My core fundamental is my AGI design uses just if-then rules by storing data with order in 1D / 2D / etc form. Ex. cat>ran. Sound>Vision. Pixel1>Pixel2/3/4 These sequences/ syntactics form all other rules, like translation, categories, and all the many rarer ones like these 2 questions: "scramble the 1st word: ?" (craembsl) and "hat key helmet, hoof mad paw, loop nice hole, book toy study, men moon woman, cat home ?" (dog).
You start the AGI off with a few of these pattern finders, it clusters matches in a network. It's all matching and clustering. The many other rules/patterns we won't/can't code are found by letting this AGI think, it can store new memories that let it carry them out. If it finds the next rarer one that is not so rare but not hardwired, it can modify its code to run that AGI component with its other components. Theirs/mine doesn't use backprop. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tcf8fbceae6bd0acd-M09969feba8d4bf0dce2a8fd9 Delivery options: https://agi.topicbox.com/groups/agi/subscription
