How OpenAI Builds for 800 Million Weekly Users: Model Specialization and Fine-Tuning

Sherwin Wu from OpenAI discusses the company's approach to balancing its API platform with first-party applications like ChatGPT, addressing potential tensions and competitive concerns. Wu shares insights on the evolution of AI models, the shift from a single "god model" to specialized models, and the importance of fine-tuning APIs for customized use cases. The conversation explores the changing views on prompt engineering, the emergence of agents as a new modality, and the complexities of pricing AI services, including usage-based and outcome-based models. Wu also touches on OpenAI's open-source initiatives, the challenges of verticalizing models for specific products, and the unique infrastructure considerations for handling both language and pixel-based models.