OpenAI Releases ChatGPT

OpenAI released ChatGPT on November 30, 2022 as a research preview built from the GPT-3.5 family. The web interface made large language models accessible to people who had never used an AI API or read a machine-learning paper.

The model had been trained not only to predict text but also to follow instructions using supervised examples and reinforcement learning from human feedback. Human preference data helped rank candidate responses, and a reward model guided additional training toward answers people generally preferred.

ChatGPT could explain code, summarize text, draft messages, and sustain multi-turn conversations, but it could also invent facts and sources with great confidence. Its popularity changed the software industry almost immediately because language models stopped being mainly a research topic and became a general-purpose user interface.