josh zeidner on Fri, 21 Jun 2002 10:01:32 +0200 (CEST)


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: <nettime> CCTV/face recognition



  Hi,

    I recently got involved with a project using face recognition, and I
can tell you from personal experience that the technology is simply not
there.  I call it acedemic vaporware (which is indicative of the general
degradation of scientific research in acedemia).  The face recognition
people have very complicated technical justifications for why it works,
however, it is simply not practical to use.  In order for most of these
systems to work, you must have a very constrained environment (ie. you
must have the background image beforehand).  Its performative state is
very similar to the speech recognition technology (which if it did in fact
do something useful would probably be in widespread use).

    At the edge of computer science exist these so called sciences of
'complexity' and 'chaos', which are trying to find a qualitative
understanding of *information*, *pattern*, and *structure* (a talented and
dedicated scientist working in this area is J. Crutchfield).  These three
terms( and many other related ones), still elude formal definition and
certainly have no quantifiable formulae.  We are very far from attaining a
scientific understanding of these things (although many would disagree
with this statement, Crutchfield would probably be among them).  As with
all scientfic 'progress', any real change will involve a major rethinking
of the means of discovery itself.

    What is also interesting is the existential aspects of this sort of
thinking.  These scientists all attempt to find a mathematical
understanding of the processes of the human mind.  Some take the classical
AI approach, but the popular modern paradigm is the 'connectionist'
approach.  Connectionism attempts to isolate the local interactions of
simple units to model the complex behavior of the whole.  Neural Networks
are the cornerstone of this philosophy.  I believe that there are many
underlying conceptual problems that are not addressed by the scientific
world concerning these areas of inquiry.  Is the human mind a complex
computer?  This has yet to be proven, and it may never be proven.

   It is well known that honey bees use a very simple language.  They use
a complex dance to communicate the location of food relative to the hive.  
This language is understandable to humans, and we are capable of
describing this simple language in terms of our own language (for instance
I can read a book on how this language works).  So we have two languages,
our own and the honey bees'.  Now would it be possible to translate this
paragraph into this honeybee language?  Certainly not.  Now humans also
speak a finite language, with total parameters beyond our grasp, much the
same way a bee is capable of using its language without understanding
fully how it works.  It is the same with human language, we can only
understand its features, the total comprehension of which must be beyond
our faculties.  Perhaps a hypothetical divine being would be able to
formulate statements about our language, but there is no way we could
understand them.  Such is the limits of artificial intellegence and
related areas such as pattern recognition.

 
  -josh

--- Forced Entertainment - Tim <tim@forced.co.uk>
wrote:
> a news story and a feature last week in The Guardian
> (UK) about face
> recognition software, CCTV surveillance and
> "exaggerated claims" by
> maufacturers Visionics/FaceIt. Esp beautiful
> explanation of the system's
> failure to spot a reporter whose face had been
> scanned for the system:


[....]




#  distributed via <nettime>: no commercial use without permission
#  <nettime> is a moderated mailing list for net criticism,
#  collaborative text filtering and cultural politics of the nets
#  more info: majordomo@bbs.thing.net and "info nettime-l" in the msg body
#  archive: http://www.nettime.org contact: nettime@bbs.thing.net