Quote:
Originally Posted by Doctordick
Kriminal99, in case you have been wondering,I haven't been ignoring you; I simply find it quite obvious that you don't understand the issue of interest in this thread “What can we know of reality”. There is a problem here which everyone (save Anssi) seems determined to ignore. For example,
You cannot so define honesty without first defining the words, "what”, “people”, .....
.....+ “word #5672” + “word #3490”. And, since I wouldn't know what you meant by the word “honesty” until after I understood your definition, I could simply express the information you intended to express in the above sentence as “word #8675” equals “word #25”+ “word #3303”+ “word #4399”+ “word #24”+ “word #7867”+ “word #2124”+ “word #675”+ “word #2”+ “word #345”+ “word #215”+ “word #2739”+ “word #2237”+ “word #10233”+ “word #2”+ “word #754”+ “word #390”+ “word #287”+ “word #5672” + “word
#3490”.
In order to understand what you meant I would have to interpret a great number of very similar expressions (a sufficient number to understand the English language). Every year, millions upon millions of children manage to decipher that kind of information in only a period of a few short years (apparently coming up with very similar conclusions). That means the problem is solvable! And, the creation of AI will depend very seriously on being able to solve that exact problem. I am discussing inherent constraints on that problem expressed in a manner which does not presume your current answer (at time t) is correct but demanding that your current answer (at any given time t) is “flaw-free” (a rather different constraint). I use mathematics because it is the most unambiguous language available to me and we need a language to communicate. To quote Richard Feynman, “mathematics is the distilled essence of logic.” It is very nice that we possess such a language, limited as it may be.
|
Nah I do understand it I promise. In fact I already came up with the solution of how define those words (you don't need the whole language) and how children do it a long time ago. I was hinting at it when I mentioned that one only need accept that the frequency with which something has happened in the past affects the frequency of it occurring in the future... Though I suppose it requires a few other memory storing and fractioning abilities without which we could probalby do nothing but drool. I solved this problem by organizing simplified information, where simplified refers to a specific algorithm that I referred to as minimalism. It turns all further philosophical discussions into math and probably would allow for an AI. All my posts were trying to respond to your arguments.
I had tried to make posts regarding that system before, but I just got banned for repeatedly referring to reality in contradiction to things moderators stated.
Anyways, the way the system works is simple and contradicts what Wittgenstein argued. Witt failed to account for people's perception comparison and pattern recognition abilities.
For illustration I will use <> to encase a perceived memory and &# to denote coincidental perceptions, which have occurred # number of times. | will denote or. Certain simple memory related abilities are assumed to be hard wired, with our automatic ability to perform these tasks as adults for evidence.
If a child sees a room with a book in it, and simultaneously hears the word book, he hasn't learned anything yet but perhaps he stores the memory.
Somewhere in his head is the record:
<Sound of word book> & <Image of a room with a book in it>
The next day he is in a completely different room with a book also laying in front of him. (preferably the same one for simplicity's sake, but not necessarily) The child stores the memory:
<Sound of word book> & <Image of a different enviornment with a book in it>
Then the child simply compares the similar parts of the two images, which just happens to be the book. Then he removes the rest of the images because of the lesser frequency with which they coincide with the word book and is just left with
<Sound of word book> &2 <Image of book>
Of course it might take more than twice, especially if its a different book or the book is being looked at from a much different angle. But with the implied comparison and pattern recognition ability, many factors can be removed through repeated experiences.
Tomorrow someone says "open the book", and the child watches someone do so. Through similar repeated occurrences he removes factors and obtains the association:
<Open the book> &(X>1) <String of images involving someone opening the book>
Later, someone in front of the child asks someone to open a different object which the child has already been able to associate with it's image. A box perhaps. This verb has not changed, and is associated with a animation that is similar to the one the child saw coincide with opening the book. Thus:
<Sound of command "Open the <Book|Box>"> &(X>3) <String of images involving someone opening something>
Now the child has something to associate with the higher level concept object, even if he doesn't do so right away. Specifically what is to be an "object" is something that can be "open the" 'ed. It is a function of what the child has thus far perceived.
Then the child hears someone say, "Grab the book" and then, and watches them do. And then they see someone being told to and grabbing other things .
<"Grab the <object>"> &(X>1) <String of images of someone grabbing an object>
and also
<Verb> & <String of images involving a person doing something>
Mind you not the word verb, but just the underlying concept just as in the case of object. And also:
<"the"> &(X>1) <Word that comes before <object> and <verb>>
And it continues such that every high level concept can be defined as a function of things previously perceived.
Such memory abilities are implied by our ability to consider something like a unicorn, which consists of a horse that has had the horn of a goat appended to its head.
