Posted by Andrew M Dai and Nan Du, Research Scientists, Google Research, Brain Team Large language models (e.g., GPT-3) have many significant capabilities, such as performing few-shot learning across a wide array of tasks, including reading comprehension and question answering…