How GPT-5 Thinks — OpenAI VP of Research Jerry Tworek

In this episode of The MAD Podcast, Matt Turck interviews Jerry Tworek, VP of Research at OpenAI, about the evolution of AI reasoning models. They discuss the concept of reasoning in AI, particularly the "chain of thought" process, and how models decide how long to "think" to balance result quality and user experience. Tworek shares insights into the development of OpenAI's reasoning models, including O1, O3, and GPT-5, and his journey from mathematics and trading to becoming a leading AI researcher. He also provides a behind-the-scenes look at OpenAI's research culture, project prioritization, and the balance between collaboration and IP protection. The conversation further explores the roles of pre-training and reinforcement learning (RL) in creating modern AI systems, with a detailed explanation of RL principles and its application in various domains, including math, coding, and general problem-solving.