Hello, Intelligent Minds!
Welcome to Acting Intelligent, where we explore the fascinating intersection of creativity and artificial intelligence. Today, I, Joseph am excited to take you on a journey through my exploration of curating general intelligence using AI. But fear not, I promise to keep it short and sweet and let you get back to whatever it is you were supposed to be doing.
As creatives, we often find ourselves grappling with the challenge of staying inspired and engaged. General intelligence, as I see it, goes beyond just technology; it’s about our ability to integrate with the complexity of the world and find meaning in the full spectrum of our experiences.
Picture this: you’re in a creative rut, everything seems dull, and boredom is taking over. It was during one such low point that inspiration struck me like a lightning bolt. While idly sitting through a university lecture on mining large-scale data, a seemingly unrelated concept triggered memories of a book I had recently devoured – “Flow” by Mihaly Csikszentmihalyi. The book explores the state of optimal experience, the sweet spot between difficulty and ease, where creativity flourishes. This moment sparked an idea – to leverage the principles of flow and the knowledge from the lecture to build something that could pull me out of creative slumps.
Now, for those unfamiliar with my work, I thrive on collaboration. Why embark on a fun project alone when you can share it with someone? With this philosophy in mind, I enlisted a friend to join me on this endeavor. Our goal was clear: create a piece of technology that could inspire and engage us in interesting activities, pulling us out of those creative ruts.
The foundation of our project was a simple User Experience (UX) – an interface where users could answer a few questions and receive personalized suggestions for hobbies, sports, or activities. The key was to suggest something new and intriguing or something they had been contemplating but never pursued. Success, in this case, would mean breaking free from creative stagnation.
But where do you get the data to fuel such a project? Enter Kaggle, with its database of 384 hobbies. While the information was limited, providing only titles and classifications, generative AI, specifically OpenAI’s GPT-3, came to the rescue. I expanded on the existing data by prompting detailed technical descriptions of each hobby, creating a rich dataset.
However, challenges awaited, particularly in generating JSON-formatted data for each hobby. Validity, repeatability, and standardization were the hurdles we faced. Yet, these challenges were steps toward enriching the dataset, forming part of an ongoing process.
To identify suitable activities for users, we employed a mixed approach. Using the long-form description and select parts of the JSON data, we calculated the cosine distance between documents to determine their similarity. The goal was to suggest activities similar to those enjoyed in the past.
The project didn’t stop there. We continued expanding the list of activities, incorporating both generated data and allowing users to manually input their enjoyed activities. This dynamic process led to a constantly growing dataset, currently boasting around 560 hobbies, activates and sports.
As I move forward, the journey involves automating the pipeline, refining user experience questions, and overcoming challenges like cleaning JSON data. While these are ongoing tasks, the progress has been significant, guided by the pursuit of flow rather than a rigid endpoint.
In closing, let me leave you with a thought from Csikszentmihalyi: “It is not the skills we actually have that determine how we feel, but the ones we think we have.” My hope is that this project not only helps me, as it has already, but also assists others in discovering new and exciting projects and hobbies.
Stay Inspired, Joseph