From: mailbox@dmitry-kazakov.de   
      
   On Tue, 27 May 2008 12:13:34 GMT, Andrey Gavrilov wrote:   
      
   > "Dmitry A. Kazakov"  writes:   
      
   >> Because according to the modern understanding of things, there cannot be   
   >> experiments which could refute the uncertainty principle. In that sense   
   >> any description must comply with the principle.   
   >>   
   > This is universal principle for clearly defined "intelligent system" and   
   > with respect to complexity (or similarity of real world) of environment   
   > (area) similar to that uncertainty principle in quantum physics depends on   
   > speed of particles, size of particles (neutrino or nucleon or atom or   
   > molecule and so on).   
      
   Well, this explanation looks much like the theory of hidden parameter,   
   which presently is assumed wrong. The uncertainty of measurements is not in   
   the complexity, but in the nature of things. The barrier here is   
   fundamental.   
      
   If uncertainty of our models of intelligence is merely rooted in complexity   
   of the corresponding processes, then nothing fundamental should prevent us   
   from describing them exhaustively.   
      
   > There are not exact bouder between stochastic and deterministic processes   
      
   No, the border is clear, it is said that quantum processes are truly,   
   non-pseudo random.   
      
   Anyway, I don't see connection between these two principles. Physics does   
   not claim that if we knew the structure of a molecule, we would not be able   
   to describe chemical reactions it participates (behaviour), or the inverse.   
      
   >> Ah, but only one intelligent thing is equivalent to the real world. That's   
   >> Mr. God. I don't think that the objective of AI were to create gods...   
   >> (:-)) Without falling into solipsism, intelligence is obviously less than   
   >> the real world. Inside solipsism, intelligence as a concept is   
   >> meaningless.   
   >>   
   > I don't understand or you don't understand me.   
   > Meaning of my statement is that there is no any mechanism for full   
   > description of universe (environment).   
      
   Yes, so my question was, why in order to describe intelligence we would   
   need a full description of the Universe? I don't see obvious reasons for.   
      
   > But even in this case if we have enough large number of inputs, outputs,   
   > neurons and samples during learning (action) of NN it is impossible to   
   > describe behavior of NN as DFA and this is nonsense.   
      
   So long the number is countable finite we can.   
      
   Another my question was: whether your principle is fundamental and there   
   exist formal reasons implying it, or else it is just an observation of our   
   inability (at the moment) to handle the complexity of the problem.   
      
   > If we implement NN in another platform (e.g., in chemical anisochronous   
   > processes as in brain) we absolutely can not describe it by DFA.   
      
   No, this plays no role. An asynchronous hardware running DFA is still DFA.   
   If it walks like a duck...   
      
   It yet has to be shown that the computational model need to be   
   asynchronous, non-deterministic, more than Turing-complete etc. So far we   
   just do not know anything certain about that.   
      
   And if your principle is true, then we will never know. Because it implies   
   that if we knew the structure of a system (per construction of), then it   
   could not work (for its behavior would be necessarily uncertain).   
      
   In a somewhat loose interpretation: people grow intelligent, because their   
   parents never learned biology... (:-))   
      
   --   
   Regards,   
   Dmitry A. Kazakov   
   http://www.dmitry-kazakov.de   
      
   [ comp.ai is moderated ... your article may take a while to appear. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)