On 10/15/07, Linas Vepstas <[EMAIL PROTECTED]> wrote: > > Yes, of course, children learn to reason from concrete to more abstract > levels, and lawyers, engineers and mathematicians working at a > particularly abstract level. The concrete levels are indeed > grounded in sensory input. > > However, once one has actually learned how to think abstractly, > its not obvious to me that sensory grounding is needed; and indeed, > trying to touch back to the grounding can prevent one from > making the next leap of abstraction. >
It's an interesting problem. Essentially, system's experience consists of (1) direct sensory input and (2) information about internal concepts that were activated. Many tasks require persistent support of 'context', that is collections of internal concepts which remain active for some time while context is relevant, independently from sensory input, and reasoning 'in context' largely depends on these internal concepts. But if some portion of internal concepts becomes completely independent of input/output, it can turn on life of its own. How does system develop itself based on real world, and not such hallucinations? There's probably a developmental stage where system trains sufficient internal 'OS' which manages internal abstractions, sensory input/output and their interaction (by appropriate layering of reflective images of concepts and symbol manipulation patterns), but in order for this stage to be achieved, influence of internal/external experience should be weighted very carefully, hopefully it will only be required for relatively simple internal 'pre-OS' organization. -- Vladimir Nesov mailto:[EMAIL PROTECTED] ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=53870003-9b608b
