GPT-3: Language Models are Few-Shot Learners
Give it a few examples in the prompt and it can do a new task without retraining.
175Bparameters, over 100× GPT-2
Why it matters
Prompt engineering was born, and OpenAI offered a model through an API for the first time, starting the commercialization of large models.
Costs and controversy
Expensive to train, sometimes confidently wrong, and not very good at following instructions.
Built on
Led to
Source: arxiv.org ↗