bigstring18.asm

PIC全系列单片机的bootloader程序

	DB	"s having many ways to deal with things. I can't remember what kind of crazy search found this. It's from Aristotle's Metaphysics"
	DB	". I was looking for something to see whether he had this idea. I will read it. 'The essence of a house is assigned in such a for"
	DB	"mula that has a shelter against destruction by wind, rain and heat. But a physicist would describe it as stones, bricks and timb"
	DB	"ers.' You see the difference the first one is a functional definition - the house is to keep the rain off me. The second one is "
	DB	"a structural description - it's something supported by something else as a certain structure. Aristotle says, 'but there's a thi"
	DB	"rd possible description which would say that it was that form in that material with that purpose or end. Which, then, is entitle"
	DB	"d to be regarded as the genuine physicist' The one who confines himself to the material or the one who restricts himself to the "
	DB	"formulable essence'' That's the function, I guess. 'Is it not rather the one who contains both in a single formula''   So here's"
	DB	" an example of multiple representation, not in terms of different data structures, but you might use the same data structure in "
	DB	"different realms of concern. John gave Mary the book. I discussed this example in my old book, The Society of Mind. I'm deciding"
	DB	" how much of that old section to copy and how much to write new. Whenever you see something happen, you can think of it has pure"
	DB	"ly physical, as Aristotle said. John gave Mary the book. First it's in John's hands and you might imagine a physical transfer sc"
	DB	"ript where it moves to Mary's hands. There's another one which I'm thinking of calling 'dominion'. Did you notice that wherever "
	DB	"you go in the world, whatever you see, you can't have it because it belongs to someone. It's very important. Almost everything b"
	DB	"elongs to someone, including the air in some science fiction stories. So what's happened is, it looks like the control of the bo"
	DB	"ok has passed from John to Mary, but we don't know whether he's loaned or maybe it was hers all along, and he's giving it back. "
	DB	"As soon as you hear that, perhaps different parts of your mind are looking for different answers to resolve this ambiguity.   Al"
	DB	"so there's been a change in the memory of John and Mary because unless he's like me, he probably remembers who he gave the book "
	DB	"to. Anyway I conclude that chapter showing a dozen different kinds of data structure that you could use for representing things."
	DB	" A lot of these data structures depend on each other, so you talk about very low level neural nets, or even lower level connecti"
	DB	"ons between things. I don't want to talk about things I call K-line semantic nets, which I think are actually about the richest "
	DB	"representations for problems you don't understand yet. That's what relational databases have in them. Then the kind of frame arr"
	DB	"ays and trans-frames and scripts that I talked about in The Society of Mind.   So the moral of all this is that when you have a "
	DB	"new problem and you say what's the best way to represent this, should I use a neural net or a relational database, the answer is"
	DB	" if you're just trying to get something done quick and it doesn't have to be very good or resourceful, then I'll suggest a way t"
	DB	"o pick one. But if you're looking for something in the future that's like a baby machine that's going to keep getting smarter, t"
	DB	"hen you'd better not have it learn by statistical associations or neural nets or whatever. You should probably start it off with"
	DB	" five or six of these things and figure out some way of negotiating and managing that the different representations give you alt"
	DB	"ernate ways of using the same kind of knowledge.    Then somewhere in my web page you'll find this diagram, which I'm very proud"
	DB	" of, although it's been around since 1993. This is a little diagram that suggests why it's sometimes better to use neural nets o"
	DB	"r fuzzy logic, and why it's sometimes better to use logical reasoning.   This is just a handy theory. You look at this problem. "
	DB	"My favorite example is you see there's some property of a situation and you ask what does it depend on. Does it depend on a larg"
	DB	"e number of small causes or a small number of large causes. What keeps this chair off the floor' It's four legs. Then you can do"
	DB	" logic. You can say if that leg were missing and it got a little force, it would tilt that way. If the front two legs were missi"
	DB	"ng, then it wouldn't need any force, it would just fall. Neural nets won't do you any good with that. Or if they do, they'll was"
	DB	"te a lot of resources. On the other hand, what keeps this jacket on the table' Well, it's probably 5-10 million little linen fib"
	DB	"ers touching the surface and each pushing up with a few milligrams of force. If 10 thousand of them disappeared suddenly, you wo"
	DB	"uldn't notice the difference. So that's a good place to use a neural net like thing.   I call this the theory of causal diversit"
