This podcast episode, recorded on the launch day of GPT-5, features a discussion between Christina Kim and Isa Fulford, both researchers at OpenAI, and Sarah Wang, a general partner at a16z, focusing on the capabilities, improvements, and implications of the new model. They delve into advancements in coding, creative writing, reasoning, and behavior, highlighting the importance of data quality and RL environments. The conversation also covers the concept of AI agents, their potential for asynchronous workflows, and the opportunities GPT-5 unlocks for startups and developers, particularly in coding and application development for non-technical users. The speakers also discuss the balance between model capabilities, data, and scale, as well as the challenges and future directions in creative writing, longer-running tasks, and agent functionalities. They touch on the changing expectations of users regarding speed and thoroughness of AI responses and reflect on the cultural shifts within OpenAI as it has grown.
Sign in to continue reading, translating and more.
Continue