top of page

Asking effective questions to ChatGPT or other chatbots can greatly enhance the quality of the responses you receive. Here's a guide on how to frame questions to get the most out of ChatGPT-4:

1. Be Specific: The more specific you are, the more targeted the answer will be.
   - Vague: "Tell me about trees."
   - Specific: "What are the differences between deciduous and coniferous trees?"

2. Use Proper Context: Giving a bit of background or context can help the model understand the essence of your question.
   - Without Context: "What's the treatment?"
   - With Context: "I've been diagnosed with mild hypertension. What are the common  treatments?"

3. Break It Down: Ask One Question at a Time:  If you have a multi-part question, consider breaking it down into individual questions or explicitly numbering the parts. This helps in ensuring that each query gets the attention it deserves. 
   - Instead of: "Tell me about the history, production process, and benefits of green tea."
   - Ask: "1) What's the history of green tea? 2) How is green tea produced? 3) What are the health benefits of drinking green tea?"

4. Clarify the Desired Format: If you want information in a specific format, mention that.
    - "Can you list the benefits of exercise?" versus "Can you write a short paragraph on the benefits of exercise?" 

5. Refrain from Overly Broad Questions: These can lead to generic answers.
   - Broad: "Tell me about the universe."
   - Focused: "What is the significance of the cosmic microwave background radiation in understanding the universe?"

6. Ask Open-ended Questions for Exploration:  If you're looking to explore a topic or stimulate thought, ask open-ended questions.   

- "What are the philosophical implications of AI?"

7. Limit Instructions When Necessary: GPT models follow instructions. If you want the model to think or respond in a certain way, guide it.
   - "Can you explain that in layman's terms?"
   - "Provide a beginner's guide to..."

8. Feedback Loop: If an answer is close to what you wanted but not quite there, give feedback and ask for clarification or further detail.

 - "That's not quite what I was looking for. Can you explain in simpler terms?"

9. Avoid Leading Questions: If you want an unbiased answer, try not to frame your question in a leading way.
   - Leading: "Why is X clearly the best method?"
   - Neutral: "What are the advantages and disadvantages of X method?"

10. Ask for Sources or Basis: If you're interested in where the information is coming from, you can ask things like

"What is the source of this information?" or "On what basis are you making this claim?"

11. Acknowledge Limitations: Remember that ChatGPT-4's training only goes up to September 2021. It won't have knowledge on events or developments after that time.

Remember that no tool is perfect. GPT-4, like its predecessors, has its limitations, but by asking good questions, you can optimize the responses you get.

*I've been playing with Chat GPT-4 a lot recently, and this article was a good exercise in generating and editing. I generated a few different versions of the prompt. (Basically the title of the article.) Some list items were missing examples so I had to get more specific to generate those. Some examples and list item phrasing I liked better than previous versions, so I did a lot of cutting and pasting. 

 

ChatGPT does not play well with others when it comes to the issue of transferring generated answers via their clipboard or keyboard commands.  Lots of time was spent cleaning up the font, spacing, bolding, italicizing, renumbering, etc. to make it easier to read. I find editing much easier than creating, so I didn't mind taking on that task. Looking at the future, eventually the tools will get more sophisticated and the users will get better at using AI tools. 

 

What will be the tipping point for the widespread adoption of generative A.I.? 

What systems and safeguards need to be in place for users to get to this point?

bottom of page