turner@webb.psych.ufl.edu (Carl Turner) (11/30/90)
In article <GREENBA.90Nov29151247@gambia.crd.ge.com> greenba@gambia.crd.ge.com (ben a green) writes: >What more is needed for an artifical person is motivation, or needs, and >an adaptive mechanism that raises the probability of the behavior that >satisfies them. This is "reinforcement". >You also need a mechanism for acquiring new needs that are based on the >old ones. So a robot that starts out inherently valuing only survival >can acquire a taste for electrical outlets ... . You've hit on the basic idea of the "Hopkins Beast" of the 1950's. This was a robot that was invented at Johns Hopkins that sort of looked like an upside-down garbage can. It would patrol the halls of whatever building it was in, and when it ran low on energy would plug itself into a wall socket and recharges its batteries. It was a nice demonstration of the idea that you get complex behavior from a simple system (the Beast) interacting with a complex environment (the halls and offices of Hopkins). Of course, it didn't "acquire" a taste for outlets but was preprogrammed to find them, else it would not have "survived" for long. >Warning! This line of thought may lead you into the dreaded world of >Behaviorism of the kind explained by B. F. Skinner in "About Behaviorism." >Ben A. Green, Jr. >greenba@crd.ge.com "Reinforcing" a machine for a behavior is the easiest thing in the world, all you have to do is increment a counter, or reset a parameter.... Behaviorist principles of behavior and reinforcement seem an appropriate domain for modelling and simulation, but computational modelling does not seem to have caught on there. Perhaps someone could explain why. Carl Turner turner@webb.psych.ufl.edu
brianc@labmed.ucsf.edu (Brian Colfer) (12/08/90)
In article <25667@uflorida.cis.ufl.EDU> turner@webb.psych.ufl.edu (Carl Turner) writes: >In article greenba@gambia.crd.ge.com (ben a green) writes: > >>What more is needed for an artifical person is motivation, or needs, and >>an adaptive mechanism that raises the probability of the behavior that >>satisfies them. This is "reinforcement". One of the challenges of AI is to create both the phylogenetics and the ontogenetics of the artificial organism (AO). I think that part of this may be found in programming the control principles (reinformcement, extinction, deprivation and punishment) into subsystems of the AO. This could be done probably with a hierarchical neural network system (NN) with groups of nodes emphasizing particular stimulus sensitivity and behaviors. [...] >You've hit on the basic idea of the "Hopkins Beast" of the 1950's. [...] > >It was a nice demonstration of the idea that you get complex >behavior from a simple system (the Beast) interacting with a complex >environment (the halls and offices of Hopkins). > >Of course, it didn't "acquire" a taste for outlets but was preprogrammed >to find them, else it would not have "survived" for long. The implication here is to scale up the program to a more complex system. Hmmm an interesting idea. >"Reinforcing" a machine for a behavior is the easiest thing in the world, >all you have to do is increment a counter, or reset a parameter.... I think that a systematic development of Skinner's ideas concerning verbal behavior will result in an effective general model for semantic processing. >Behaviorist principles of behavior and reinforcement seem an appropriate >domain for modelling and simulation, but computational modelling does not >seem to have caught on there. Perhaps someone could explain why. Besides the ideologic hangups people have about Behaviorism.. while the principles are simple the interaction results in an incredibly complex network of behavior patterns. It is similar to designing an ecosystem. I think it is just a huge job with lots of tedious work that will pile up to make an interesting whole. >Carl Turner >turner@webb.psych.ufl.edu -- Brian Colfer | UC San Francisco |------------------------| | Dept. of Lab. Medicine | System Administrator, | brianc@labmed.ucsf.edu | S.F. CA, 94143-0134 USA | Programer/Analyst | BRIANC@UCSFCCA.BITNET | PH. (415) 476-2325 |------------------------|
valis@athena.mit.edu (John O'Neil) (12/09/90)
In article <16562@cgl.ucsf.EDU> brianc@labmed.ucsf.edu (Brian Colfer) writes: >I think that a systematic development of Skinner's ideas concerning >verbal behavior will result in an effective general model for semantic >processing. Please then explain the verbal behavior contained in "War and Peace" by Tolstoy. Make reference, if you wish, to Napoleon as the stimulus for a book written several decades later. John O'Neil Organlegger "From head to toe, you know where to go." Spleens a specialty.
greenba@gambia.crd.ge.com (ben a green) (12/10/90)
In article <VALIS.90Dec8170548@e40-008-8.mit.edu> valis@athena.mit.edu (John O'Neil) writes: In article <16562@cgl.ucsf.EDU> brianc@labmed.ucsf.edu (Brian Colfer) writes: >I think that a systematic development of Skinner's ideas concerning >verbal behavior will result in an effective general model for semantic >processing. Please then explain the verbal behavior contained in "War and Peace" by Tolstoy. Make reference, if you wish, to Napoleon as the stimulus for a book written several decades later. Can't do it, of course, but that's not the zinger you may think it is. After all, physics is a mature science. Yet with it, we cannot predict next week's weather with any accuracy. Even if psychology were as advanced as physics, we would need a complete record of Tolstoy's life and the details of his environment. Let's wrangle over more reasonable problems, especially problems for which there are competing solutions. In other words, does anybody have anything better than Skinner's ideas -- anything with comparable scope? -- Ben A. Green, Jr. greenba@crd.ge.com Speaking only for myself, of course.
