On Wednesday, May 17, 2000 10:27 PM, Stevan Harnad
[SMTP:harnad@coglit.ecs.soton.ac.uk] wrote:
> On Wed, 17 May 2000, Lorincz, Andras wrote:
>
> > sh> For toilers, the knowledge is
> > sh> grounded in having learned the features of the edible mushrooms; so the
> > sh> toilers can forage alone; for the thieves, the only "feature" of an
> > sh> edible mushroom is the vocalization of the toilers: "edible". So when
> > sh> there are no more toilers, the thieves cannot eat.
> >
> >la> I buy this.
> >
> > sh> The solution is always to have the bottom-level categories learned by
> > sh> toil, by everyone. The categories acquired by "theft" need to be
> > sh> higher-order categories, describable by (e.g., boolean) combinations of
> > sh> the names of the lower-order categories, as in the Cangelosi/Harnad
> > sh> paper.
> >
> >la> I do not see this if the thieves can "associate" other features to
> >la> the vocalization of the toilers: "edible". On the other hand, if
> >la> "association" qualifies as "toil" then I see it. However, this seems
> >la> to contradict with your final conclusion:
>
sh> No contradiction. "Toil" is learning the hard way, by active trial and
sh> error, with corrective feedback (from the consequences of making
sh> errors). The toilers learn by trying to eat mushrooms, and getting sick
sh> if they eat inedible ones (or going hungry if they fail to eat edible
sh> ones).
sh>
sh> They thereby learn (the hard way) the mushroom-features (sensory) that
sh> distinguish the edible and inedible ones.
sh>
sh> As long as there are toilers around who vocalize (correctly), the only
sh> "feature" a mushroom needs to have in order to make it edible is the
sh> auditory feature, which is a toiler nearby, vocalizing "edible." Take
sh> away the toilers and you take away that feature.
sh>
sh> Now you are right that in eating the mushrooms that have the (auditory)
sh> feature "edible" and not-eating the mushrooms that lack the (auditory)
sh> feature "edible," it is possible that the real (visual) features are
sh> being passively learned by differential association alone, and with
sh> very easy features (such as black/white), that might be enough -- but
sh> for those trivial feature-learning problems, there is not much
sh> difference between toil and theft in the first place.
sh>
sh> For nontrivial features, requiring nontrivial learning (e.g.,
sh> chicken-sexing), active trial-and-error toil (reinforcement learning)
sh> and feedback are needed, and not mere passive exposure and
sh> association.
sh>
sh> By way of an intuitive example, if a school-teacher had 30 students in
sh> every class, and taught 7 different classes a day (and never for longer
sh> than a few months), and every child every day wore his name written on
sh> his forehead -- and every day the teacher had to interact with each
sh> child several times by name -- what do you think would be the "hit
sh> rate" of that interaction if, on the last day of class, all the
sh> students came in without their forehead-labels?
Hit rate will be bad, no doubt, for an average person. For some,
however, it will be zero. I know people who can learn six pages
of the phone book with ten errors in one night. So this is not
the issue, the problem is somewhere else. The hit rate is not
important if we are after general concepts, or you should argue
why the hit rate is more than being merely a quantitative issue.
sh> That hit-rate would reflect the strength of passive association. (I
sh> don't think it would be very high! Facial features are hard enough to
sh> learn WITH active trial-and-error toil and feedback; passive association
sh> would not stand much of a chance.)
sh>
sh> Not zero chance, I agree; but in the context of what it would take to
sh> sustain language, with nontrivial features and and categories to learn,
sh> close enough to zero as to be a nonstarter (in my opinion).
sh>
sh> (The chicken-sexing counterpart would be to sit passively by the
sh> assembly-line watching the grand-master examine the chickens and say
sh> whether they were male or female, for several months, and then to try it
sh> on their own -- compared to having done it on their own for all those
sh> months, and having gotten the feedback from the grandmaster each time
sh> only AFTER having tried to say which was male and which was female
sh> themselves.
sh>
> > sh> The solution is always to have the bottom-level categories learned by
> > sh> toil, by everyone. The categories acquired by "theft" need to be
> > sh> higher-order categories, describable by (e.g., boolean) combinations of
> > sh> the names of the lower-order categories, as in the Cangelosi/Harnad
> > sh> paper."
> >
> >la> unless you allow (as a special case) that "lower-order categories" could
> >la> be raw sensory information and not more.
>
> I don't know what "raw sensory information" means; but what I mean here
> is certainly sensory features (of the object -- not of its associated
> label: it is the label we are actively trying to learn to apply to the
> right objects).
>
> Describing something (with words/symbols) only helps if you already
> have grounded SOME symbols, at least, in something other than just more
> descriptions. Otherwise it's all just meaningless squiggles and
> squoggles. (That's the symbol grounding problem.)
>
> But I don't think mere passive exposure and association is enough to
> ground symbols either.
How about the first two words learned, say 'dad' or 'mam'?
BTW, my son is different. In his case the first word was 'auto'
and he meant it.
