Ben's 2006 book is an ok one. Very long though and mostly all over the place and nothing much new (for me), mine I'm making will be very short! I'm actually done now, checked enough of it...
Ben says what I say but in slightly different words: "the goal of achieving personal happiness is so important that inference is pushed along paths that lead to beliefs that increase this goal" Yes, prediction is helped by "favorite words", PPLM and Blender force prediction with Dialogs to drive the model! It is circular reasoning, that you want X so you predict X is the answer. But it does help. You look "around" X. Same for recently activated embeds, saying "cat cat cat cat...." is circular prediction but it does help - if you look "around" the embed space "cat in cage with dog and cat, ran off with cat to dog". Saying recent things is likely the answer. This is re-generative/replicative prediction, it parrots the topic hence cloning itself, but with mutations, it [re-]generates what should be there that goes with it. "I don’t believe any philosophy or any science is going to find the ultimate essence of the universe, break down the universe into its essential components, give an ultimate explanation for why things are the way they are, etc. I don’t believe there is any one big meaning underlying this universe we see all around us" The title of the book, take note! This book was written in 2006, I hope you don't believe that. There's patterns, EVERYWHERE :) ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T90840a08c888a2d2-M0177d3e9c3e1a9c0b38d8eb9 Delivery options: https://agi.topicbox.com/groups/agi/subscription
