SIGNAL//DESK
AI/MLsrc: curated AI glossary

Prompt engineering

Prompt engineering is the art of talking to an AI in just the right way so that it gives you the exact answer you are looking for, much like giving clear instructions to a new assistant.

The process of iteratively refining prompts, including context, constraints, and few-shot examples, to optimize an AI model's performance and ensure the generated output consistently meets specific quality and formatting requirements.

The practice of designing model inputs — instructions, context, examples, output format, and constraints — to steer a model toward correct, reliable, and well-formatted outputs.

evolution

  1. 2017-06 · history
    Transformer Architecture

    The 'Attention Is All You Need' paper introduced the mechanism that made models sensitive to specific input sequences, laying the groundwork for prompt-based interaction.

  2. 2020-05 · history
    GPT-3 Few-Shot Learning

    The release of GPT-3 demonstrated that models could perform tasks by providing examples in the prompt, effectively birthing the practice of prompt engineering.

  3. 2021-09 · history
    Chain-of-Thought Prompting

    Researchers introduced techniques to encourage models to generate intermediate reasoning steps, significantly improving performance on complex logic tasks.

  4. 2022-11 · history
    ChatGPT Public Launch

    The widespread availability of a conversational interface turned prompt engineering into a mainstream skill for interacting with Large Language Models.


← all terms