ChatGPT launched as a research preview on November 30, 2022. OpenAI adapted its instruction-following work into a conversational format, letting people ask follow-up questions and work through a problem across multiple turns. The preview was also a way to learn directly from how people used the system and where it failed. The initial model was fine-tuned from the GPT-3.5 series using human-written conversations and comparisons between possible responses. That combination of a simple chat interface and iterative feedback became the starting point for ChatGPT's subsequent development.
