Overview
Prompt engineering is the art of designing and developing prompts. A prompt is user text, a string of words like simple statements, code, etc, submitted to an AI model/tool to get unique responses. Prompts are a critical component of the AI conversation.
Prompt Engineering
Prompt engineering is the art of creating effective prompts for the AI LLM( Large Language Model) to get the intended generated output responses from the model.
The prompt user should understand the model’s capabilities and limitations. Prompt engineering skills allow the user to understand the model better and craft useful prompts for the LLMs.
LLM examples
Some of the examples of LLMS are as follows:
- GPT-3/GPT-4
- Google Bard
Types of Prompts
Prompts can be broadly classified into the following types:
- Effective prompts
- Ineffective prompts
Effective prompts get high-quality responses from the AI model compared to ineffective prompts.
Prompting Engineering Techniques
In this section, we will outline different prompting techniques. Prompting engineering techniques are strategies used to guide the behavior of AI models like LLMs in generating desired responses.
Some prompting techniques are as follows:
- Zero-shot Prompting
- Few-shot Prompting
- Chain-of-Thought(CoT)
- Role-based Prompting
- Self-consistency
- General knowledge Prompting
- Tree of Thoughts
Zero-shot Prompting: Providing the model with a task without any previous examples.
Few-shot Prompting: Give the model a few examples to illustrate the task before asking it to perform.
Chain-of-Thought (CoT) Prompting: Encouraging the model to “think aloud” or show its reasoning process step by step.
Role-based prompting: involves tailoring prompts or instructions according to individuals’ roles or responsibilities within an organization.
Self-Consistency: Asking the model the same question multiple times to ensure consistent answers.
Generated Knowledge Prompting: Using the model’s ability to generate new content based on the learned information.
Tree of Thoughts: In this approach, the model self-evaluates the progress of intermediate steps through a tree of intermediate thoughts to solve the reasoning
task. This method combines tree search algorithms and tree process
algorithms like backtracking.