Howell: MindCode project collection
05Mar2024 : [genetic, architecture, function, process, SNN, etc]
Related links to some of my work are provided below. All of this is in very early-stage development even though some of it has been worked on several times since the late 1990's, early 2000s :
What if questions
Dating back to the late 1990s, early 2000s, I had a growing feeling that the mainstream of Artificial Neural Networks (ANN) research had become "conceptually trapped" in a growing range of [simple, powerful, successful] toolsets grounded in [universal function approximation, statistics, information theoretics] which was very good at solving tough problems, but less and less relevant to understanding [biology, neuroscience]. For most researchers, that is fine, as most are probably interested in powerful toolsets, for example, in [math, science, engineer, technical]. Even so, these toolsets seemed to become easy [excuses, impediments] in a competitive research environment to avoid more [fundamental, tough] work.
There is, of course, and enormous community of researchers in [biology, neuroscience] who care very much about understanding natural systems, and a smaller group withing the ANN community who work towards that. Still there seems to be quite a gap
I was particularly interested in how genetics works for [architecture, function, process, "operating systems", ???]. Additionally I was fascinated by ongoing evolution of exciting toolsets such as recurrent nets and kernel systems, and how [specialised, modular] systems might improve understanding (long before "Explainable AI" became a big theme).
My "MindCode" project gradually evolved over the decades, never getting much of my priority time, as I thought others would rapidly fill in the gaps.