| on the foolishness of natural language speaking |
[Apr. 4th, 2014|01:02 pm] |
to user yarikas, primarily.
suddenly i realized one interesting problem of natural language comprehension by machines.
one can not measure a software success by talking to a "talking machine".
make a trivial mental experiment: imagine two machines (perfected in conversations with humans) talking to each other.
it is fucking ENLIGHTENING experience.
u c?
what tectonic plates of context we involve in any however trivial chat!
we humans use all the life experience to make meaning of words.
and it is a human agent who makes meaning of any hypothetical conversation with a machine. we removed human agent, for a moment, and we end up with a complete "waste of bytes".
and there is no magic at all. no soul, no holy spirit, no divine breath.
the basis of our understanding is recognition of SIMILAR EXPERIENCES!!!
even in simple conversations we appeal to "shared memoirs".
even to specify objects of our speech, we call to the memory of our vis-a-vis, sometimes explicitly: "do you remember, that guy..." "do you remember we have been there..." "do you remember i told you..." -- otherwise, we do so implicitly.
every "mustard grain" of understanding depends on shared (or reproduced) experience. we speak of EVERY object, EVERY property, EVERY relation, implying our addressee has any experience with these. otherwise, we try to establish a relation of SIMILARITY between these objects we talk about and the objects we suppose our addressee has known. "imagine! it is as good as..."
we can talk at all ONLY because we live in basically the same physical environment
the difficult part is that the vast majority of these memories required for a however slightly meaningful conversation are constituted of our senses... they are physical rather than logical.
what is logical, on the contrary, can be easily reconstructed, therefore does not require to be memorized... that hard. am i close to the very essence of logic as opposed to experience?
so what?
Dear yarikas, i was talking about the mainstream of nowadays IT, that you have traumatized me with.
the "big data" fad.
what we must do IN THE FIRST PLACE (before any application comes to us) is to establish a method of looking for SHARED KNOWLEDGE among data nodes.
no more and no less. once we master this skill, everything become possible, otherwise nothing.
you can think of it as a "high level" relation of equivalence -- very basic of any meaning indeed. |
|
|
| Comments: |
1st things 1st
i asked u to post your task about dancing contest on my maillist, and i still want you to do so, thanx.
well, i talked about us, and we are "physically-based", it is a matter of fact. i can not be wrong about it.
yes, there are lots of AI in virtual realities, yet they are unable to understand us, and i exactly explained WHY.
and i concluded that this principal "incompatibility" gives us a vital clue to make them able to understand each other in their worlds.
what r u opposing to?
Task? Dancing contest??
I have rough memories about some task about dancing contest, but I think it was not my, and clearly can't remember what is it and who's it was. And when you asked me something about it.
Real opposition is: we don't need physical same environment. It could be electronic same environment, mem-based same environment and so on.
you blew my mind up with the tast of representing "dancing pairs" along their attributes such as: costume colour, scheduled position, earned score and stuff.
yes we need, at least for now.
we only mastered our language in the only physical reality.
everything beyound this reality is fucking UNEXPLORED for OUR language.
it does not imply, however, a prohibition for exploration of "Virtuality".
my point is that we (in order to master a language in another reality) must take the same steps we took in OUR reality.
and the correspondence between physical reality and IT's domain of knowledge is another issue.
a bonus issue :)
Right you are. The sence of words is a matter of context. Even if we are talking with machine we PROJECT our memories and experience on her words.
And, yes, SHARED KNOWLEDGE is the basis of conditional (conventional?) information - the degree of similarity?
and it is my "issue", i can not find any application for my "being right" :( | |
![[info]](http://lj.rossia.org/img/userinfo-lj.gif){kind=small}
![[info]](http://lj.rossia.org/img/userinfo.gif){kind=small}