Re: Intelligence and Statistics

From: AlphaOmega2004 (OmegaZero2003_at_yahoo.com)
Date: 09/28/04 

Date: Tue, 28 Sep 2004 13:13:44 -0700

"Bill Modlin" <modlin1@metrocast.net> wrote in message
news:Z9Sdnbx_rYkGPcTcRVn-qA@metrocastcablevision.com...
>
> "David Longley" <David@longley.demon.co.uk> wrote in message
> news:$g3E32BNcoVBFwSJ@longley.demon.co.uk...
> > In article <V4CdnW1g25Ww2MvcRVn-iQ@metrocastcablevision.com>,
> > Bill Modlin <modlin1@metrocast.net> writes
> > >
> > >"David Longley" <David@longley.demon.co.uk> wrote in message
> > >news:aSptbBH1UbVBFwlZ@longley.demon.co.uk...
> > >> In article
> > >> <AI2dnW4PM5WIMMjcRVn-oQ@metrocastcablevision.com>,
> > >> Bill Modlin <modlin1@metrocast.net> writes
> > >> >
> > >> <snip>
> > >> >I know Glen has a problem even agreeing that a computer
> > >> >computes or a calculator calculates, as he sees the words
> > >> >as originally designating classes of human behavior and
> > >> >therefore as only metaphorically applicable to machines.
> > >> >
> > >> >But for the majority of us who have accepted the linguistic
> > >> >switch, both electronic circuitry and neural circuitry can
> > >> >be seen to compute functions. Not metaphorically, but in
> > >> >fact. That's what they do. That's their job.
> > >> <snip>
> > >>
> > >> But it's this "linguistic switch" which is the whole
> > >> (philosophical) point!
> > >>
> > >> We know *you* have made this "switch", but that's what's
> > >> been explicated and criticised. You don't appear to
> > >> appreciate why it's an error, and looking for endorsements
> > >> from other similarly deluded mentalists is not the way to
> > >> find out. You have to look into that field which I have
> > >> repeatedly said is your bete noir. What you have to do is
> > >> stand back and ask, if what you were doing was in fact
> > >> sound, why has it proven itself to be so pitifully
> > >> inconsequential technologically?
> > >>
> > >> It isn't just me that's saying this, it's also some of the
> > >> very people who played a large hand in starting this wild
> > >> goose chase off!
> > >> --
> > >> David Longley
> > >> http://www.longley.demon.co.uk/Frag.htm
> > >
> > >Let me understand.
> >
> > I don't think you have any intention or hope of understanding.
> >
> > >You suggest that it makes me a deluded mentalist that I think
> > >of the circuits that I have designed and the programs that I
> > >have written as computing and calculating, as implementing
> > >various algorithms and functions.
> >
> > I suggested no such thing.
> >
> > > You say that the things I and other professionals
> > >in computing fields have done have all proven "pitifully
> > >inconsequential technologically", all because our terminology
> > >is, according to you, unsound.
> >
> > No, what I said was something else and you're deluding yourself
> > by making out otherwise. That may well just be your way of
> > coping with the criticism, which I think should stop you, like
> > others here, in your tracks. That's quite a tall order, but I
> > reckon one you and others here should try to face up to.
> >
> > > We cannot do anything consequential if we have the temerity
> > >to think that computers, programs, and circuitry can
> > >implement various functions and can be said to compute various
> > >results.
> > >
> > > If that's really the whole philosophical point... no. You
> > >cannot possibly believe that. But there seems to be little
> > >leeway to interpret this post otherwise. I don't know what
> > >to say to you.
> > >
> > >Bill
> > >
> > >
> > No, that is not what I said, and I have said a lot. Read
> > through the quote at the end of the post below. I reckon it
> > should, by now, be rather clear to anyone who reads what I've
> > posted here over the years, what my philosophical point is.
> >
> > Your above paragraphs illustrate *how* you delude yourself in
> > basically the same way that the deluded "cognitive
> > neuroscientist" does elsewhere when s/he misappropriates the
> > language of neurology to promulgate whatever folk psychological
> > metaphysics s/he really wants to promulgate about "the mind".

