The story
On November 30, 2022, the San Francisco company OpenAI released ChatGPT, a free chatbot that answered questions, wrote essays, poems and computer code, and held conversations in plain language. OpenAI called it a 'research preview'. It reached 1 million users in five days and an estimated 100 million monthly users by January 2023, at the time the fastest growth of any consumer app. For many people it was the first time artificial intelligence felt like something they could talk to.
ChatGPT was a large language model, a neural network trained to predict the next word in a passage, applied to a vast amount of text from the internet and books. Its design came from the Transformer architecture published by Google researchers in 2017. OpenAI, founded in 2015, had scaled that design up in its GPT series; GPT-3 in 2020 had 175 billion adjustable settings, or parameters, and could write fluent text, but it was hard to steer and often rambled or produced offensive material. Microsoft's multibillion-dollar investments paid for the enormous computing power required.
The key addition was training the model to be a helpful assistant. OpenAI used a method called reinforcement learning from human feedback (RLHF): people wrote example answers and ranked the model's responses, and the model was adjusted to prefer the kinds of answers people rated highly. Wrapping this in a simple chat window made it easy to use. The model still made confident mistakes, sometimes called hallucinations, inventing facts or sources that did not exist.
The release set off a race. Microsoft invested about $10 billion more in OpenAI in early 2023 and added the technology to its Bing search engine. Google launched its own chatbot, and Anthropic, Meta and others released competing models, while OpenAI followed with the more capable GPT-4 in March 2023. Teachers rewrote assignments, companies tested AI for writing, customer service and programming, and governments debated rules on safety, copyright, privacy and jobs. Demand for AI chips made Nvidia one of the world's most valuable companies.
Why it mattered
- The European Union passed the AI Act in 2024, the first broad law regulating artificial intelligence.
- AI agents emerged that could use software tools and carry out multistep tasks on a user's behalf.
- Open and cheaper models, such as DeepSeek-R1 from China in 2025, intensified global competition.
- Investment in data centers, chips and electricity for AI surged to hundreds of billions of dollars a year.
- Everyday writing, coding, studying and searching began to change in workplaces and schools.
Sources
Drafted by AI (Claude) from the sources above and checked against them by AI; no historian has reviewed this page. Spot a mistake? Suggest a correction. How pages are made
Community notes
Loading notes…