	DB	"y. What annoys me is it's not that I like this theory, but I've been peddling it. I say, tell me a similar theory for when it pa"
	DB	"ys to use the genetic algorithm or a blah, blah. In other words, what went wrong in AI is that people went off around 1980 in di"
	DB	"fferent directions. If somebody is doing grammar, instead of grammar they're doing statistical associations of words and cluster"
	DB	"ing theory to find the word types and so forth. That's very nice, but I want to know, when is that better than using a simpler t"
	DB	"extbook type grammar and what do you mean by when' For what kind of problem' If you're trying to transcribe a conversation that "
	DB	"you dictated from hearing two of your roommates talk, you'll find that it isn't even made of sentences very much. People stop in"
	DB	" the middle of the sentence when they know that you probably got the idea and so forth. Then statistical things might help, or e"
	DB	"lse something that represents conceptual ambiguity very graphically and clearly.   In other words what I'm saying is not that th"
	DB	"is is the right diagram, but how come you don't have 5 or 10 of them and why don't people think more about that sort of thing. L"
	DB	"et me put it another way - artificial intelligence...of course there's nobody in charge of it, so you can't say what's it for. C"
	DB	"learly some people are trying to make machines smarter at particular things. Lots of people do that. A few people are trying to "
	DB	"make machines generally more resourceful. I only know of about a dozen such people. Some people are trying to make theories of h"
	DB	"ow the human mind works. That's cognitive science.   But in any case, if you look at those, you might have [] of higher level qu"
	DB	"estion, which is, what are the kinds of problems that a person can solve' What are the big classes of problems' I don't know. So"
	DB	" you see the reason why this theory of causal diversity is nice, [] it says well in some problems you're trying to understand an"
	DB	" effect that has a lot of small causes. You see all 9 holes in this diagram are meaningful. If there are just a few causes, then"
	DB	" they have small effects. Then it's not a problem. That's the empty square. If there's a medium number of problems with small ef"
	DB	"fects, then a linear theory will probably do, and maybe a simple statistical one. If there are a lot of causes with small effect"
	DB	"s, then you need some elaborate connection, neural network or something like that.   It's hard to find anything else that will d"
	DB	"o. If a small number of causes with large effects, then that's clearly logic. If you're in the middle, that's where classical, a"
	DB	"rtificial and traditional heuristic search worked. The chess program. Actually it's a small number of causes with large effects."
	DB	" But that small number is about 32 chess pieces and they're all interacting in rather horrible ways. Loss of a pawn has a large "
	DB	"effect eventually. So classical AI handles those problems by big searches perhaps. You can use case-based reasoning or analogy-b"
	DB	"ased reasoning. This middle ground is where traditional AI was notably successful. But not many people do it anymore. In this co"
	DB	"rner, the problem is hopeless. If there are a lot of causes with large effects, then you're screwed. But there's a way to get ar"
	DB	"ound it, which is to find a different representation like happened in physics when most people couldn't cope with the Heisenburg"
	DB	" matrix formulation for quantum mechanics when there are a lot of variables. But if you put it in terms of the Schroedinger equa"
	DB	"tion, it looks much neater and everybody felt better. Obviously there were still 20 variables you couldn't do a damn thing. But "
	DB	"you could prove meta-theory about it more easily. So making a new representation is the big deal if a problem seems hopelessly c"
	DB	"omplicated.   That's mostly what I wanted to say. So the question is why we didn't get HAL in 2001' I think the answer is I beli"
	DB	"eve we could have. [] I once went to an international conference on neural net. There were 40 thousand registrants. I don't know"
	DB	" how many go to an international conference on genetic algorithms or genetic programs, but there are many thousands of people wo"
	DB	"rking on that. Tens of thousands of people try to make slightly better rules-based systems. That's another very useful represent"
	DB	"ation. But very few try to make a system that will in some very smart ways make new rules on the basis of experience. That's the"
	DB	" obvious thing to do. Randy Davis tried it for awhile, but I haven't heard of much work in that direction. I'm sure there is. Bu"
	DB	"t you see in these areas that have real growth potential, if you had an international conference, for example, on using multiple"
	DB	" representations for common sense reasoning, I've only been able to find 6 or 7 people in the whole world. I'm sure there's at l"
	DB	"east twice that many that I don't know of, but not thousands.   So the reason nothing much happened, from my point of view, it's"
	DB	" not fair to the people who made lots of progress in qualitative physics and certain kinds of linguistics and heaven knows what."
