← All inventions
Computing

Large Language Models (AI)

2017 · United States · United States

The transformer design published in 2017 let neural networks learn language from huge amounts of text, leading to systems such as GPT and ChatGPT.

In 2017 researchers at Google published the transformer, a neural network design that handles language well and can be trained on very large amounts of text.

Later models, such as OpenAI's GPT series and Google's BERT, were trained at larger and larger scales. ChatGPT, released in November 2022, brought large language models to a mass audience.

Why it matters: It let computers write, summarize and answer questions in everyday language, and raised new questions about accuracy, bias and work.

Credited to: Google researchers and others

Related: Internet (ARPANET)

No. 100 in our list of 100 inventions. Learn more