Photo by cottonbro studio on Pexels

An independent developer claims to have made significant strides in the pursuit of Artificial General Intelligence (AGI) with a symbolic kernel, currently undergoing debugging. The project, reportedly completed in just 38 days, boasts a symbolic natural language processing (NLP) system that eschews neural networks, coupled with live knowledge retrieval, a suite of twenty reasoning types, and real-time learning, all governed by a strict ontology. Notably, the system leverages Google’s Gemini model as an ‘oracle’ for information, but is designed not to directly utilize Gemini’s generated text as the basis for its reasoning. The developer acknowledges ongoing challenges with NLP accuracy, but emphasizes the system’s adaptive capabilities, including the ability to dynamically adjust its datasets and knowledge base. While still in its nascent stages and under active development, the project represents a potentially novel approach to AGI development. The original discussion surrounding the project can be found on Reddit. [Reddit Post: https://old.reddit.com/r/artificial/comments/1lvr6px/p_symbolic_artifical_general_intelligence_kernel/]