Generative AI has been marketed as a magic wand able to do your thinking for you. In a commercial from Google Gemini, a father asks the chatbot to write an inspiring letter to give to his daughter’s hero, Olympic hurdler Sydney McLaughlin-Levrone. This scenario sparks an interesting conversation about the role of AI in our most sincere moments and the use cases presented in many AI commercials. For those who enjoy the act of writing, it also raises the question of how a tool like this can be used without losing the personal touch.
While it’s possible that some of us would rather automate all their thinking, I started to wonder if this was really the best way to use GenAI. From my experience using the Generative AI @ Pitt tools, it’s far more effective when you (a real person!) are in the driver’s seat doing critical and creative thinking for yourself. Today’s tools simply can’t access your real-world know-how, subject matter expertise, or dot-connecting abilities. Those are the traits that set you apart — even in the AI era.
Better Prompts, Better Output
Learning how to write better prompts is one of the main ways to improve your GenAI experience. There are plenty of resources that can help improve your prompts (some of the best are available for free with your Pitt credentials on LinkedIn Learning), but one of my favorite methods is to treat your chats like a conversation. Think of your chats like collaborating with a smart, always available coworker. GenAI can help you to work faster, but the ultimate responsibility for the quality, creativity, and accuracy of the work — and the name attached to it — is yours alone.
It’s easy to start a chat with a prompt like,“Create a first draft of a blog about how to guide GenAI tools to better output.” When I gave this prompt to Google Gemini, I got a reasonable, though very generic and incredibly lengthy, first draft. If you’re just starting to use GenAI, the speed and coherence of this interaction can be mind-bending, but as a regular GenAI user (as well as an experienced blog writer), I wasn’t so impressed.
Instead of asking GenAI for the end result, start at the beginning by first identifying the ways you could use help with your task. Need to brainstorm? Provide your initial thoughts and ask for help refining them. In research mode? Give it your assumptions and see where your knowledge gaps are. Already have a document drafted? Ask it to point out the strongest and weakest points.
If I was speaking with a colleague, I might ask them for help sorting through my early ideas or whether the direction I chose is the right one, so that’s exactly how I would approach a conversation with GenAI.
Here’s an example of a prompt that I could have used to get started on this blog:
“I’m writing a blog about how to guide GenAI tools to better output intended for a University community of students, faculty, and staff. I’m thinking of including information about how to write better prompts as well as some information about how GenAI works. The goal is to educate readers about how to more effectively and intelligently use GenAI tools. What related topics might my audience find helpful to discuss in this blog? Please provide 10 to get started.”
This prompt is doing a lot of heavy lifting, and while writing it, I was thinking critically about what I already knew and what I needed to find out. This makes the work of writing the prompt part of the creative process, rather than cutting out my own creative thinking. As I wrote it, I started to generate ideas and determine what I thought would work based on my prior knowledge and experience. All this crucial brainstorming work happened before I hit “send.”
When I receive a response from GenAI, I ask myself if I agree or disagree, and if so, why? By being thoughtful about what I apply from its responses (and not just copy and pasting), I continue to think critically and open the door to new ways of thinking. The way you reflect on a response can sometimes be more valuable than the response itself.
GenAI Finds the Ordinary; You Find the Extraordinary
GenAI is essentially really advanced predictive text. What seems like “thinking” is the tool selecting the next most probable word in response to your prompt based on everything it knows about language (which is a lot). This is undoubtedly impressive technology, but by its very nature, it’s best at finding what’s average. For most of us, the hard work is figuring out how to go beyond the obvious to make new connections.
GenAI’s predictive nature is also why it needs to be fact checked. Because it’s predicting, it’s not really thinking, feeling, or deploying logic. Whether or not the next word is true doesn’t matter to a GenAI tool. All that matters is whether it’s the most probable.
Once you understand GenAI’s predictive powers, it’s easy to see why a specific, tailored, and targeted prompt will be better than a generic one. The more information you can provide GenAI, and the more zoomed in your request, the more useful the response will be. GenAI might seem like magic, but it’s not a mind reader. It’s working with a lot of data; you’re the one who will provide focus.
Brain Power Makes the Difference
It’s clear that it takes a lot of work to use GenAI effectively, which isn’t a bad thing. Rather than viewing GenAI as simply a generator of end results, I find its real power is in its ability to spur new ideas through intelligent conversation and collaboration. That means you can’t just shut your brain off when using GenAI tools. In fact, if you want to most effectively use GenAI, you should think more about the things that encourage you to grow your knowledge and perspective.
— Rachel Bachy, Pitt Digital
In the spirit of applying the tools we provide, Google Gemini was used to brainstorm and debate initial ideas, as well as to provide the example first draft mentioned in the article.
