Public chatbots

In November 2022 OpenAI releases ChatGPT, a chat window on top of a large language model. The model family, GPT, is older. The new fact is that millions of people can use it without being researchers. Google, Anthropic, and others ship rivals within months.

The contribution is access, not a new machine. The underlying model is the 2017 transformer trained on web-scale text, with extra training so it answers in a conversation. It writes, summarizes, and codes well enough to change office work. It also invents false details with confidence. That limit is why this timeline does not treat its answers as a source.

A chatbot does not weld, drive, or care for a person. Connecting one to tools and to robots is a later, unfinished step. The 2022 product is text.

The labor effect is the one in the McKinsey and Goldman forecasts: some tasks shrink, especially routine writing and junior knowledge work. It is not a date when jobs as a category end.