I know the above seems like low level is too low but really the 3rd paragraph above is the actual how AI works, the beneath that is the details it runs, the 2 paragraphs above the 3rd are more implementation details e.g. how Backprop works and what criteria they use as in second paragraph (to ground the Backprop). The 3rd paragraph is the actual reason for any algorithm though, because we use matches and predictions from sentences to get the final result, backprop is explained like a mess and so is other mechanisms, this is not what the key picture is.
For the most part GPT seems pretty close to my AI and I think they or others do understand their AI well, but the explanation and way it works still lacks, even if I can get it our of a paper or person it still will disappoint me how GPT exactly works, backprop just seems a bit the wrong way, I don't like it, I have another powerful way clearly being uncovered from what I can see so far. I still think they don't understand why they sue backprop etc, none can explain them, in fact I'm possibly the best AGI researcher that has ever lived...literally no one else can explain AGI... Matt's side and some articles on GPT explain a lot but the GPT explanations are just gotten out the numskulls, like they tried their hardest and is broken and barely made it out their mouths onto paper. Not impressed at all with explanations of GPT. Ya. And their way they do it seems a bit wrong. That's how I feel. I still know what to do, I pretty much know every mechanism in GPT and not in GPT. The implementation method too, but mine is different as said. I don't even know if they know why they use some things n GPT. But I see them and know why they are there. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T73cb0deded02df8c-M55214a593f6eee98968c9022 Delivery options: https://agi.topicbox.com/groups/agi/subscription
