On Tuesday, May 21, 2024, at 10:34 PM, Rob Freeman wrote: > Unless I've missed something in that presentation. Is there anywhere in the hour long presentation where they address a decoupling of category from pattern, and the implications of this for novelty of structure?
I didn’t watch the video but isn’t this just morphisms and functors so you can map ML between knowledge domains. Some may need to be fuzzy and the best structure I’ve found is Smarandache’s neutrosphic...So a generalized intelligence will manage sets of various morphisms across N domains. For example, if an AI that knows how to drive a car attempts to build a birdhouse it takes a small subset of morphisms between the two but grows more towards the birdhouse. As it attempts to build the birdhouse there actually may be some morphismic structure that apply to driving a car but most will be utilized and grow one way… N morphisms for example epi, mono, homo, homeo, endo, auto, zero, etc. and most obvious iso. Another mapping from car driving to motorcycle driving would have more utilizable morphisms… like steering wheel to handlebars… there is some symmetry mapping between group operations but they are not full iso. The pattern recognition is morphism recognition and novelty is created from mathematical structure manipulation across knowledge domains. This works very well when building new molecules since there are tight, almost lossless IOW iso morphismic relationships. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T682a307a763c1ced-Me455a509be8e5e3671c3b5e0 Delivery options: https://agi.topicbox.com/groups/agi/subscription
