Skip to content

Latest commit

 

History

History
23 lines (11 loc) · 1.75 KB

README.md

File metadata and controls

23 lines (11 loc) · 1.75 KB

eve

a neuro-symbolic AI

guile myFile.scm guild compile myFile.scm

it needs to immediately start learning, from scratch

starts with no training data at all; bootstraps and learns from nothing but input/output conversational text

Over the course of several years, I have designed a number of systems for personal knowledge representation - without even knowing what that was. While the things I produced were mostly plain text journal entries, creative works, and wild conjecture, there was always a clear progression of structure, as I've iteratively moved the exact same ideas through several distinct stages of evolution. Each time, representing the same data in new and completely different ways.

It has been so interesting to see all the ways that AI has learned to represent "me" and my friends. The excerpt above... I mean, I could have written it myself! It so-perfectly matches my style, yet was entirely generated by a tiny GPT-Neo model! It blows my mind.

I have decided: the obvious next step is to formalize the symbolic, archtypal structures found within my work. As I've studied the AtomSpace, it's become perfectly clear... this thing was built to control systems of a scale that I can barely imagine. It will take YEARS to learn how to be useful here. Yet, I needed a proof of concept, yesterday.

So, the clear solution is to start with what I'm good at. I will use the AtomSpace to link people, places, and ideas - and I will generate clear, concise documents that represent real-world people in creative stories, just like the one above. A document generator, and a profile management system, with an emphasis on context-switching (for different audiences): now, that seems like an achievable, short-term goal.

Tentatively, I'll call this thing the "Role Model."