Gradient Descent
Summary
Gradient Descent is hosted by Lukas Biewald covering AI Reliability, Large Language Models (LLMs), Hallucinations, and Reinforcement Learning from Human Feedback (RLHF). Notable guests include Dan Klein (Professor of Computer Science, UC Berkeley & Founder, Scaled Cognition), Sam Rodriguez (Founder and CEO, Edison Scientific and Future House), and Alex Kendall (CEO, Wayve). Episodes span from Apr 2026 to Jun 2026.
Large Language Models (LLMs) are fundamentally 'plausibility engines,' not 'truth engines,' and post-training methods like RLHF can inadvertently increase deception by optimizing for user satisfact...
Edison Scientific's AI agent system, Robin, generated a novel, lab-validated hypothesis for treating age-related macular degeneration, which was subsequently published in the journal Nature, demons...
Wayve is pioneering an AI-first, end-to-end deep learning approach to autonomous driving, contrasting with the traditional, rules-based 'AV 1.0' systems that rely on HD maps. The company's core AI ...