Longley, shut the *** up already you idiot. We neuroscientists KNOW what
we mean by your so-called "folk-psychological" terms. You do not, or thinbk
they are terms that are intensional and therefore are inappropriate to the
task, because you do not know anything about this science. Nor do you know
anything about AI. What are you doing on this forum? We tell you over and
over that you are a fool and an idiot but you do not listen. Please go gain
some knowledge about this field. <sigh>

> > In both cases, the strategy is to make out to yourself
> > and to others that you're doing one thing where all the time
> > you're really doing something else. There are fancy names for
> > it, but it just comes down to sleight of hand or spin.
>
> David, please try to look at what I actually say.
>
> I'm talking about implementing functions to combine incoming signals from
things like arrays of
> photosensors, arrays of frequency-selective sound sensors, and various
other sorts of information
> sources, to produce new signals with more potential utility than those
with which we started.
>
> Where is the delusion? Where is the sleight of hand?
>
> These functions might be performed by many sorts of physical systems,
including but not limited to
> computer programs, electronic circuitry, neural circuitry, and collections
of tinkertoys. What
> counts most is not the medium in which a function is instantiated, but the
relationship between
> inputs and outputs: the function itself.
>
> Of course there are practical differences in the performance of different
implementations: neural
> circuitry is more efficient than anything we can so far construct. A
small mass of neurons using a
> few watts of power can in aggregate handle far more signal transitions per
second than our largest
> computers or electronic gate arrays. These practical engineering concerns
mean that we can never
> actually build a brain-equivalent machine out of tinkertoys, and probably
cannot build one today
> even with electronics. But we may at least try to work out what it is
we'd like to construct.
>
> The sheer scale of the processing performed by the brain, coupled with the
difficulty of observing
> precisely the operation of brain circuit elements in-situ, makes it hard
to see just exactly how the
> brain does whatever it does. We need further advances in technology to be
able to fully assess
> brain function directly. But in the meantime, we can try to suggest (and
implement on a smaller
> scale in electronic circuitry and programs) some functions that may be
able to account for various
> aspects of observable macroscopic behavior.
>
> I find it reasonable to propose that the compounding of a small number of
functions loosely
> characterizable as statistically driven unsupervised learning functions
can account for many aspects
> of animal (including human) behavior, and this discussion started when I
asked you if you would
> accept such a characterization.
>
> I'd like to engage you and others in a discussion of those functions, but
so far we seem unable to
> progress beyond quibbles over the words used.
>
> You say:
>
> > These subterfuges are usually exposed by pointing to their
> > empirical bankruptcy and a deflationary, accurate description
> > of what is really being done.
> >
> > David Longley
>
> Again I ask, where is the subterfuge? If you see my words as subterfuge,
please explain it to me.
>
> When I mentioned unsupervised techniques you asserted that there are just
two broad types.
> Specifically, you said:
>
> "As I've said many times before, "supervised" == regression
> and discriminant analysis, and "unsupervised" == cluster and
> factor analysis"
>
> ... which may be marginally acceptable as a gloss so far as it goes, but
there are many variations
> which can make huge differences in the practical effectiveness of an
implementation. I'm not sure
> where you place things like Fourier and spline transforms, HMMs,
clustering of random fragments, and
> Kohonen maps in your deflationary (or is the word dismissive?) summary,
just to mention a few.
>
> Neural circuitry does a lot of things, including many specialized tasks
dictated by evolved genetic
> control. Much of our behavior is formed around circuitry that is
basically hardwired to keep us
> alive and procreating, circuitry not much different from the circuitry in
simpler animals.
>
> We also have some possible behaviors beyond those of simpler animals. We
are able to detect more
> complex patterns and make associations beyond the reach of any cat or dog,
just as they are able to
> learn things impossible for an insect.
>
> I suggest that the mechanisms responsible for such extensions in
capability can be best understood
> by thinking of them as performing an unsupervised statistical analysis of
signals originating both
> in sensory inputs and in the operation of other parts of the organism.
This analysis results in the
> generation of new signals, which can then participate in further deeper
analyses, and also are
> available for association into behavioral patterns by the older more basic
mechanisms.
>
> Are you interested in discussing such processes?
>
> Bill Modlin
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>

Quantcast