meta information operations
Max
NARS(this is mainly the way it could be used).
For the purposes of the analytical engines and cooperations of several
NARS instances managed by one central NARS, could be really useful to
naturally operate with meta information of the inferred facts in NARS.
For example to check inbound information according to the common sense
rules.
Pei Wang
please provide a more detailed explanation.
Thanks!
Pei
> --
> You received this message because you are subscribed to the Google Groups "open-nars" group.
> To post to this group, send email to open...@googlegroups.com.
> To unsubscribe from this group, send email to open-nars+...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/open-nars?hl=en.
>
>
Max
I try to provide the example:
If we are going to check the inbound information via some rules:
using meta syntax:
<<(*, inbound.confidence, 0.8) --> greater_than> ==>
^we_can_trust_it>.
or if we want to check that NARS was able to infer some fact in some
range of NARS cycles:
<<(*, some_fact, Self, number_of_cycles) --> ^could_infer> ==>
we_are_good>.
I guess these two examples of the meta information could be used to
manage the NARS inference process in native way.
Possibly this could let us construct the system with what ever logical
abstraction layers we need, to implement for example learning to learn
processes.
Thank you,
Max
Max
I try to provide the example:
If we are going to check the inbound information via some rules:
using meta syntax:
<<(*, inbound.confidence, 0.8) --> greater_than> ==>
^we_can_trust_it>.
or if we want to check that NARS was able to infer some fact in some
range of NARS cycles:
<<(*, some_fact, Self, number_of_cycles) --> ^could_infer> ==>
we_are_good>.
I guess these two examples of the meta information could be used to
manage the NARS inference process in native way.
Possibly this could let us construct the system with what ever logical
abstraction layers we need, to implement for example learning to learn
processes.
Pei Wang
Now I see what you mean --- your "meta information" is closely related
to the "Self-monitor and self-control" operations introduced in
http://code.google.com/p/open-nars/wiki/OpenNarsOneDotFour
See the following.
On Mon, Aug 30, 2010 at 5:27 PM, Max <max.t...@gmail.com> wrote:
> Ok,
> I try to provide the example:
> If we are going to check the inbound information via some rules:
> using meta syntax:
> <<(*, inbound.confidence, 0.8) --> greater_than> ==>
> ^we_can_trust_it>.
The operator "^believe" will do something similar: when the
truth-value of judgment J is higher than a threshold, it will be
converted into a binary statement "I believe J", here "I" means the
system itself. It is not limited to inbound judgments.
Whether the system will trust an information source is a different
matter --- it mainly depends on the system's past record, rather than
in its confidence.
> or if we want to check that NARS was able to infer some fact in some
> range of NARS cycles:
> <<(*, some_fact, Self, number_of_cycles) --> ^could_infer> ==>
> we_are_good>.
Yes, the system will gradually develop knowledge about what it can/cannot infer.
> I guess these two examples of the meta information could be used to
> manage the NARS inference process in native way.
> Possibly this could let us construct the system with what ever logical
> abstraction layers we need, to implement for example learning to learn
> processes.
Agree.
Pei
Max
I have run through the document, it is really helpful.
Could you please crate some further examples of the Self-monitor and
self-control operators.
On 31 авг, 00:37, Pei Wang <mail.peiw...@gmail.com> wrote:
> Max,
>
> Now I see what you mean --- your "meta information" is closely related
>
> See the following.
Pei Wang
> Thank you.
>
> I have run through the document, it is really helpful.
> Could you please crate some further examples of the Self-monitor and
> self-control operators.
I surely will, but probably not soon --- I'm super busy at the
beginning of a semester.
Self-monitor/self-control is a huge topic that needs a lot of work.
For example, one obvious question is: what is the minimum set of
operation needed for human-level self-monitor/self-control (aka
"consciousness")? "To be able to completely rewrite ones own source
code" was proposed by some people, but to me, that is a very bad idea.
Intelligent systems should be much more conservative than that. Taking
human beings as an example: we cannot even erase a specific memory (in
normal situations), and I believe there is a reason for that operation
not to be found by evolution.
Pei