From the archive
OpenAI releases ChatGPT as a free research preview
ChatGPT introduces conversational follow-up questions, corrections, and a public interface for testing an AI assistant.
OpenAI released ChatGPT as a free research preview on November 30, 2022. The service put a conversational interface around a language model, with follow-up questions and corrections in the same conversation.
A conversation rather than a single prompt
ChatGPT was trained to answer questions, acknowledge mistakes, challenge incorrect assumptions, and refuse some requests. Its dialogue format made it possible to refine an answer over several exchanges rather than submit a separate prompt for every revision.
The initial announcement included examples of explaining code and asking for more information when a request was ambiguous. That made the preview relevant to everyday writing and technical questions as well as experiments with language models.
Free access during the preview
The public preview was free while OpenAI gathered feedback. The announcement also described problems with plausible but incorrect answers and sensitivity to prompt wording. The launch was a chance to test conversational AI, with results still requiring checking.