Introducing UnityGPT

UnityGPT is an AI tool designed to interact with unconventional research. Unlike big AI models which are trained primarily on mainstream internet content, UnityGPT is trained on extensive research in fields such as consciousness, physics, UFOs/UAPs, frontier energy, ancient archaeology, and alternative healing modalities, to name a few. It is also provided with canonical texts from the world's wisdom traditions.

The primary goal of UnityGPT is to assist researchers in finding relationships between these distinct subjects, enabling the development of new theories and facilitating a deeper, more interconnected understanding of our world.

Unity Model

At the core of UnityGPT is a vast volume of technical publications primarily from established peer-reviewed journals. These include technical articles from the Society for Scientific Exploration, the Institute of Noetic Sciences, and the Resonance Science Foundation, as well as scholarly books on a variety of subjects. There are roughly 80,000 embeddings across several thousand texts.

Hidden Connections

Each text has been chunked, converted to high-dimensional vector space, stored in a single index, and mapped according to cosine similarity. This approach extends beyond simple keyword recognition, enabling UnityGPT to grasp and reveal the deeper meaning, similarities, and associations within the text. A large language model enables NLP engagement with the embeddings.

I believe unification requires us to pay particular attention to persistent anomalies, and to venture beyond our domain silos, seeking insights and inspirations from seemingly unrelated areas. I hope UnityGPT can help.

Roadmap

This is an early beta. The near-term roadmap includes full citations for each response, and model improvements with larger training sets. I’ll post updates on twitter.