![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
guslI'm a few days away from handing in the final version of my thesis.
* Are you interested in Kuhn's description of normal science, and how scientific derivations are found? Would you like to how see reuse in scientific derivations is analogous to reuse in natural language and reuse in programming?
* Are you interested in how scientific derivations are justified when formalization fails? Have you always been unhappy with physicists for not being logical, and wondered how the heck their "derivations" are justified?
If you answered YES to the above, you will probably find my thesis interesting.
Here it is:
Automating Normal Science: Reusing Exemplars in Quantitative Explanations.
Warning: I'm still working on this document. The beginning is pretty smooth already, so you can get a good idea of what it's about... and the rest will probably be easily understandable, if sometimes unsteady, slightly disorganized and slightly incoherent reading.
The more comments I get the better. Please be frank. You can also email me if that will make you more frank. :-) I thank and appreciate all comments.
If there's anything that I should explain better, please email me! Having readers helps keep my motivation up. For what it's worth, you will be credited, unless you indicate that you'd rather be anonymous.
* Are you interested in Kuhn's description of normal science, and how scientific derivations are found? Would you like to how see reuse in scientific derivations is analogous to reuse in natural language and reuse in programming?
* Are you interested in how scientific derivations are justified when formalization fails? Have you always been unhappy with physicists for not being logical, and wondered how the heck their "derivations" are justified?
If you answered YES to the above, you will probably find my thesis interesting.
Here it is:
Automating Normal Science: Reusing Exemplars in Quantitative Explanations.
Warning: I'm still working on this document. The beginning is pretty smooth already, so you can get a good idea of what it's about... and the rest will probably be easily understandable, if sometimes unsteady, slightly disorganized and slightly incoherent reading.
The more comments I get the better. Please be frank. You can also email me if that will make you more frank. :-) I thank and appreciate all comments.
If there's anything that I should explain better, please email me! Having readers helps keep my motivation up. For what it's worth, you will be credited, unless you indicate that you'd rather be anonymous.
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
(no subject)
Date: 2005-09-26 01:46 am (UTC)I noticed a section where you describe "theoretical" versus "empirical" laws. There's a word that gets used by physicists a lot "phenomenological", or "phenomenology" that fits exactly with the description you give for an empirical law. I'm sure they could probably both be used in that context interchangeably, but in case you wanted to throw in another term there... phenomenological is often used in science to mean that (http://en.wikipedia.org/wiki/Phenomenology_%28science%29). For instance there is one archive online for "High Energy Theory" and a separate one for "High Energy Phenomenology"; I've always found that interesting and wondered if in the future there will be at least 3 divisions between science: experiment, phenomenology, and theory.
What do you think of Thomas Kuhn by the way? I have not read The Structure of Scientific Revolution yet (although I probably will soon) but I have found a lot of the things I've heard him say second hand quite offensive to my view of science.
(no subject)
Date: 2005-09-26 10:32 am (UTC)yes, "empirical" = "phenomenological" law / model
Re: Kuhn
I used to have the same impression as you. From the little I know, I think he's one of the best philosophers of science of the 20th century. Unfortunately, his message can be easily misused by relativists.
(no subject)
Date: 2005-09-26 05:29 pm (UTC)Q: “Introduction: What this thesis is about.
A: This is redundant. Just say Introduction.
Q: Your opening sentence is declarative, and good. But the next sentence “Thus, by implementing…” Drop the “thus” you are not an old man pontificating. You speak to reader through out your paper as if you are pacing back and forth in a classroom “and if we look this way you will see…blah blah blah” Words like “we” “one” (as in “when one does”) is academic slime. Also “As we will see in Ch. Philosophy.” My vote is drop it unless you really feel your audience expects it. Which if they do, that is a shame.
Q: “one can be said to be”
A: Why are you being indirect? You write passively. Just say what you want to say. You are such a “P” : )
Q: 1.7 Goals
A: This is closer to what your introduction should be. Take the body of this and move it up.
Q: “Unlike theories, laws can be stated in few words”
A: This is only because there does not need to be room for mistakes. As in “hind sight is 20/20.” Sort of like how people that are giving you directions to a place they have not been in a while may be correct (and this is important, but subtle here), but will also require more words to describe because you want to give the other person every chance to get there. Overall I feel your entire paragraph on Laws is opinion, and itself a theory, and thus wordy. I’m not saying this because it is fun to do (although it is), but ask yourself, what is the law of laws? Now, state it succinctly.
Q: The words “neat” and “scruffy”
A: I don’t know why you chose these words. But in America at least “neat” has several meanings, and the one that will come to mind first is probably not what you want. It is not “organized” or “fitted” as you might like. Rather it is a common expression which is also sometimes “Neato” which is what kids used to say about 20 years ago, like “groovy” and “rad.” “Hey man, that is way neato” or “That new stylized skateboard is neat, let me play with it.”
And “scruffy” conjures a dirty dog, like Benji (from the movie), or Brad Pitt when he has not shaved for 2 weeks.
After reading your paper, may I suggest, if an American market is ever who you want to read this, something more like “elegant” and “shabby”? Otherwise ignore me.
(no subject)
Date: 2005-09-26 05:30 pm (UTC)Side note…I’m in Greece right now, so I keep accidentally trying to figure out what equation a sign is trying to present LOL. I consider this a parsing problem on my part. Although I am getting pretty good at reading modern Greek. And not everything is reducing polar coordinates, or calculating circles.
