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

Filip Kozera, Founder and CEO of Wordware, is on a mission to revolutionize how we interact with AI through natural language as the new programming interface. He discusses the architecture of AI agents, emphasizing the need for 'graceful recovery' systems that involve humans when agents hit knowledge limits. The conversation explores the shift to user-centric workflows and the challenges of data silos in SaaS platforms. Filip's vision for the 'word artisan' potentially transforms non-technical users into AI managers, reshaping knowledge work.

Jared Quincy Davis, Founder and CEO at Foundry and a former DeepMind core deep learning team member, discusses transformative 'compound AI systems' that merge diverse AI models for superior performance. He introduces 'laconic decoding' and explains how these systems can boost efficiency while cutting costs. The conversation covers the interplay between AI algorithms and cloud infrastructure, the evolution of ensemble models, and the potential of hybrid systems. Davis emphasizes co-design and innovative strategies to revolutionize the AI landscape and enhance developer experience.

Kwindla Kramer, co-founder and CEO of Daily, shares his insights on building real-time conversational voice AI. He discusses the full stack of voice agents, emphasizing the importance of a modular approach for better latency and cost-efficiency. Kwindla delves into challenges like interruption handling and natural dialogue dynamics. He also highlights the future of voice AI in use cases, hybrid edge-cloud pipelines, and exciting advancements like real-time video avatars. It's a comprehensive look at the dynamic world of voice technology!

In this conversation with Fatih Porikli, Senior Director of Technology at Qualcomm AI Research, he unveils cutting-edge innovations from the CVPR conference. He discusses DiMA, a groundbreaking system using large language models for safe autonomous driving, dramatically reducing collision rates. Fatih also dives into SharpDepth, enhancing depth prediction through diffusion distillation. He highlights impressive on-device demos, from text-to-3D mesh generation to real-time video fabrication, showcasing the future of AI and computer vision.

Vijoy Pandey, SVP at Outshift by Cisco, shares insights on creating the "Internet of Agents" to improve collaboration among diverse agent systems from vendors like Salesforce and Microsoft. He discusses the challenges of integrating these systems and introduces AGNTCY, an open-source project aimed at enhancing interoperability. Vijoy breaks down the four phases of agent collaboration and reveals SLIM, a new transport layer ensuring secure, real-time communication. The conversation sheds light on overcoming semantic challenges and the importance of evolving communication protocols in AI.

In this enlightening discussion, Ben Wellington, Deputy Head of Feature Forecasting at Two Sigma, shares his expertise in AI-driven equity feature forecasting. He delves into the intricacies of identifying and quantifying measurable features to improve predictive accuracy. The use of satellite imagery for data points like vehicle counts unveils unique insights. Ben emphasizes the importance of strict data timestamping to avoid temporal leakage and discusses the transformative role of large language models in enhancing data analysis. He also offers a glimpse into the future of agentic AI in finance.

Join Jason Corso, co-founder of Voxel51 and University of Michigan professor, as he unpacks the fascinating world of automated labeling in computer vision. Discover FiftyOne, a tool for visualizing datasets and enhancing data quality. Jason reveals how zero-shot auto-labeling can rival human performance, offering significant efficiency gains. He also dives into the challenges of label quality, decision boundaries, and the innovative 'verified auto-labeling' method. Plus, learn about synthetic data generation and the exciting future of agentic behaviors in AI!

In this insightful conversation, Charles Martin, the founder of Calculation Consulting and an AI researcher merging physics with machine learning, introduces WeightWatcher, a groundbreaking tool for enhancing Deep Neural Networks. He explores the revolutionary Heavy-Tailed Self-Regularization theory and how it exposes phases like grokking and generalization collapse. The discussion delves into fine-tuning models, the perplexing relationship between model quality and hallucinations, and the challenges of generative AI, providing valuable lessons for real-world applications.

Logan Kilpatrick and Shrestha Basu Mallick from Google DeepMind dive into groundbreaking advancements from Google I/O 2025. They discuss the Gemini API's impressive features like thinking budgets and thought summaries, enhancing voice AI’s expressiveness with native audio output. The duo shares insights on the challenges of building real-time voice applications, including latency and voice detection. They also send a playful wish list for next year's event, dreamily aiming for enhanced language capabilities to foster global inclusivity.

Sebastian Gehrmann, head of Responsible AI at Bloomberg, dives into the complexities of AI safety, particularly in retrieval-augmented generation (RAG) systems. He reveals how RAG can unintentionally compromise safety, even leading to unsafe outputs. The conversation highlights unique risks in financial services, emphasizing the need for specific governance frameworks and tailored evaluation methods. Gehrmann also addresses prompt engineering as a strategy for enhancing safety, underscoring the necessity for ongoing collaboration in the AI field to tackle emerging vulnerabilities.