On Sun, Dec 18, 2005 at 11:20:29AM -0500, Pei Wang wrote: > Of course, most of the limitations of NN can be avoid by generalizing > the concept to such a level. However, at the same time, such a general > notion does not support the claims of advantages of NN, either. How
Biological cognition is based on network processing, too. > can someone argue that this concept can compete with the classical > symbolic model of cognition? Can semantic network be counted as a kind Because you're reading this message in realtime NNs are clearly a quite powerful model. In fact, since we don't have any human-equivalent symbolic processing systems, the burden of proof is reversed. > of neural network? Once again, when the extension of a concept gets > larger and larger, its intension gets smaller and smaller. The space of automata networks is vast and almost utterly barren. The probability of hitting a fertile spot with an educated guess is basically nil. Of course biological tissue processing does some fancy tricks ANNs can't yet. > > As for whether NN's are the best architecture for AGI right now, I > > agree that they are not. My reason is that no one knows how complex, > > abstract knowledge can be efficiently, adaptably represented in NN's. But you don't have to understand the representation in order to be able to build very successful, superhuman intelligences. The evolution process is not sentient, and rather straightforward, yet it can produce at the very least human-level intelligence. > > I'm sure there *is* a way to do it, but since no has discovered it yet > > (via either mathematical theory, computational experimentation, or > > neuroscience), basically in my view NN-based AGI is a non-starter. > > Agree. Disagree. It is curious why 'mathematical theory' is supposed to be useful for AGI. There has been hitherto no attempt to map a parameter space, nor is there sufficiently powerful hardware to execute 10^9 node networks in realtime. > > Once this conceptual problem is solved, then NN-based AGI may become a > > good strategy. However, I still think it will wind up being worse > > than probabilistic logic based AGI, at least on von Neumann computers > > (or networks thereof), because NN architectures don't take good There won't be any AGI on any current memory-starved, few-threaded hardware. > > advantage of the particular strengths of von Neumann computers (which I presume you're including Harvard along with von Neumann. I do not see many advantages or strenths of sequential machines for AGI, I must admit. > > can do very precise operations very quickly in serial, a quite > > different strength from the human brain, and one which is in large AGI is not cryptography. > > part wasted when computers are used for NN computation) > > Don't fully agree, but that is a separate issue. -- Eugen* Leitl <a href="http://leitl.org";>leitl</a> http://leitl.org ______________________________________________________________ ICBM: 48.07100, 11.36820 http://www.ativel.com 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE ------- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
signature.asc
Description: Digital signature
