GPT Prompting
The more detailed your prompting for ChatGPT the better your result. Here are some of the ones we use for GPT that work well.
You Tube instructors have been great on teaching me prompting. You just need to sift through to the good ones. For me, those are not too technical, but more practical where I can grasp concepts and implement ideas with some additional knowledge as to why I am doing this. My prompts are getting more detailed.
1) I give it a role. "You are an expert in African history specifically in colonization..." Or "You have 20 years of experience ..." or "You are a professional butterfly swimmer..."
2) Then I give it a goal. "I need you to ...."
3) I give it details on the goal and I get pretty granular. If i need an explanation of something complicated, I ask to be spoken to as if I am a 5 year old.
4) At the end of the prompt, I ask Chat GPT to ask me questions for clarification before it generates the answer. This means I can get as much specific information as possible in the response which will be much more tailored to what I am looking for.
5) I learned from a You Tube instructor the importance of these words "step by step" when interacting with ChatGPT. "Give me a step by step response." Those words forces it to really outline with total clarity what I am looking for. It's something referred to as 'zero chain of thought.' When I've tried it, it's definitely added more juice to the response.
We will share more this week, as we have a GPT team call on #prompting best practices.
Interested to learn more?
Sign up today to get notified when the course becomes available on presenting on the future of comms and AI
More from the blog
Revolutionizing Journalism: The Rise of AI in Newsroom Innovation
Some takeaways from the Africa Bloomberg Media Innovators event I attended in Cape Town. Great insights to consider if you're trying to navigate AI.
Trust in God, But Lock Your Car
Eyes wide open. The limitations of Artificial Intelligence that are crucial to know, and how you may be a target for poisoned data
Future of Work in Newsrooms
Invest in fact-checking, Understand AI opportunities and limitations and try new things to ensure data hygiene
RAG, Three Letters and a Big Difference. Huge.
Retrieval-augmented generation (RAG) is an artificial intelligence framework that retrieves data from external sources of knowledge to improve the quality of responses (source: techtarget)