Artificial intelligences learn to speak thanks to “language models”. The simplest models allow the autocomplete function on the smartphone: they propose the next word. But the prowess and progress of the most modern language models such as GPT-3, LaMDA, PaLM or ChatGPT are breathtaking, with for example computer programs able to write in the style of a given poet, simulate dead people, explain jokes, translate languages, and even produce and correct computer code – something that would have been unthinkable only a few months ago.

Fabian Suchanek, Professor at Télécom Paris, give more details about these new models on The Conversation.