For those of you interested there are two presentations this Sunday on Artificial General Intelligence (AGI) by Monica Anderson and Stefan Reich. Monica is someone that I am well acquainted with and I an administrator for her Facebook group on Model-Free Methods for Artificial General Intelligence (MFM-AGI). She is an amazingly intelligent and sharp researcher with decades of experience and spent many, many years programming in Lisp. The presentations are this upcoming Sunday. If you plan on attending I strongly encourage you to read Monica’s blog on “Artificial Understanding” beforehand:
"This event will feature two independent AGI developers passionately laboring on their own separate projects to produce their AGI. They are not developing under the auspices of academia
nor big tech. We are pleased to welcome Monica Anderson, Director of Research at Syntience Inc., and Stefan Reich of the Gazelle project.
Epistemology is the branch of philosophy that discusses matters such as truth and falsehood, sources of knowledge, information versus noise, problem solving methods and their limitations, and processes such as learning, abstraction, understanding, reasoning, and saliency (knowing what matters).
She will be using vocabulary explained on the site, so reading this ahead of the session would be useful.
These insights have been used to create Understanding Machines that learn any human language simply by reading un-annotated text, and using a million times less energy than GPT/3 or BERT.
Monica spent the past 20 years (except when working for Google) on developing a machine that can learn any language, including Japanese, DNA, or MIDI. A prototype of Syntience Understanding
Machine One (UM1) is available in the cloud and API example and test Python code is available by request.
Stefan's project, Gazelle, deliberately runs counter to the sub-symbolic trend in AI/AGI:
The Gazelle system implements the most recent advances within the symbolism approach to AI. Gazelle builds on the ideas developed in groundbreaking symbolist projects like Cyc -- while addressing their shortcomings.
Gazelle aims to prove that a symbolist AI is capable of delivering conversational AI so advanced it can reliably be used as a replacement for traditional user interfaces. Gazelle has been in development for 7+ years and recently had a development sprint with a 4-person team.
Gazelle is based on the art of pattern matching. Texts and patterns are the primary data types. Everything is a first-class citizen in the Gazelle database and connected to everything
else, including inputs, patterns, procedures, mathematical operations, input transformers, evaluations, logs, meta-procedures, assumptions and statements (both about the world and about the system itself). This heterogeneity and connectedness is the key to
how the database works and how it begins to self-organize step by step.