	DB	".. but what happens now' What's the worst of all'   Here is a word for the simulation people. I think simulations are immensely "
	DB	"important. All over the world there are probably 40 thousand students working on making stupid little physical robots with their"
	DB	" programming. Nothing seems ever to be learned from this. When I was a young student, I met W. Gray Walter, a scientist in Brist"
	DB	"ol, England, who had made these little turtles that reacted to light and crawled around and did a lot of the things that your ty"
	DB	"pical mini-robot does. He showed what they could do. It became clear that they couldn't do very much. After awhile this early pe"
	DB	"riod of the late 1940s and early 1950s, that got abandoned. It started up again in force in the 1980s. I can't for the life of m"
	DB	"e see why any student would do this. I'm very angry at a professor who would ask a student to. Because in the 10 or 15 years sin"
	DB	"ce 1980, maybe there have been 3 or 4 interesting discoveries. People will tell you we've discovered that you have to make the l"
	DB	"imbs compliant or they won't work. That was known in the '70s. Tell me something else that you've learned from building a physic"
	DB	"al robot, and I'll tell you someone in the 1970s who wrote a big paper on that. So the student is wasting a whole year or three "
	DB	"soldering connections and working with bad components. Every now and then the robot will go down the hall and actually find a do"
	DB	"or and go through it, if that's what you're programming it to do. But you don't know why because next time it won't. That's why "
	DB	"you'll find that these robotics people treasure their videos - because it won't work tomorrow. The reason is that the real world"
	DB	"...but they say in the real world, you know, how do you know your simulation contains all the important things' The answer is te"
	DB	"ll me an important thing you've discovered and I'll make sure that it gets into the simulation. Nobody ever tells me one. Yes, t"
	DB	"he gears could have backlash, and I bet that your best CAD program can allow for that. I don't know. Yes, the floor might be sli"
	DB	"ppery and I'll bet there are 7 kinds of friction that you can understand. But then you can have the robot go over these 100 time"
	DB	"s varying a few parameters and understand the problem. But what happens in physical robotics is you never get to do the same thi"
	DB	"ng twice. There's no science. There's no replicable experiment. It's just like ESP, meaning usually it doesn't work, but it work"
	DB	"s if you're happy enough to have a video.   So I think simulations are immensely important. If you see a student who says I'm bu"
	DB	"ilding another robot, tell him 40 thousand people are doing that. In 15 years they've discovered 5 things (giving them the benef"
	DB	"it of the doubt). That means that in 400 thousand years, if you divide that...let's say 40 things...then you have one chance in "
	DB	"10 thousand of discovering something. Boy, isn't that an incentive' It's in common sense reasoning where we discover 5 things a "
	DB	"month (I'm not kidding), all 7 people. That's where the student should go.     From ARTIFICIAL INTELLIGENCE GAME DEVELOPERS CONF"
	DB	"ERENCE, MARVIN MINSKY.   [This was a test to fill PICs memory with data, and instead of something like 'The quick fox jumped ove"
	DB	"r the lazy dog' I selected something more interesting. And now a serial test:]  abcdefghijklmnopqrstuvwxyz ABCDEFGHIJKLMNOPQRSTU"
	DB	"VWXYZ abcdefghijklmnopqrstuvwxyz ABCDEFGHIJKLMNOPQRSTUVWXYZ abcdefghijklmnopqrstuvwxyz ABCDEFGHIJKLMNOPQRSTUVWXYZ abcdefghijklmn"
	DB	"opqrstuvwxyz ABCDEFGHIJKLMNOPQRSTUVWXYZ abcdefghijklmnopqrstuvwxyz ABCDEFGHIJKLMNOPQRSTUVWXYZ abcdefghijklmnopqrstuvwxyz ######."
									;ARTIFICIAL INTELLIGENCE GAME DEVELOPERS CONFERENCE 2001 MARVIN MINSKY 
Table1_end

		
          END