Google blogs?
3:43 PM Edit This 0 Comments »
Whoa, Google's taken over something else. I'd protest, but it has auto-save now. :) And at least it'll be harder to lose my password.
I completely forgot I was working on this. One post per semester, I suppose it works! I'm a junior now and thrilled to get to take upper division classes. Algorithms is really neat, Networking seems a well-chosen elective so far and the SCI advisors committee, or whoever they are, even let me into a grad class this time.
Software Engineering, on the other hand, is not my thing. First time I think I've had a great teacher for a subject I genuinely find boring. Why? It seems all about following strict methodical rules in a corporate environment. I like inventing processes as I go. Writing out the same spelling words repeatedly in elementary school, I'd start writing them backwards and from the middle just for a change. But it may be the key to my problem with getting lost in conceptual "levels." As much as I've said I think using a reductionist approach, I really switch very quickly back and forth from a birds-eye view to what the worm sees, and everywhere in between. I get so confused writing essays because I can't look at how a sentence fits into the whole without seeing the whole and then noticing how it fits into the sentence. Code has fewer possible layers of meaning, that helps. Unless you could find a meta-compiler that has a semantic interpretation for entire programs...but computers don't work like brains (goto: next few paragraphs, I just got ahead of myself). Anyway, I think that class might be good for me so I'm going to try and keep an open mind. Also, "agile" methods--alternate ideas for efficiency, like programming in pairs-- are actually pretty interesting.
Anyway, I finished GEB (well, most of it). My new find is On Intelligence, by Jeff Hawkins. It was published in 2004 so I'm surprised I hadn't heard of it--until an ex-roommate was reading it for a objectivist discussion group meeting and recommended it. Some roaming around the internet unearthed skepticism and opinions that Hawkins is abrasive, full of himself, and too sure of his theory.
Well, he might be wrong, but I found the book extremely well-written and very interesting. Notably to me on the well-written thing, he explained the concept behind feed-forward neural nets in two sentences. I remember struggling with my write-up for the parallel neural net code I wrote in MPI last semester, and it's not trivial to explain things so clearly. And his ideas make a lot of sense. He explains that the brain sees all incoming sensory data as the same thing: they're all just patterns, which is exactly what my friend Anh Vu and I got to in a recent discussion on consciousness. He goes on to suggest that these patterns in space and time are also treated in the same way. The brain has one algorithm for learning sequences of patterns, and it applies it to all patterns it receives in a hierarchy. This also seems very intuitive. I've always thought of concepts as a higher-level view of smaller constructs, which can themselves be made into other concepts. And Hofstadter (Firefox, I love you for knowing how to spell that) lists most eloquently cascades of that kind of example: speaking ant colonies, tiny copies of the word 'reductionism' making up 'holism,' and even neurons that form thoughts (I should look up that last one specifically, he was looking for something specific there but I can't remember what).
I completely forgot I was working on this. One post per semester, I suppose it works! I'm a junior now and thrilled to get to take upper division classes. Algorithms is really neat, Networking seems a well-chosen elective so far and the SCI advisors committee, or whoever they are, even let me into a grad class this time.
Software Engineering, on the other hand, is not my thing. First time I think I've had a great teacher for a subject I genuinely find boring. Why? It seems all about following strict methodical rules in a corporate environment. I like inventing processes as I go. Writing out the same spelling words repeatedly in elementary school, I'd start writing them backwards and from the middle just for a change. But it may be the key to my problem with getting lost in conceptual "levels." As much as I've said I think using a reductionist approach, I really switch very quickly back and forth from a birds-eye view to what the worm sees, and everywhere in between. I get so confused writing essays because I can't look at how a sentence fits into the whole without seeing the whole and then noticing how it fits into the sentence. Code has fewer possible layers of meaning, that helps. Unless you could find a meta-compiler that has a semantic interpretation for entire programs...but computers don't work like brains (goto: next few paragraphs, I just got ahead of myself). Anyway, I think that class might be good for me so I'm going to try and keep an open mind. Also, "agile" methods--alternate ideas for efficiency, like programming in pairs-- are actually pretty interesting.
Anyway, I finished GEB (well, most of it). My new find is On Intelligence, by Jeff Hawkins. It was published in 2004 so I'm surprised I hadn't heard of it--until an ex-roommate was reading it for a objectivist discussion group meeting and recommended it. Some roaming around the internet unearthed skepticism and opinions that Hawkins is abrasive, full of himself, and too sure of his theory.
Well, he might be wrong, but I found the book extremely well-written and very interesting. Notably to me on the well-written thing, he explained the concept behind feed-forward neural nets in two sentences. I remember struggling with my write-up for the parallel neural net code I wrote in MPI last semester, and it's not trivial to explain things so clearly. And his ideas make a lot of sense. He explains that the brain sees all incoming sensory data as the same thing: they're all just patterns, which is exactly what my friend Anh Vu and I got to in a recent discussion on consciousness. He goes on to suggest that these patterns in space and time are also treated in the same way. The brain has one algorithm for learning sequences of patterns, and it applies it to all patterns it receives in a hierarchy. This also seems very intuitive. I've always thought of concepts as a higher-level view of smaller constructs, which can themselves be made into other concepts. And Hofstadter (Firefox, I love you for knowing how to spell that) lists most eloquently cascades of that kind of example: speaking ant colonies, tiny copies of the word 'reductionism' making up 'holism,' and even neurons that form thoughts (I should look up that last one specifically, he was looking for something specific there but I can't remember what).