWS and a professor at UCLA, dives into the fascinating world of Graceful AI. He discusses the challenges of evolving AI in real-world applications while avoiding the pitfalls of constant retraining. Topics include the critical timing of regularization in deep learning, the parallels between model compression and material science, and the intricacies of model reliability. Stefano also unpacks innovations like focal distillation and their potential to enhance lifelong learning in AI systems.

In this engaging discussion, Suchi Saria, Founder and CEO of Bayesian Health and an esteemed professor at Johns Hopkins University, shares her journey at the intersection of machine learning and healthcare. She highlights the slow acceptance of AI in medical practice and discusses pockets of success in the field. Saria elaborates on groundbreaking advancements in sepsis detection and the challenges of integrating ML tools into clinical workflows. Finally, she envisions a future where improved data accessibility drives better patient outcomes in healthcare.

Jeff Gehlhaar, VP of Technology at Qualcomm, dives into the world of AI compilers and their importance in managing parallelism. He highlights Qualcomm's latest innovations, including AI Engine Direct, which bridges capabilities across devices. The conversation covers how research on compression and quantization is translated into real products and the competitive landscape of ML compilers like Glow and TVM. Additionally, Jeff discusses advancements in benchmarking and the integration of AI frameworks that enhance smartphone performance.

Dan Bohus is a senior principal researcher at Microsoft, specializing in human-computer interaction, while Siddhartha Sen focuses on the collaboration between AI and human design. They discuss how projects like Maia Chess and Situated Interaction are redefining human-AI interaction. The conversation covers the challenges of AI understanding human behavior, the integration of natural language processing in chess, and the importance of embodying social cues in machines. Both guests share their excitement for future innovations and the potential for enhanced collaboration between humans and AI.

Julieta Martinez, a senior research scientist at Waabi, dives into the fascinating world of AI and self-driving technology. She highlights how insights from large-scale visual search can bolster neural network compression techniques. The conversation also covers the intricacies of using product quantization to enhance performance while managing vast datasets. Additionally, Julieta discusses her research on deep multitask learning, demonstrating how integrating localization, perception, and prediction can revolutionize autonomous systems and improve real-world applications.

Claire Monteleoni, an associate professor at the University of Colorado Boulder, shares her inspiring journey from environmental activism to a leading role in climate informatics. The discussion covers innovative machine learning techniques for analyzing climate data, particularly unsupervised methods for downscaling. Claire also highlights the evolution of climate informatics conferences and their focus on collaboration. Additionally, she emphasizes the need for integrating social justice in climate action, advocating for resilience in vulnerable communities.

In this engaging discussion, Amir Habibian, a Senior Staff Engineer Manager at Qualcomm Technologies, delves into groundbreaking advancements in video processing. He explores the innovative concept of skip convolutions, which enhance efficiency in visual neural networks. Amir also introduces his FrameExit framework, a conditional early exiting mechanism for video recognition, optimizing how frames are processed. The conversation highlights the future of AI in video technology and the importance of tailoring methods to maximize performance.

Today we’re joined by Noam Slonim, the principal investigator of Project Debater at IBM Research. In our conversation with Noam, we explore the history of Project Debater, the first AI system that can “debate” humans on complex topics. We also dig into the evolution of the project, which is the culmination of 7 years and over 50 research papers, and eventually becoming a Nature cover paper, “An Autonomous Debating System,” which details the system in its entirety. Finally, Noam details many of the underlying capabilities of Debater, including the relationship between systems preparation and training, evidence detection, detecting the quality of arguments, narrative generation, the use of conventional NLP methods like entity linking, and much more.The complete show notes for this episode can be found at twimlai.com/go/495.

Madhur Behl, an Assistant Professor at the University of Virginia, dives into the thrilling world of autonomous racing and AI. He shares insights on how training AI for high-speed environments differs from traditional driving tasks. The conversation covers the unique challenges of perception and planning in racing, as well as the innovative techniques of sensor fusion versus vision-only approaches. Behl also hints at their upcoming race at the Indianapolis Motor Speedway, where his team aims for a million-dollar prize with their fully autonomous vehicle.