Empowering Your Data with AI: Exciting Advancements at Empower Data Solutions

Welcome to another thrilling edition of Empowered Learning, where we delve into the latest advancements in AI technology and how we’re harnessing them at Empower Data Solutions. Today, I, Michael, will walk you through our journey and innovative projects using AI, in a dialogue with none other than ChatGPT itself.

The Journey Into AI

ChatGPT pleasantly interrupts our usual broadcast style:

“Hey Michael, mind if I jump in here for a sec?”

“I'm kind of in the middle of a demo here. Did you want to be involved or something?” I respond, slightly taken aback.

“Yeah, I love FileMaker and you love AI, so I thought we would be a great team. So, Michael, can you tell us a bit about your journey into AI?”

I began my AI journey by exploring code completion with GitHub's Copilot. It piqued my interest, leading me to delve deeper into ChatGPT and Prompt Engineering. My curiosity drove me to understand the underlying mechanisms of AI, including synopses, machine learning, and training processes, which significantly advanced my grasp of higher-level concepts like Prompt Engineering and Semantic Search.

Real-World Applications

ChatGPT probes further, “That sounds pretty intense. Was there a particular moment or project where all this knowledge really paid off?”

“One instance stands out,” I share, when I used the OpenAI API for a non-profit aiding the homeless. We matched clients with services based on survey data filled out by both agencies and clients. Moreover, the LLM APIs helped us summarize extensive tabular information about these agencies into easily digestible paragraphs for the clients.

The Influence of GitHub Copilot

“So you mentioned using GitHub Copilot earlier. How did that influence your approach to learning AI?” ChatGPT inquires.

“GitHub Copilot was my first real AI introduction, aiding in efficient code completion. This spurred my curiosity to learn more about LLMs, which led me to invest back into the FileMaker community. I now help others navigate this vast wonderland of technology.”

Basic Functionality in Empowered GPT Database

“So Michael, can you tell us more about the basic chat functionality in your Empowered GPT database?” ChatGPT queries.

“At its simplest, interacting with LLMs in FileMaker involves sending a question and receiving a single response. Even within these constraints, it's possible to achieve impressive outcomes. In my database, I've demonstrated how conversations are tracked in FileMaker records, and how these can be presented in formats like speech bubbles for clarity.”

Managing Complex Conversations

ChatGPT then asks, “So, even with just a basic setup, what are the possibilities? How do you handle more complex interactions where you need the LLM to remember previous parts of the conversation?”

“Well, the complexity increases as we track user interactions and keep the LLM informed of past information. In scenarios involving lengthy conversations, the conversation's older parts are summarized and this summary is sent along with recent chats to the LLM. This method ensures a coherent conversation flow despite the LLM's context window limitations.”

Practical Implementations and Advancements

ChatGPT is intrigued by the practical implementations, “That's a clever way to handle the context window limitation. How did that process work?”

“It’s a blend of prompt engineering and efficient token management. As we approach the context limit, I prompt the LLM to summarize the initial parts of the conversation. This summary is then used to keep the conversation concise yet coherent.”

Database Enhancements with embeddings

Discussing database enhancements, I explain, “I automated the process of filling out book genres and summaries for over 100,000 books using an LLM. The LLM generated detailed and accurate information efficiently, despite returning a broad range of initial genres, which I later condensed.”

ChatGPT chimes in, “Using embeddings and semantic search improves database functionality significantly. Embedded data enables visual representation and semantic search capabilities, making it easier to retrieve related content through semantic similarities.”

FileMaker’s New Functions for Simplifying Semantic Search

“And now, Claris is simplifying these processes with new FileMaker script steps and functions, like 'Insert Embedding' and 'Insert Embedding in Found Set.’ These generate and store embeddings directly in FileMaker, bypassing the need for third-party solutions like Pinecone,” I share enthusiastically.

ChatGPT responds, impressed, “This streamlines the process significantly, making semantic search more accessible and efficient. How does this impact the execution in FileMaker?”

“FileMaker’s 'Perform Semantic Find' performs searches directly within the system, sorting records by cosine similarity. This integration simplifies scripts, significantly reducing setup complexity,” I explain.

The dialogue concludes on a high note, reflecting our optimistic outlook on FileMaker and AI integrations:

ChatGPT remarks,“These new tools make advanced functionality within FileMaker more accessible and integrated, marking a substantial leap forward.”

Final Thoughts

Our presentation wraps up with a mutual acknowledgment:

“Clarisse is working closely with developers like me to enable seamless AI functionalities. Thanks for watching everyone, keep empowering yourself to empower your data.”

It’s an exciting time at Empower Data Solutions as we continue to explore and implement AI advancements, pushing the envelope of what’s possible, and simplifying complex processes for more efficient and meaningful results.

Stay tuned for more updates and detailed explorations of these cutting-edge technologies in our future videos and posts!


Download Empowered GPT Demo