Regards,
Andras
Date: Thu, 18 May 2000 10:59:20 +0100 (BST)
From: Stevan Harnad <harnad@coglit.ecs.soton.ac.uk>
To: "Lorincz, Andras" <alorincz@matavnet.hu>
Cc: 'Stevan Harnad' <harnad@coglit.ecs.soton.ac.uk>,
Margareta Lutzhoft <b96marlu@ida.his.se>,
hlist -- Bereckei Tamas <Btamas@btk.jpte.hu>,
Bezeczki Gabor <bezecz@tina.iti.mta.hu>,
Angelo Cangelosi -- Angelo Cangelosi <angelo@caio.irmkant.rm.cnr.it>,
Angelo Cangelosi <angelo@soc.plym.ac.uk>,
Csak Laszlo <csak@mksh.u-szeged.hu>, Dudas Kata <dudas_kata@yahoo.com>,
Farkas Agnes <f_agi@hotmail.com>,
Fiser Jozsef <FISER@BCS.ROCHESTER.EDU>,
Gervain Judit <h633301@stud.u-szeged.hu>,
Gyongyi Mikola <mgyongyi@hung.u-szeged.hu>,
Hofer Adam nincs cime <Btamas@btk.jpte.hu>,
Ivan Zsuzsanna <manka1@freemail.hu>,
Jelasity Mark <JELASITY@INF.U-SZEGED.HU>,
Kaldy Zsuzsa <kaldy@ruccs.rutgers.edu>,
Kontra Miklos <kontra@sol.cc.u-szeged.hu>,
Kotel Emoke <kotele@sol.cc.u-szeged.hu>,
Kovacs Agnes <agikov@hotmail.com>,
Kovacs Gyula <KOGYU@PHYS.SZOTE.U-SZEGED.HU>,
Kri Szabolcs <SZKERI@PHYS.SZOTE.U-SZEGED.HU>,
Mekis Peter <mekis@sol.cc.u-szeged.hu>,
Olah Gyongyver <gagarin@mksh.u-szeged.hu>,
Pleh Csaba <pleh@edpsy.u-szeged.hu>,
Pocs Kata Rita <pkatas@mars.arts.u-szeged.hu>,
Racsmany Mihaly melyik a jo cim? <racsmany@izabell.elte.hu>,
Sandor Klara <sandor@jgytf.u-szeged.hu>,
Stacho Laszlo Pal <h633874@stud.u-szeged.hu>,
Szilagyi Sandor <szilagyi@lett.ubbcluj.ro>,
Thuma Orsolya <thumao@mail.alba.hu>, Toth Eva <tothe@edpsy.u-szeged.hu>,
Zemplen Gabor <zemplen@hps.elte.hu>
Subject: Re: Cangelosi/Harnad Symbols
On Thu, 18 May 2000, Lorincz, Andras wrote:
> sh> For nontrivial features, requiring nontrivial learning (e.g.,
> sh> chicken-sexing), active trial-and-error toil (reinforcement learning)
> sh> and feedback are needed, and not mere passive exposure and
> sh> association.
> sh>
> sh> By way of an intuitive example, if a school-teacher had 30 students in
> sh> every class, and taught 7 different classes a day (and never for longer
> sh> than a few months), and every child every day wore his name written on
> sh> his forehead -- and every day the teacher had to interact with each
> sh> child several times by name -- what do you think would be the "hit
> sh> rate" of that interaction if, on the last day of class, all the
> sh> students came in without their forehead-labels?
>
>la> Hit rate will be bad, no doubt, for an average person. For some,
>la> however, it will be zero. I know people who can learn six pages
>la> of the phone book with ten errors in one night. So this is not
>la> the issue, the problem is somewhere else. The hit rate is not
>la> important if we are after general concepts, or you should argue
>la> why the hit rate is more than being merely a quantitative issue.
Read Luria's "The Mind of a Mnemonist" and Borges's "Funes the
Memorious" in order to see the down-side of rote memorization of
everything. In a word, it is a trade-off with abstraction and
generalization.
Luria, A. R. (1968) The mind of a mnemonist; a little book about a
vast memory [by] A. R. Luria. Translated from the Russian by Lynn
Solotaroff. With a foreword by Jerome S. Bruner. New York, Basic
Books [Re-issued: Harvard University Press]
http://qsilver.queensu.ca/~phil158a/memory/luria.htm
http://psyche.cs.monash.edu.au/v2/psyche-2-10-cytowic.html
http://www.lgu.ac.uk/psychology/elander/SHERASH.HTM
http://grandin.com/inc/visual.thinking.html
cogsci.soton.ac.uk/~harnad/Hypermail/Foundations.Cognition/index.html
cogsci.ecs.soton.ac.uk/~harnad/Hypermail/Thinking.Psychologically/index.html
cogsci.soton.ac.uk/~harnad/Hypermail/Thinking.Psychologically96/0041.html
http://www.bridgewater.edu/~atrupe/GEC101/Funes.html
Borges, J. L. (1969) Funes the memorious. In Yates, D. & Irby,
J. E. (eds.) Labyrinths. New York: New Directions.
http://slate.msn.com/Concept/96-07-15/Concept.asp
cogsci.soton.ac.uk/~harnad/Hypermail/Foundations.Cognition/author.html
cogsci.ecs.soton.ac.uk/~harnad/Hypermail/Thinking.Psychologically/date.html
> sh> But I don't think mere passive exposure and association is enough to
> sh> ground symbols either.
>
>la> How about the first two words learned, say 'dad' or 'mam'?
>la> BTW, my son is different. In his case the first word was 'auto'
>la> and he meant it.
Dad, mum and auto are trivial categories. Try chicken-sexing, or
naming all the people you've seen...
Stevan
This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:07 GMT