Now for the meat of my feedback:
Is there an example of that you can prove where one path resulted in failed conclusions by “reusing,” and another example (where they stepped outside the box) to reveal the truth of a formula? It seems easy to confine science to its abuse of reuse, but the fact is that science is best conducted for the scientific collective in baby steps. It is not bad, just methodical. It only takes a few geniuses to set the hoard onto the next big job.
And, an entire area I may have missed, that may be staring me right in the face in your paper, but in the 1000 days a child learns a language, in fact 800 of those days are making repeated mistakes, and the other 200 are feeling secure enough in the truth. So it seems a good chunk of your paper should be going into depth of how the system will learn from failure. I’m a big fan of Chomsky, but he was simply wrong about the human brain. He confused his brilliance in one area with his ability in another (biology and neurology). Humans are simply complex machine that can be used for all sorts of things, we keep learning things us apes can do that have no pattern or example in evolution. Not to imply that this proves anything like something other than evolution, just simply that we are sort of like Swiss army knives, and one day we realize that the bottle opener makes for a pretty good screw driver, someone had to invent the screw, as someone had to invent the bottle.
The same pathways we use for language are mostly overlapped strongly with many other fun tricks in our brain that can be used for other purposes. Like a person that can play chess very well. Is there a chess module? Is there a dance module?
Which brings me to the hardest problem of your thesis (which I happen to like the topic and the concept a great deal). At first the idea of reducing language to the known parts, and then rebuilding them into a proto-language that can be rebuilt into any language seemed pretty straight forward, and yet we have failed to even get close so far (I suspect this will in fact change eventually, but only with the use of a very sophisticated group of tricks, and huge memory banks, compared to what we have today).
There is of course a pattern to formulas, but I have a tough time believing there is a discernable pattern to “correct” or “useful” patterns.
In language this is what Koans are about. “What is the colour of the number 3?” Again it is the failures in language that reveal the truth, not the successes.
(no subject)
Date: 2005-09-26 06:20 pm (UTC)Yes, but DOP is still pretty good. FWIU, the best computational linguistics methods since the 90s have been based on DOP. But I would like to understand the patterns of failure. One obvious barrier is that our present NL systems lack "general intelligence" and "common sense". But I think DOP basically "solved" the problem of parsing. Today, parsers are really good because of DOP. Warning: I am probably biased... and most definitely ignorant.
(I suspect this will in fact change eventually, but only with the use of a very sophisticated group of tricks, and huge memory banks, compared to what we have today).
well, indeed! This is why DOP can maybe provide an indication of what to choose, but the result needs to hold up empirically too. DOP is only one heuristic in the process of scientific discovery.
(no subject)
Date: 2005-09-26 06:43 pm (UTC)I want to connect it to the first sentence somehow.
I took a philosophy class last semester, and while the prof liked my presentation (interesting, lively discussion), he said I was being too much of a "J", that I was an "extremist" and should label myself as such. While this last bit of advice is BS, otherwise I think that's the way people expect to read philosophy, and I have no reason to be forceful: it would just piss off the people who disagree or dislike the terminology.
"neat" and "scruffy"
Date: 2005-09-26 06:46 pm (UTC)What I call "scruffy results" are results that are indeed "dirty", because they were derived by using questionable assumptions.
The stuff on "laws", etc, I'm mostly taking out of reference books on Phil of Science.
(no subject)
Date: 2005-09-26 06:51 pm (UTC)A: could you rephrase this? I'm still searching for examples for many of the claims that I make, but I'm don't understand what you mean here.
Q: learning from failure.
A: This is interesting, but beyond my scope... because it involves judging experimental results.
(no subject)
Date: 2005-09-26 08:32 pm (UTC)Yeah, I guess you have to play to your crowd...
(no subject)
Date: 2005-09-26 08:45 pm (UTC)However, to answer your question, you mention you are looking for examples. A common problem programmers have is that they will begin to think of how to solve a problem, and upon looking at the first few examples that come to mind either conclude that the permutations are endless, or that a simple formula will work. Both are usually wrong. As humans, even smart humans, they tend to follow the same path of logic, which as far as I’m concerned is simply one or two more steps that separate a really smart person from a normal person.
“if she floats she’s a witch”
A smarter person will realize this is not a very good test because it does not prove she is a witch.
An even smart person will realize that it is the question itself that is not very good because it assumes that there are witches in the first place.
In your examples, you are testing for what you know to be true. I’m asking you to add a solid example in your thesis of why a nonsense formula would be removed fro the tree. Is it alone true because the formula simply has no matching pattern?
(no subject)
Date: 2005-09-26 09:01 pm (UTC)A: there are no "nonsense formulas". There are formulas that are known to hold in a given situation, and formulas that are not known to (and usually don't).
You could also do analogical reasoning, but that's very unrealiable.
I am going to add an example or two like this: using naive syntactic formalizations, you would use bad formulas rightaway (unless the user were really careful, but the whole point of the new semantic formalizations is that you do have to do all this extra work anymore), whereas with semantic formalizations, formulas that don't apply are filtered out. ("No, I'm not orbiting around the Moon! So stop telling me that my potential energy is high!")
(no subject)
Date: 2005-09-26 09:27 pm (UTC)(no subject)
Date: 2005-09-27 12:51 am (UTC)(no subject)
Date: 2005-09-27 12:53 am (UTC)I'm still a couple of sleepless nights away, I think.