Pile system (2nd wave Cybernethics)

[Faré]

I've been recently contacted by Peter Krieg from the Pile System www.pilesys.com

TUNES is more like 1st wave and a half ("symbolic" stuff, but with
humility on the results and faith in external evolution to make the
system better). These guys are more like 2nd wave: faith that you can
directly program a non-symbolic program that can then evolve
internally.

PS: as for me, I'm more overworked than ever in my job at ITA
software, but this is rather good, as I am becoming a better
programmer and manager, and get to work fully in Lisp and partly on
distributed systems.

[ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ]
So that there be reality, there must be an observer.
"I am, therefore someone thinks."

[Oleg Cherevko]

Post by Faré
I've been recently contacted by Peter Krieg from the Pile System www.pilesys.com
TUNES is more like 1st wave and a half ("symbolic" stuff, but with
humility on the results and faith in external evolution to make the
system better). These guys are more like 2nd wave: faith that you can
directly program a non-symbolic program that can then evolve
internally.

Tanks for the useful reference.
Don't you think that the basic Pile building block (binary relation from
binary relations to binary relations) is strikingly similar to what
Brian Rice tried to do with his Arrow System?

[Oleg Cherevko]

Post by Oleg Cherevko

Post by Faré
I've been recently contacted by Peter Krieg from the Pile System www.pilesys.com
TUNES is more like 1st wave and a half ("symbolic" stuff, but with
humility on the results and faith in external evolution to make the
system better). These guys are more like 2nd wave: faith that you can
directly program a non-symbolic program that can then evolve
internally.

Tanks for the useful reference.
Don't you think that the basic Pile building block (binary relation
from binary relations to binary relations) is strikingly similar to
what Brian Rice tried to do with his Arrow System?

... or strikingly similar to most other relational systems?

No.
Perhaps I should clarify my terminology a bit.
A binary relation R is an association between elements of two sets S1
and S2. In fact, R is a set of some tuples (s1, s2) where s1 is taken
from S1 and s2 is taken from S2. Now, if we consider rather special case
of S1 == S2 == ST, where ST is a "set of tuples" and consists of the
very tuples that constitute the relation R we will get the main
reflexive structure behind Pile.

As far as I remember, this is similar to Brian's arrows that have other
arbitrary arrows as their endpoints (i.e. an arrow points from one arrow
to another arrow).

[Kyle Lahnakoski]

Post by Oleg Cherevko

Post by Oleg Cherevko
Tanks for the useful reference.
Don't you think that the basic Pile building block (binary relation
from binary relations to binary relations) is strikingly similar to
what Brian Rice tried to do with his Arrow System?

... or strikingly similar to most other relational systems?

No.
Perhaps I should clarify my terminology a bit.
A binary relation R is an association between elements of two sets S1
and S2. In fact, R is a set of some tuples (s1, s2) where s1 is taken
from S1 and s2 is taken from S2. Now, if we consider rather special
case of S1 == S2 == ST, where ST is a "set of tuples" and consists of
the very tuples that constitute the relation R we will get the main
reflexive structure behind Pile.
As far as I remember, this is similar to Brian's arrows that have
other arbitrary arrows as their endpoints (i.e. an arrow points from
one arrow to another arrow).

Thanks, but my interest is not in the reflexive/reflective core of Pile;
like you mentioned it has been done. Putting objects (arrows) into a
database(relational store) has also been done.

What has not been done (as far as I know) is how to solve the
lack-of-spatial-locality problem that comes from essentially using a
database for storing those relations. Direct memory-addressed objects,
despite all their drawbacks, are often lucky to be physically close to
objects it references..This closeness reduces cache misses. This
closeness does not naturally happen in a database; where a single object
can be scattered over several relations, and so too are the objects it
references. Without fine control over how relations are stored; your
data access speed is restricted to the speed of the physical storage
capable of holding the whole program image. This usually turns out to
be main memory speeds which is an order of magnitude slower the the L1
cache.

If someone on this list has seen how to get the benefits of relational
representation along with the benefits of localized data, then please
pass me a link.


Thanks

[Kyle Lahnakoski]

Post by Oleg Cherevko

Post by Faré
I've been recently contacted by Peter Krieg from the Pile System www.pilesys.com
TUNES is more like 1st wave and a half ("symbolic" stuff, but with
humility on the results and faith in external evolution to make the
system better). These guys are more like 2nd wave: faith that you can
directly program a non-symbolic program that can then evolve
internally.

Tanks for the useful reference.
Don't you think that the basic Pile building block (binary relation
from binary relations to binary relations) is strikingly similar to
what Brian Rice tried to do with his Arrow System?

... or strikingly similar to most other relational systems?