INSTRUCTOR is a novel approach for computing text embeddings based on task instructions, enabling the generation of task-specific embeddings without additional training. It utilizes the Generalized Text Representation models and multitask instruction-based finetuning to adapt quickly to various language processing tasks.