MT:Can you think of a single analogy or metaphor, in addition, that is purely 
symbolic?
  Ben:I don't really understand your definitions of the terms analogy, metaphor 
or symbolic...


  Analogy: He eats like a pig. 
  He started posting in this forum like a bull in a china shop.

  Metaphor: The sky was crying.
  The quality of mercy is not strained, it droppeth as the gentle rain from 
heaven.
  We are not getting very far in this conversation.  ["getting v. far" is a 
physical journey metaphor applied to linguistic exchanges]

  Symbols: Abstract signs which bear no resemblance to their objects - words, 
algebraic/logic letters, morse code, numbers

  My argument is that productive analogies and metaphors are overwhelmingly 
though not exclusively graphic and image-based. I've still to see any examples 
of productive analogies purely derived from symbols (although I'm not denying 
them). I guess:  "112233 is like 111222333" is a purely symbolic analogy - but 
it is also a METAsymbolic analogy - IOW an analogy directly about symbols.  I'd 
like to see examples of networks of symbols of the kind beloved by Minsky & 
others (& perhaps Vlad & Cyc) producing purely symbolic analogies of value 
about the world  

  That brings us to your:
  Ben:"As a very simple example, if Novamente's evolutionary learning engine 
has learned a procedure to play fetch, Novamente's inference engine may allow 
it to utilize this procedure as guidance in learning to play hide-and-seek, so 
that the system may learn to play hide-and-seek more easily than if it had 
never learned to play fetch. This sort of inferencing however involves a number 
of probabilistic inference rules all acting together in a coordinated way, it's 
not a single inference-step that I can just paste into an email."
  Sounds like a good analogy. If it can play fetch, it can play hide-and-seek. 
[And exactly the sort of thing that a true AGI must do - absolute heart of AGI].

  The question, wh. I wouldn't think that complex to answer,  is: how did it 
connect the action/activity of fetch, to the activity of hide-and-seek? (And 
that connection surely has nothing to do, essentially, with probability). [My 
first impression is that that's a graphic/imaginative analogy. It certainly 
would be, I'd argue, if a human or child made it]. Did the system/pet draw the 
analogy itself, without being told?

  P.S. COMMENT: This - going from fetch to hide-and-seek - is the first 
specific problem example I can remember in this forum of the general problem of 
AGI, i.e how do you go from learning and/or solving one activity, to learning 
and/or solving a totally different kind of activity. And you're not going to 
discuss it?!!!  You/we/everyone should be discussing this sort of thing all the 
time! 

  In the final analysis, to use another imaginative analogy, it's not much use 
talking endlessly about the wonderful engine of your guided missile, if you 
never talk about what it can actually hit - in this case, the actual problems 
that Novamente and other AGI's can solve. And if you don't, you're wasting a 
lot of valuable collective brainpower (worth much more presumably than your 
personal team).






-----
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=52937691-13b628

Reply via email to