Sparking Creativity

Sparking Creativity
Photo by JD Weiher / Unsplash

The language we use when describing AI assistants often leans towards the dramatic, cold, or dystopian. But this week, I was reminded how generative AI can be used in more fun ways - it can foster out-of-the-box thinking and, dare I say, promote engagement within your community (as it has for me).

I don’t have any notable experiments to share this week, but I did have an experience with ChatGPT that I want to share. It enabled me to do more as a volunteer at my son's school without a huge time commitment. I also had the pleasure of hosting a new weekly AI Research Lunch & Learn. These discussions about AI Research energized me and provided a fantastic way to connect with peers in AI.

Research Insights

The LLM Discussion Framework offers an innovative solution for those frustrated with the creativity of large language models (LLMs). This framework proposes a three-phased discussion among a group of AI agents: brainstorming, discussing, and then converging on an idea. Researchers found that AI agents with diverse roles and expertise were more creative together than independently. The results are fascinating yet intuitive, drawing inspiration from similar Social Sciences research on human creativity.

This research particularly benefits those incorporating AI-powered brainstorming and ideation features into their products. However, given most users currently interact with single-agent systems like ChatGPT, most of us aren’t able to put these findings into practice (yet).

For more details, read: Lu, Li-Chun et al. “LLM Discussion: Enhancing the Creativity of Large Language Models via Discussion Framework and Role-Play.” (2024).

My Experience

Over the years, I've enjoyed decorating display cases at my son's elementary school several times a year. As a crafter, I relish the creative outlet and the opportunity to create displays that foster a sense of belonging for the kids.

Despite my enthusiasm, I've never had the bandwidth to create a calendar and enlist volunteers to help, resulting in underutilized display cases —decorated only a few months of the year.

This year, I volunteered to create a calendar for the display case. I knew right away that ChatGPT could assist with this task, and I was right. In less than 15 minutes, I had a draft calendar ready to share without any modifications needed.

This was a perfect example of a use case where a directional result was sufficient; the formatting and wording didn't need to be perfect, and the exact events and ordering were flexible.

After reviewing the results (which were excellent), I followed up by asking ChatGPT for a supply list and details on which displays each supply would be used for.

Generative AI, in this instance, proved to be an invaluable tool, enabling me to contribute more effectively to my son's school and enhancing my engagement with the community.


Need help implementing your AI strategy? Whether you’re looking for additional engineering resources to build a custom solution or someone to fine-tune a set of prompts to enable your team to do more with the AI tools you already have, Mellonhead can help. Email me or Book a consult.