ИИ и технологий

Aligning language models to follow instructions

We’ve trained language models that are much better at following user intentions than GPT-3 while also making them more truthful and less toxic, using techniques developed through our alignment researc