Do you have an interest in using ChatGPT? Have you heard of Prompt Engineering, but not really sure what it means or how to apply it to ChatGPT? If so, then this blog post is for you! In this article, we'll explore the concept of Prompt Engineering and why it matters for those looking to leverage ChatGPT. We'll also provide guidance on best practices to ensure users get the most out of their interactions. Join us as we break down how prompt engineering can supercharge your use of ChatGPT today!
ChatGPT can either be incredibly smart or not so much. When asking it questions or using it to help you with tasks prompt with being an expert in whatever you think applies. Throughout this blog I will use examples the whole way through so you can see each step in practical use. When I first start a conversation with ChatGPT I prompt it as being an “expert”. In this case I told ChatGPT “You are a Marketing Expert, an SEO Expert, and a Content Writing Specialist.” Now obviously this varies based on your needs but what I have done here is shifted the machine’s focus and narrowed it down to be proficient in what I needed it to become.
A feature that many don’t include while prompting is giving sources for the AI to work with. Now we still have to keep in mind that ChatGPT is only up to 2021 with information so you can’t use “new” sources. However, there are extensions coming for some users and some other 3rd party options that can read current webpages. So, depending on how agile your field is you can provide some level of baseline sources. These sources paired with the expertise step above allows users to get some seriously accurate answers.
Sources also don’t need to be “physical” like books, webpages, etc. Sources can also be classified as writing types, food cuisine, etc. When trying to, for example write a short story, a source can be giving the prompt of “in a Dr. Suess style” or “like Stephen King”. This gives the AI another way to source how it will write the short story. Sources much like making it an “expert” give more focus to the AI which in turn gives more precise answers and less room for its “thoughts” to drift.
While ChatGPT is a revolutionary tool in countless ways you should still treat it somewhat like a child. What I mean is you don’t ask children a hyper complex question and expect to get an ironclad answer. The same is said for ChatGPT. Your question(s) should be broken down into smaller fragments, remember it is a conversational AI. Lead the AI down the proper path of what you want it to achieve.
Giving overly complex questions to the AI can cause some issues like it only answers one part of the question, doesn’t answer the question, etc. So wait until part one of your question is answered the way you want/need it to be but then the next part should include something like “referring to the previous question…..”. This helps the AI connect dots between you beginning and ending questions to guide the AI.
When it comes to asking questions, I don’t mean the questions you’re trying to get the answer too. Let’s say it gave a half correct answer to your original question or even missed the mark entirely. What you could do is ask the AI things like “How can I improve this prompt?”, “What if I change out X for Y?”, Asking questions like help push the AI towards a proper path or can even help you improve your existing prompt. Again, remember this is a conversational AI so use the token limits to its fullest to get exactly what you need.
A great question to ask the AI if you aren’t getting the results, you want is twofold. One is “How would you rate this prompt?”. Depending on the answer you can then ask the question I listed above, “How can I improve this prompt?”. These two questions not only train the AI through your conversation but also train you as well on how to prompt for future inquiries you may have.
The average user, most likely yourself, knows what ChatGPT is but I would bet most of you are missing out on what truly drives the AI. OpenAI is extremely community driven and the ultimate place that it culminates is OpenAI’s Discord. For those of you that may not know what Discord is in short, it’s a social platform where users can come together in groups and share, post, IM, etc. The group is run by OpenAI so you can converse directly with developers there and see other things users have posted. There are channels dedicated to Prompts and Prompt Engineering where you can see prompts others have created or if you need help can post your questions there. Here is the link to join the group: https://discord.gg/openai If you don’t know how to use Discord, I would strongly suggest learning at least the basics so you can navigate through the channels. It can be one of the ultimate “how-to” places for you to learn how to masterfully use ChatGPT, Dall-E, the API, etc.