JUST a draft…… it goes to in memorization of Shailesh
Ideas about AI.
1, during evolution, human has this skill of transfer information into knowledge. But is running on a slow crappy computer(brain) with low memory and horrible inter-connection. …if it can be simulated on modern computer, with computer’s high inter-connection and speed, will be supper. 2, Right now, All AI development I learned so far are based on mathematic. From English language sentiment categorizing to picture pattern matching….those ANN or SVM are all based on mathematic logic reasoning and calculation. Those are not the way when I learn stuff and get intelligence.
Yesterday, I consciously trying hard to record the process of how I learning a new thing…..So that I may be able to see the process of how do normal people transfer information into knowledge when they learn a new thing…..I think this is a key piece that is missing to make AI really work. So I learned a new terminology called Selenium. And this is what I recorded: 1, I remember I heard it years ago. 2, I saw on web there is a digram, three layer, selenium is in the middle. So I know it’s a key thing that connecting up and down. Therefore I decide to learn this Selenium. So far, What I did was, I try to draw a structure diagram/picture in my head. And now when I keep reading things about this Selenium,
I try to PLACE this selenium into that structure. And, during that, for every new piece of info I got, I was also trying to fit it so that I’m also validating the correctness of that structure. Based on this process, I think I should build an AI, not purely mathematical based(like traditional deep learning(ANN)), but this way
Step 1: crawling web and dig out a new terminology Step 2: read about this terminology and trying to place it into a network diagram, whilst also validating/revising the digram itself. Step 3: Repeat from Step 1