Philosophical Foundations

[Sergio Navega]

Brendan Macmillan wrote:
>
>Sergio Navega wrote:
>
>> a) Recognizing regularities
>> (perception of reocurring patterns)
>> 
>> b) Grouping of regularities according to several similarity criteria
>> (conceptualization and categorization)
>> 
>> c) Development of causal models
>> (rules, theories, formalization)
>> 
>> d) Exploring uncharted territories
>> (creativity)
>
>Surely "recognising regularities" is part and parcel of compression.
>

Agreed. 

>"Grouping of regularities" is based on recognition of some common theme 
>between them - ie a regularity. see (a).
>

In a way, categorization could be seen as recognition of regularities
over the set of recognized regularities. Things get a little bit more
complicated when we remember that categories often group things that
are *similar*, not equal. The concept of *similar* is something that
often puzzles us and most of the time is utterly dependent on 
context. Imagine that you cut the trunk of a tree. This is a typical 
exemplar of a stump. Now if I put a table next to that stump, we
suddenly get that stump as being another thing: a chair. 
The same object may change its category just by switching contexts.
It is this dynamical aspect that categories must group.

>A causal model which gives better predictions will also give better
>compression, which it also recognises regularity. see (a).
>

Yes, a causal model can favor compression but I don't see a direct
link between it leading to better compression just because it is
better. The quantum model of the atom is certainly better than
Bohr's model, but it can hardly be considered as being "more
compressed".

>However, "exploring uncharted territories" is more an attitude and an
>application of intelligence rather than intelligence itself.
>

I could agree with you if we look to creativity as the generation of
innovative and useful ideas, as seen from an "external" position.
But I propose an extension to the meaning of this word, to encompass 
the exercise of generation of information in a way that may not
be directly perceptible "from outside". My concept is part of 
a model in which creative combination of concepts by the agent
is used to try different routes in one's mind. These "mental
experiments" are used by the agent to try to make sense of
the huge quantity of information it is subject to. Most of what
I think about creativity comes from Douglas Hofstadter's "Fluid
Concepts and Creative Analogies".

>
>A counterview: good communication requires redundancy - the
>opposite of compression.  Great speeches, for example, roll waves and
>waves of the same meaning repeated and reinforced over and over again.
>A person who is able to communicate well is certainly exhibiting
>intelligence; and part of that skill lies in the redundancy they
>introduce.
>

Interesting comment, I agree. There's something beautiful in 
repetition, redundancy, something that can even be traced back
to our emotional origins. I have a pretty recent case to illustrate.
I used to listen only to a specific kind of music: a classic composer
of this century, Shostakovich. I love his symphonies. Then suddenly
I discovered another one I appreciate: Steve Reich, a modern composer
known to be a member of the "minimalist" school of classic music.
Minimalist music is extremely redundant (a good example is
Reich's The Desert Music). It is this redundancy with just small
variations over time that attracts me so much.

>Of course, one could say that the process of this skill itself can be
>explained with compression, regardless of the nature of the result.
>
>
>And now, a theme of my own:
>The important problem in AI by compression is the ability to recognise
>patterns, of any form, that have not been prespecified by the
>"programmer". Ie a language in which to describe the pattern. Any
>Turing complete language is up to this task.
>

Recognition of patterns not specified by the programmer is what 
I think is one of the fundamental points behind anything that
aspires to be called intelligent. Too bad that most AI systems
of today don't use this as a basic premise.

Regards,
Sergio Navega.