Alan, I think you are trying to resolve a very high level problem. Would not it be simpler to start resolving basic problems like an animal have. Hunger, thirst, .... and try to build a system to model a simple animal. After this we can talk about more sophisticated systems ??
You are saying that : "AI would start out by expressing its listening behavior" But a recently born child do exactly the contrary, he does not listen, he cries... he is expressing himself....he already have his simple behaviour.... My approach is to first try to resolve simple problems to validate the model and then try to reproduce our educational process to , step by step, learn our world.... Olivier -----Mensagem original----- De: Alan Grimes [mailto:[EMAIL PROTECTED]] Enviada em: sexta-feira, 17 de janeiro de 2003 20:01 Para: [EMAIL PROTECTED] Assunto: Re: RES: [agi] Unmasking Intelligence. Olivier - (JConcept / Cetip) wrote: > But why is it necessary to reproduce our internal brain way of working > to build an intelligent system ? One one level, it would be very advantageous to replace biological architectures with much more powerful/scalable/reliable/efficient approaches. On the other, I don't see how the basic layers can be disposed of with such prejudice. I don't say that they all must be replicated perfectly but rather that in any sucessful AI entity you will find analogous units of functionility, even though there may be of different structures and purposes... The intelligent compiler problem I mentioned attempted to illustrate this point. While a cortex or a cognitive matrix might be necessary for higher intelligence, it is not _SUFFICIENT_ to produce useful behaviors... Such a system only exists as a part of a cybernetic system (running a cybernetic program) that directs the cognitive matrix, which represents the raw faculties, towards the things we need it to do. Such a cybernetic program in this case, would configure itself based on the command line arguments and, by means of a complex interaction with the cognitive matrix, generate the behavior of either optimizing or translating a program. In a more general intelligence, this cybernetic program would be a meta program which directs the matrix through the process of identifying and solving problems towards the goals dictated by the user... The first such behavior would be to await input... So the AI would start out by expressing its listening behavior... The user would say something like this: "I would like us to begin a new project." AI: "Okay." User: "I want to begin desgning my next body, please do the necessary research." AI: "Polling internet databases........ I think I will require a 60 petabytes of memory and a 2048 Q-bit quantum core for this problem." [during this process, the AI expresses a large number of complex behaviors as it educates itself on the necessary subjects...] User: "Okay, I give you permission to purchace the necessary materials and construct that equipment.". etc.... -- I WANT A DEC ALPHA!!! =) 21364: THE UNDISPUTED GOD OF ALL CPUS. http://users.rcn.com/alangrimes/ ------- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] ------- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
