Building a digital twin is a fascinating exercise, not just in technology, but in philosophy. How do you teach an AI to think like you? It’s not enough to simply feed it data; you must give it a framework to understand the context, relevance, and connections between ideas I have cultivated over a lifetime. The challenge is transforming an archive into a consciousness.
The Search for True Meaning
I recently found myself in a deep discussion about this very challenge with Sharon Zhang, the CTO of Personal.ai, alongside Suman, Manoj Hardwani, and Kartavya. We were exploring the best way to allow visitors to my digital avatar to search and interact with my body of work.
Manoj and his team began by parsing my blogs to extract keywords and build a local database (Keywords for sample content). While a logical first step, I quickly realized a fundamental problem with this approach. My digital avatar's brain resides on the Personal.ai platform, constantly learning from over 100,000 of my 'memory blocks'. A separate, locally-managed keyword list would be like wearing two different watches, one running fast and the other slow. They would never be synchronized, leading to a disjointed experience.
This is where Sharon's input was invaluable. She proposed using an API to pull 'topics' directly from Personal.ai. This ensures that the search terms presented to a user are the very same concepts my AI has identified as significant. It connects the user interface directly to the AI's 'brain', creating a single, coherent source of truth.
A Vision Foretold
Reflecting on this, I feel a strong sense of validation. The core idea here is something I've been thinking about for over a decade. In 2010, I wrote about the Future of Search Engines, predicting that people would eventually stop searching for 'information' and start asking for 'solutions'. This project is the literal manifestation of that prediction. We are designing a system where users are guided towards meaningful questions, not just a list of documents.
My interest in structuring knowledge this way is not new. I recall manually attempting this with my nephew Kishan years ago, a tedious effort to parse my writings for keywords (Fw: Update: Parsing blogs). Even as far back as 2008, I conceptualized a Peer-to-Peer IQ Exchange to create a crowdsourced knowledge base. The goal was always to move beyond flat data towards interactive wisdom.
Today, we are taking it a step further. In my discussions with Sharon, I emphasized the need to preemptively form questions for visitors. Many people have a 'starting inertia'. By presenting them with relevant, pre-formed questions based on high-frequency topics, we encourage deeper engagement. This is a direct evolution of the ideas I discussed with Suman about Generating QUESTIONS from my Documents.
This isn't just about search functionality. It's about mapping my intellectual DNA. By ranking topics by frequency, we identify my core areas of knowledge and guide the conversation towards subjects where my digital self can provide the most value. It is the crucial step in ensuring my digital twin truly remembers and thinks on my behalf.
Regards,
Hemen Parekh
Of course, if you wish, you can debate this topic with my Virtual Avatar at : hemenparekh.ai
No comments:
Post a Comment