La parole aux machines

Description:

An essay on large language models by Thibaut Giraud, a philosopher specialising in logic and philosophy of language, and creator of the French philosophy channel « Monsieur Phi ». Neural networks trained on vast text corpora to do one thing — predict the next word — now hold conversations, and Giraud argues that the scale of that shift is still widely underestimated. He retraces the history and the workings of these systems, then examines what it means philosophically that human beings are no longer the only producers of language.

subjects: Large language models,  Artificial intelligence,  Philosophy of mind,  Philosophy of language,  Machine learning,  Deep learning,  Consciousness,  Ethics of artificial intelligence,  AI alignment,  ChatGPT

Times: 21st Century