Fundamental Compressionist Philosophy.

[Detlef Morgenstern]

Hello Andrew,

Sorry I let you wait for an answer. Must keep a foreground job
running which drains a lot of %CPU. What I want to reply to is a bit
scattered about several documents and mails of yours and other CasC
people. So I decided to pack the essence of what I want to say into
this document - not heavily citing others.

-- On Academic Style --
I see no need for endless apologizes for some contribution not being
perfectly academical. In my opinion, good will, drive and vision
count a lot. And, of course, mental self-discipline. Compare the
academic approach to triangulation of an unknown land. But you must
leave the shore to discover another continent.

-- On Switches --
>From a superficial view, there is no news in what you say about
switches. Zoom into any functioning computer, and you will find
boards with integrated circuits (IC) on them. ICs mostly contain
transistors, i.e. switches. Hence, the question is not 'to switch or
not to switch'. I think, there is a second level question in your
switch idea, which I would circumscribe as this:
'How can we build a Universal Computing Machine (UCM) arranging
switches in an advantageous manner?'

It is obvious, that the Turing Machine (TM) has several
disadvantages:
- It is serial. Which makes it slow.
- It requires a very complex translator to transform simple jobs into
a TM program (tape).
- It reqiures a very, very complex translator to transform complex
jobs (as the above translator) into a TM program. This complexity
recursion forbids doing something demanding on a TM.

Which UCM (topology) do I consider 'advantageous'?
(1) It provides a lightweight translation for basic computing
operations as
  (a) pattern matching
  (b) pattern generation
  (c) information dispatching
or - even better - it **is** an implementation of a pattern
matching/pattern generating device.
(2) It works massively parallel.
(3) It supports (inductive) compression (see below "On Inductive
Compression").

-- On Computing Atoms --
Is an isolated switch a computing device? Two switches together?
Three??
>From which complexity level on would we say, "Yeah, it computes!"?

Consider a living cell. A small entity of which one could say 'it
lives'. Of course, you can split a cell into molecules, the molecules
into atoms, the atoms into sub-atomic parts, and so on. But there is
a level beneath wich the 'essence of life' gets lost.

For me, an isolated switch is no computing device. It lacks the
'essence of computing'.

What is this essence?
- Can detect patterns.
- Can generate patterns.
- Can move patterns.

A computing atom must reflect these capabilities.

I try to map all the other computing around
- clause
- function
- predicate
- quantor
- resolution
- unification
- deduction
- abduction
onto the above three basic capabilities. And I think, this can be
done.

I suggest the following benchmark measure for the 'minimality' of a
UCM.
Define a set of tasks which can be solved only if a UCM is actually
universal.
Implement the benchmark set on each candidate machine.
How many switches does each candidate machine consume?

(A cell on a TM tape is a memory cell, it can be implemented as a
flip-flop - which consumes n switches. Any computing implementation
(be it physical or emulated) can be broken into an integer number of
(physical or emulated) switches.) 

-- On Essence --
Something has been brought to its essence, when it cannot be
compressed any further - still doing the same.
Essence is about acting, not about being.
It is about producing patterns, not being patterns.

-- On Inductive Compression -- 
If we want to 'mechanically' distill the essence of something, we
must compress it. But there is a strict constraint for this
compression: After completing compression, the compressed instance of
our something 'must work as before'. Which means: When it is
confronted with a given set of patterns, it produces the same
patterns as it would have done before compression.

It can be checked whether an arrangement of switches does the same as
another one. ('The same' means: It reacts with the same output to a
given input.)

If one of the arrangements consumes ten times less switches compared
to functionally equivalent ones, this one will be more abstract. And
this will reflect in its structure.

Inductive compression is accompanied by emerging of abstract
structures. It forms abstractions. And it requires nothing more than
a process on a nearly physical level: (controlled) compression.

I am sure, we will reach THE breakthrough following this direction.

-- On CasC --
You see, I separated induction and compression from the essence of
computing. This is not the CasC mainstream. But I see no problem in
this kind of separation. I get a lot from this CasC discussion, and I
hope, I can give some useful ideas in return.

Best wishes,

Detlef




__________________________________________________________________
Do You Yahoo!?
Gesendet von Yahoo! Mail - http://mail.yahoo.de