The last time I tried to use generative AI in earnest—specifically ChatGPT and Microsoft Copilot—I wrote a long and detailed prompt asking for a table with 20 items. One came back with 23 items. Part of the instructions was to gather information about best-selling memoirs, but both chatbots mixed in novels. While many people complain about AI hallucinations, I’m still waiting for it to follow basic instructions. Suffice it to say that my experiences have been underwhelming.
As a technologist, however, it’s my duty to revisit certain technologies every so often to give them another chance. That’s what happened recently after I watched a MasterClass series on AI. Upon listening to one good example after another, I realized how I could make AI more useful to my needs.
1. Ask Questions That Have No Wrong Answers
Many AI proponents say that brainstorming is a good use case for AI. Ask ChatGPT, Copilot, or Google Gemini for ideas about…well, anything. Here’s an example:
What are some interesting ideas for a birthday party theme for an eight-year-old that’s whimsical but doesn’t involve anything scary or violent, like dragons or swords? Include some suggestions for the decor, a birthday cake, and entertainment. It all must be appropriate for kids ages six to ten.
If there are no wrong answers, which should be the case when you’re brainstorming, then these sorts of open-ended questions tend to work well.
Just keep in mind that this is just the start. If it’s a topic that has been written about extensively online, like kids’ birthday party themes or creative ideas for gifts that don’t cost money, you might get technically correct or broadly acceptable answers, but not ones that are particularly satisfying.
2. Follow Up About Specifics
To drill down into what’s actually useful, take a page from improv comedy and follow up with a “yes, and…”
If a chatbot gives you mostly hackneyed ideas related to the first prompt, find one that isn’t terrible and use it for your next prompt. Ask for new information about it or a revision. You’re more likely to stumble upon something slightly more interesting or that at least fits your criteria better.
If you don’t do this, you effectively have to start over. Even if you spend more time crafting a better and more specific prompt, you’re not necessarily going to get more creative ideas. So, you’re putting yourself in a loop of disappointment.
In the example from above, Copilot suggested an outer space-themed birthday party with an activity of making “moon sand” and a space-themed obstacle course. These aren’t the most brilliant and creative ideas I’ve ever heard, but they’re not bad, either. So, here’s a way I might “yes, and” my way to something usable:
Recommended by Our Editors
I like the idea of an outer space-themed party, but I don’t know what moon sand is or how we would make it at a birthday party. Can you explain in detail how it would work? I also like the idea of a space-themed obstacle course, but I want to make sure I can build it using things I already have in my home. What kind of obstacle course might you suggest we could make for eight-year-olds inside my house using the space theme?
Finally, the AI gave me very specific ideas. Of the seven suggestions for an “obstacle course,” one sounded like a good fit, and the moon sand activity requires only two inexpensive ingredients (flour and oil).
Understand the Limitations of AI
Advocates of generative AI for personal and business use always say that for it to be beneficial, you have to understand what it can and can’t do. You also have to know the right language to use when talking to it.
I still disagree with several of the things generative AI supposedly can do, seeing as it doesn’t know how to count to 20 reliably or distinguish a work of fiction from a memoir. But if you ask it a question that has no wrong answers and you give it the old “yes, and…” treatment, you just might end up with something compelling.
Get Our Best Stories!
Sign up for What’s New Now to get our top stories delivered to your inbox every morning.
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.
About Jill Duffy
Contributor
I’ve been contributing to PCMag since 2011 in a variety of ways. My column, Get Organized, has been running on PCMag since 2012. It gives advice on how to manage all the devices, apps, digital photos, email, and other technology that can make you feel like you’re going to have a panic attack.
My latest book is The Everything Guide to Remote Work, which goes into great detail about a subject that I’ve been covering as a writer and participating in personally since well before the COVID-19 pandemic.
I write about work culture, personal productivity, and software, including project management software, collaboration apps, productivity apps, and language-learning software.
Previously, I worked for the Association for Computing Machinery, The San Francisco Examiner newspaper, Game Developer magazine, and (I kid you not) The Journal of Chemical Physics. I was once profiled in an article in Vogue India alongside Marie Kondo. I’m currently pursuing a few unannounced long-form projects.
Follow me on Mastodon.
Read Jill’s full bio
