Probably one of the most important things with AIs is the prompts.
If your prompts are good then your results will be good. If your prompts are wrong your results will be wrong.
To make your prompts better apply this list to your prompt workflow.
-
Clear Instructions: Be precise. If you need short answers, specify. If the responses are too simple, request expert-level detail. Show the format you prefer.
-
Use Details: Providing more context can help get more relevant responses.
-
Adopt Personas: Asking the model to assume a specific character or style can tailor the responses more closely to your needs.
-
Delimiters and Formats: Use clear markers to differentiate parts of your input, or specify a format for the model to follow.
-
Reference Texts: Supplying text for the model to reference can lead to more accurate and less fabricated answers.
-
Decompose Complex Tasks: Break down intricate tasks into simpler ones for more accurate outcomes.
-
Reasoning Time: Allow the model to ‘think’ through a problem before responding, which can lead to more reliable answers.
-
External Tools: Supplement the model’s capabilities with tools for tasks like searching for information or executing code.
-
Test Systematically: To confirm improvements, evaluate changes against a broad set of examples.
Important TIP:
good results not just depends on your prompts but it depends on your model too. If your model is not Intelligent or big enough doesn’t matter how good your prompt is your results will be bad every single time.
Personally I would recommend using at least GPT3,5 level intelligence model.