>> A question to you who have a certain experience
>> with this 'breeding' practise: can you still follow the (textual)
>> output of your algorithms - I mean do they "make sense"?
>
>Not really - no. They can contain large amounts of code that does not get
>executed, or is irrelevant. This can be minimised by scoring code length, but
>it's generally there to some extent. This can be seen as being "junk dna" and
>may be important to have around - but more on the dodgy biological parallels
>below. They can also employ ways of doing things that are truely arcane and
>hard to follow.
I see that this is a quite unique thing: a programmer will normally
try to get rid of this type of idle-code as fast as possible. But for
live coding as well as for gp it makes a big difference how much one
has to do to activate it again - the difference might be one
character. Similarly, sometimes it's not good to optimize code (too
early, but in our case, maybe ever) because certain ways of changing
it only come up due to its present form. On the other hand,
obfuscated code is hard to "think in", so it also makes certain steps
impossible.
>> Another side of this is that the terminology of genetics (and
>> immunology) is full of metaphors with a very political taste,
>> usually neo-liberal: competition, survival, defence etc. This
>> terminology also naturalizes certain behaviours - like hereditary
>> descendence - which should nearly automatically be challenged by a
>> process of live coding. How does literature and nature metaphors
>>"interbreed"?
>
>The use of biological language in these cases is more than little misleading,
>as I am constantly reminded by my geneticist girlfriend. There is really no
>parallel between the methods of GP, genetic algorithms or any of these
>techniques and real life genetics. They take a certain aspect of adaption and
>steal a lot of the terminology for processes that we really understand little
>about yet even in current biology.
>
>It would be great to be able to seperate this language between the biology,
>computer science and social politics, as they really share very little
>meaning. It's really important to get this message through to people, I think.
I can't agree more. Maybe it is this language to a certain degree
that allows people to think in such ways (Maybe good that it is so
obvious). But generally it is also important to keep apart a broad
statistical way of thinking (selection pressure etc.) from a more
individual event structure, like a new idea or simply something that
happens.
>> Writing programs on the fly, I got the impression that the relation
>> between statistical and deteministic properties of the algorithm
>> gets quite relevant. If there is too much randomness, modifying the
>> program does not yield any new insight in the process. But this is
>> the case for a certain durational range - very fast changes are
>> overlooked easily enough. I would guess that gp + ip is really good
>> for these ranges.
>
>I'm not sure I follow you, but it would be possible, if only viewing the best
>code for the latest poplation, to watch the development, and improvement in a
>way that would be almost understandable I think - maybe not in specifics, but
>in a general sense.
Yes, what I meant is that, say if you consider pink noise, you cannot
immediately understand what is going on. But because it happens so
fast, it does not take long to pick out the typical characteristic of
this noise. Playing back pink noise at, say, a sample a minute, it
would be nearly impossible to reason about. The selection rules (be
it the ones you use when selection manually or the automatic
selection rules) act as part of the audible synthesis process. So it
would be interesting under what conditions one finds it possible to
reason about their logic.
>The thing that GP and livecoding share in opposition to the traditional act of
>writing software, is that they are more concerned with the program being part
>of a dynamic process. Nothing is ever finished.
most definitely.
--
.
Received on Wed Apr 13 2005 - 14:22:28 BST