Terminus,
I would like to add one link more:
http://www.brainatlas.org/This is the effort financed by Paul Allen to create a complete, detailed map of the human brain.
As for it (AI), I am wondering, what abilities of the human brain should be programmed as an overriding procedures. (I am not referring to how it will be done, but to what should be done.) As for now, we've talked about the ability to collect information (recognizing characters, etc.).
Yes, I have pressed more in analysis, because NN more simple and common uses are analysis processes.
However synthesis of information is also acomplished with NN.
I personally think, that one of crucial ideas (of future AI) should also be the ability to thing (compute) in an abstract way. That is, in the opinion of many, what differs people from animals. You have to admit, peskanov that even if it is true, that our brains are complex computers, than it is also apparent, that no other animals express the ability to think abstractly in our way.
Should I? The seeds of abstraction seems to be in the animals also...
For me, the concept of number is the supreme abstraction of the human mind, and probably the driving force in our capacity for abstraction. The number is not directly related to any object, action or sensation of the real world.
However there exist evidence that certain innate hability for numbers is already hardwired in our brain as consequence of evolution. It seems we use an special part of out brain to count until three, and it seems this counting hability is also present in some superior animals.
http://www.apa.org/monitor/apr99/math.htmlFor you abstract thinking is radical departure from the animal kingdom, but for me is a question of degree. An evolutive hability, which probably developed and grew with the aparition of the human languages.
Sadly, there no other human species alive today to compare our respective capacity for abstraction.
How, gentelmen, would that be implemented in AI? What do You think?
I have defended reaching artifical human thinking by imitation of existent biology. This method does not require our understading of it's mechanism, only mere replication.
...you don't have to understand how Linux works to copy a CD containing it.
After acomplishing that, you can try to augment it's capacities through try-and error, doing methodical architecture changes, like in the genetic algorythms (for example).
Ps. peskanov pardon for asking, but I won't take it anymore : What, if I may ask, is Your speciality? Are You a programmer? Apologies for asking only You, but me, Deckard, and dzi know eachother from Polish section, and I know that Deckard is a hard-core computer scientist, and dzi is whatever-You-wish-him-to-be
Ok, but you should not use the expression "I won't take it anymore". In English it's used in violent situations (google it for checking).
I am a profesional videogames programmer.
I studied a short universitary course of 3 years, (in a very pragmatic, work-oriented school). There, I also studied expert sytems and prolog, but very lightly, and several algorythms and techniques used commonly in AI (like recursion or path finding).
I have programmed the AI for a few commercial videogames, but this kind of AI is usually quite simple. Usually I program in the graphics part of the games, the 3D engine, not in the AI part.
I learned about NN reading some IEEE transactions from one of my brothers who studied telecommunications. I became fascinated with them and I usually read about the matter regularly and keep with the news from time to time.
Right now I am working in the preproduction of a fighting game for ps2, and I am planning to try NN in the AI for the first time in my career. I want to see if I will able to train some NN using the corpus of data produced by real persons playing the game. If I fail I will have to program it in the classical way
BTW, I also pertain to a very small cultural movement called "demoscene", which basically consist in creating real time computer graphics presentations, programmed in C and/or assembler. I mention it because is quite popular in Poland and maybe you know it.
PD: I don't claim to be an scientis of the AI field, just a person interested in the matter an who works in it from time to time.