"Most of us use ChatGPT wrong. We don’t include examples in our prompts. We ignore that we can control ChatGPT’s behavior with roles. We let ChatGPT guess stuff instead of providing it with some information... We need ... high-quality prompts ... [so here's] 4 techniques used in prompt engineering."
There's even a video.
"Few shot standard prompts ... [include] examples of the task". Structure:
"the actual answers in the examples are not important, but the labelspace is... all the possible labels for a given task... providing random labels from the labelspace... instructs the model on how to format the answer to the prompt."
Role Prompting... set a role for ChatGPT... start your prompt with the words “Act as a … ”
"'write [topic] in the style of an expert in [field] with 10+ years of experience'... add descriptors (adjective)... to tweak your prompt. "
"first generate knowledge and only then" ask it to write for you... [eg] 'Generate 5 facts about “AI will not replace humans” '... Use the above facts to write a witty 500-blog post ..."
"chain of thought prompting, the model is induced to produce intermediate reasoning steps before giving the final answer", as forcing it to explain its "reasoning(?) often leads to more accurate results... GPT-4 outperforms old ChatGPT in advanced reasoning capabilities, so you may or may not need Chain of Thought Prompting in GPT-4"
See main image for example, above.
More Stuff I Like