What is Prompt Engineering?

Prompt engineering (PE) is a concept in artificial intelligence, particularly natural language processing (NLP). It’s a technique where the description of the task is embedded in the input, for example, as a question instead of it being implicitly given. This method has been gaining popularity in recent years as it allows for more efficient and accurate training of language models.

How Prompt Engineering Works

Prompt engineering converts one or more tasks to a prompt-based dataset and training a language model with what has been called “prompt-based learning”. Another way of implementing prompt engineering is to use a large “frozen” pre-trained language model and only learn the representation of the prompt. We call that “prefix-tuning” or “prompt tuning”.

Advancements in Prompt Engineering

The GPT-2 and GPT-3 language models were important steps in the field of PE. In 2021, multitask PE using multiple NLP datasets showed good performance on new tasks. Prompts that include a train of thought in few-shot learning examples show a better indication of reasoning in language models.

Text-to-Image Prompting

In 2022, machine learning models like DALL-E, Stable Diffusion, and Midjourney were released to the public. These models take text prompts as input and use them to generate images. That way we got new category of PE related to text-to-image prompting. With the broad accessibility of these tools, we expect that PE will continue to evolve and improve in the future.