My own personal AI learning journey started with a fundamental shift of focus. Initially, I was stuck with debating whether I should use ChatGPT, Gemini or Copilot (something I feel many of you can relate to).
Further along though, I realized I was focusing on the wrong thing - particularly as a beginner. Regardless of the platform, the real driving force behind the success of GenAI adoption wasn’t the tool, it was the quality of prompting.
So, if you’re new to GenAI (just like me not too long ago). Drop the product debate, pick whichever you have access to and start practicing on prompting. To help you in that journey, here are 3 lessons I wish I picked up sooner.
Tip #1: Prompts are not the same as Google searches
Do you typically write prompts as concisely written as: "write me an email draft ..." or "find the best ..." only to end up with sub par results? I have. So, sooo many times.
It’s a personal habit from my own Google searches. If you check my history, it’s a long list of random mix of keywords pretending to be a sentence (e.g. John Mayer ooh oohh ohhh song 2000s) OR I write as I think/speak, which isn’t any improvement from the latter.
For Google searches, a haphazard query is fine. In fact, we often intentionally exclude specifics to broaden our results. In GenAI, however, this practice can lead to frustrating results.
Subconsciously, our prompts typically come with a hyperspecific expectation. We know what we want, how we want it to look like, and we expect nothing less from GenAI. For example, if I want ChatGPT to write a draft email, I have a specific tone and length in mind. If I want a list of cafes from Gemini, I have specific considerations (e.g. what season that matcha batch was harvested). The question is, am I effectively communicating these expectations in my prompts? Honestly speaking, I typically do not.
In GenAI, details are key. Don’t rush the prompt. Instead, take a minute to collect and organize your thoughts. Doing so will go a long way in getting better results (or at least better first iterations of it).
Need help in organizing your thoughts into prompts? That leads us to ...
Tip #2: Follow a Prompt Formulae
Transforming thoughts/needs into a well-structured prompt is not as easy as it sounds. At least in the beginning. But what significantly helped was the use of prompt formulae. Don’t worry, it’s not a math equation!
Prompt formulae really are just reminders. It’s a list of categories of information (typically in a neat abbreviation) that serves as a simple reminder about what info you need to include in a prompt. Think Google’s: Persona, Task, Format & Context (PTFC) framework here.
In line with the first tip, using a prompt formula allowed me to better organize my thoughts. By providing you a clear framework to work with, writing a prompt slowly became as simple as filling out a form. It’s an effective guideline to follow, particularly if you’re still getting started with GenAI.
There are a ton of prompt engineering frameworks and formulae, such as:
Following prompt formulae has led to me getting better results more consistently and I’m quite sure it will do the same for you. Will it give you the best result right out of the gate? Not necessarily BUT this leads us to the third and final tip ...
Tip #3: It’s A Work In Progress
Using GenAI will always be an interactive process. While the first two tips will get you as close to your desired outcome, let’s simply be realistic, there will always be room to improve it.
My intention for my 3rd tip also deals with my personal point of view on the larger existential debate as to whether AI would replace us humans or not (quite a tangent to go on but hear me out). My take is, it definitely will - for some tasks, that is. But, such is the nature of technology. It’s meant to free us up so we can focus on other bigger matters.
So for my last tip, it’s simply for you to think of whatever result you get as a first draft.
Whether you’ve tasked ChatGPT to write an email, Copilot to analyze a spreadsheet, or Gemini to list locations of nearby cafes, these are really just first iterations. It’s up to you to further refine that by either working on it on your own, or my personal preference, to learn how to co-author/collaborate with GenAI.
How do you do collaborate further with GenAI? Here are a few tips:
-
Share your feedback on results - There’s a reason why there’s a thumbs up/down button for every result. Use that to communicate back what went right/wrong so as to better train your GenAI collaborator. You won’t hurt its feelings (I hope so).
-
Provide follow-up prompts - Not quite satisfied? Revisit your prompt formula and see what went wrong, what info to include or exclude. It’s a great activity as well that typically leads to improving future prompts!
-
Use Canvas - Finally, particularly useful for written outputs (e.g. written drafts), use the Canvas tool (available in Gemini & ChatGPT). This allows you to use GenAI as a proofreader/editor. Offers you options to recommend suggestions, edits or even change the tone and length of your content in an instant.
Overall, these were some of the tips I wished I learned sooner rather than later. I’m sharing it here with the intention of hopefully helping you speed up your own AI learning journey.
Want to continue your AI learning journey?
Join the Asia-Pacific AI Nonprofit Learning Community, a learning initiative led by Infoxchange in collaboration with TechSoup and its regional partners to help educate nonprofits on how to use AI for good. Join the community today and get access to free webinars and an online forum where you can connect with AI experts and other nonprofits to share ideas and ask questions.
Join the APAC AI Nonprofit Learning Community here.