Just for fun. I think that people used to say the internet was a passing fad and would die out and fail. Human language processing is IMPOSSIBLE. The strange thing is that even if you show some people the completed contradiction, they still deny it and strangely many people will accept that delusion because it strokes their ego.
This link at Wikipedia is an example of the attempt to define the problem as a perfect solution. There is the problem, there is no absolute or perfect solution in infinite forms. The simplest interface is the human interface and it is very limited in scope. It has really three options, 1. Point 2. Click 3. Type. Now the results can be quite complex in their factorial combination, but at its central point of action it is deadly simple. I can't make any method that does not result in a combination of those things ( excluding different input types which are equally dull in their expression ). A person could rant on about syntactical hierarchy of language and symbolic systems for centuries and still be no farther along. I can cast an infinite number of sentences that have little or no meaning as is obvious from my normal posts. That which cannot be understood or is not something that affects me or I can't act on have little meaning to me. What I need to know is what things can I do to satisfy my interests. Mine are knowledge and application. So it is simple for me to exclude about 99.99999999% of everything on the internet as noise. What I need must relate to what I already know and be new to me and as such it is in that list like thing that is my experience and if it has no extension or connection there , I can't use it anyway. I could pass up something very important, but I will never know until I reach that level of understanding that allows me to see that this is my next step.
I have been swirling about this concept for at least 10 months. The ideas of context free grammar, compilers, Python, C++, translation, DNA, logic, and a thousand other concepts. The result is that I have all the knowledge and tools I need to implement a computer language that is human in form. I like the idea of having languages that use mov ax,1234h; jmp 0fffffh:f000h and std:cout and if ( -e "zombie" ) then fi or many other cool coded things like scheme and lisp and python. The problem is that it isn't natural in any way. It is no problem for me, but other people probably have a life and don't want to know whether the object class inheritance name overloading is consistent across machine code implementations.
The problem as I see it is that computers have gotten millions of times smarter and Microsoft and many others do not want language to be easy as it would screw them into the ground and take away their advantage. Many people just go along and think it would be impossible to have a human language interface to the computer. I feel very stupid because I should have gotten this long ago. I have devised an English language programing method which anybody can use to program their computer and share knowledge in the same way it is done in open source.
The key to making it work is the method I just discovered with image recognition. This is supposedly the impossible dream. The solution comes from what I have done in dimensional space to read an entire image and screen and explode it out as objects of specific and easily separable types. If I have a screen with 2048x1024x32bit colors ( which is what I use ) that is 8 million bytes of information. It would seem that it would require a super computer to do what the mind does easily, but it actually resolves quite easily. It then makes it very easy to act within that object space. The 240 million pixels that bombard me every second become a few objects (windows) with names and within those windows are other objects and so on until it becomes a letter.
So this is how it works. I have devised a language that allows me to script the action of a person and record that action as a symbolic English stream that is just sentences with no special semi-colons, indents, <> or underscore-underscore__camelCase(){} but just exactly the thing you would do to explain how to do something to a noob.
The magic comes in the translation of sentences to meta code and methods with images. What possible things can a person do? Click mouse, move mouse, type. Now tell me how difficult is to define primitives that implement 3 types of action if it is possible to identify what is to be clicked. And that is the big difference. Once it is possible to navigate the visual landscape as objects, the interface is simple and I could just as easily implement it as LOLcode, srsly. The action of interface to the computer is the simple interface and easily simulated, and yet it produces complex action.
It actually makes it possible for a person to define common actions across key short cuts that a person cannot remember. It combines the simplicity of text interface with the absolute memory of the computer for trivia.
Open firefox. Click Blogger Sign in. Click new post. Type Universal Language in the Title: field. Give me the keyboard.
It is an obvious progression and why nobody saw this before is odd. People create meta code, meta-meta code and scripts that write python and C and yet it is far easier to write code that says "Open terminal. Type find -iname "zombie". and have the C program that reads the script recognize the syntax that it expects in a specific context. The code need only be written once and even if it can't do everything, it does most of what people might want to do. In some ways it is already done in a more complex way. If I get a debian .deb file, I know what it is doing and it could just as easily be spoken or written in English and translated by an interpreter.
I don't think of it as a universal translator, it doesn't understand ideas or would not even have a chance of understanding what I just wrote, but it knows what a mouse is, it knows how to invoke a program, it knows what move the mouse means, it knows what click is and it can find things on the screen to click. So, start terminal and type ls. The syntax need not have anything to do with the actual program and "leik start skypezorz dude." "yeah and connect to irc :)"
It doesn't have to be perfectly defined in syntax as long as it knows what people are like and ignores things like "rm -rf /". It doesn't have to execute everything it encounters and can simply say that it doesn't get what this is about and stops so you either add the code in the interpreter or do it yourself. On top of that it doesn't spew hexadecimal at you and expect you to have a response. That is stressful, like a math test on the thing you failed to study.
The nicest thing is that it acts somewhat like inheritable intelligence.
I am going to make an implementation of it this weekend and should have a working prototype language by Sunday or leik maybe Monday srsly.
0 comments:
Post a Comment