valis@athena.mit.edu (John O'Neil) (12/11/90)
In article <GREENBA.90Dec10081926@gambia.crd.ge.com> greenba@gambia.crd.ge.com (ben a green) writes: In article <VALIS.90Dec8170548@e40-008-8.mit.edu> valis@athena.mit.edu (John O'Neil) writes: In article <16562@cgl.ucsf.EDU> brianc@labmed.ucsf.edu (Brian Colfer) writes: >I think that a systematic development of Skinner's ideas concerning >verbal behavior will result in an effective general model for semantic >processing. Please then explain the verbal behavior contained in "War and Peace" by Tolstoy. Make reference, if you wish, to Napoleon as the stimulus for a book written several decades later. Can't do it, of course, but that's not the zinger you may think it is. I didn't think it was a "zinger"; I was just trying to make a point in an (admittedly marginally) humorous way. After all, physics is a mature science. Yet with it, we cannot predict next week's weather with any accuracy. Even if psychology were as advanced as physics, we would need a complete record of Tolstoy's life and the details of his environment. Even if physics were as advanced as physics, we couldn't predict next week's weather with any accuracy. With *all the details* of Tolstoy's environment, you seem to believe in principle that a deterministic behaviorism could predict the verbal behaviorism evinced in _War and Peace_. This seems to me akin to showing which butterfly in China caused Hurricane Bruno -- not just difficult, but even in principle quite impossible. That is a basic problem with "verbal behavior": when one allows the stimulus to be an event 50 years in the past, the notion of "stimulus", which is at the heart of behaviorism, becomes completely muddled. Let's wrangle over more reasonable problems, especially problems for which there are competing solutions. In other words, does anybody have anything better than Skinner's ideas -- anything with comparable scope? The question is whether the scope of behaviorism includes these phenomena at all, or whether it is necessary to posit more complex internal structures to the mind than behaviorism allows. John O'Neil Organlegger "From head to toe, you know where to go." Spleens a specialty.
greenba@gambia.crd.ge.com (ben a green) (12/12/90)
In article <VALIS.90Dec10173017@e40-008-8.mit.edu> valis@athena.mit.edu (John O'Neil) argues:
Even if physics were as advanced as physics, we couldn't predict next
week's weather with any accuracy. With *all the details* of Tolstoy's
environment, you seem to believe in principle that a deterministic
behaviorism could predict the verbal behaviorism evinced in _War and
Peace_. This seems to me akin to showing which butterfly in China
caused Hurricane Bruno -- not just difficult, but even in principle
quite impossible.
The butterfly example is good; I won't quibble over the difference between
difficult and impossible in that context.
That is a basic problem with "verbal behavior": when one allows the
stimulus to be an event 50 years in the past, the notion of
"stimulus", which is at the heart of behaviorism, becomes completely
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
muddled.
Many think of behaviorism as a matter of stimulus and response, and it
may have been for some in the early days. But it's not for the radical
behaviorists of Skinner's school. For them, most "responses" of any
interest have no immediately associated stimulus. They have
conditional probabilities of emission. Events in our lives change us
by affecting these probabilities ever after. Some day neurologists will
discover the process.
I could attempt a precis of modern behaviorism, but it wouldn't be as
good as Skinner's book, _About Behaviorism_, already available in
paperback in your neighborhood. See especially the early chapter
listing myths about behaviorism. I'd enjoy further discussion with
anyone who takes the trouble to avoid these elementary errors.
--
Ben A. Green, Jr.
greenba@crd.ge.com
Speaking only for myself, of course.G.Joly@cs.ucl.ac.uk (Gordon Joly) (12/12/90)
In article <GREENBA.90Dec10081926@gambia.crd.ge.com> from greenba@gambia.crd.ge.com (ben a green) >Can't do it, of course, but that's not the zinger you may think it >is. After all, physics is a mature science. Yet with it, we cannot >predict next week's weather with any accuracy. Even if psychology were >as advanced as physics, we would need a complete record of Tolstoy's >life and the details of his environment. > >Let's wrangle over more reasonable problems, especially problems for >which there are competing solutions. In other words, does anybody have >anything better than Skinner's ideas -- anything with comparable >scope? > >-- >Ben A. Green, Jr. >greenba@crd.ge.com Sure; as a colleague proposed is a recent seminar :- "Academic Psychology is Folk Psychology". This means ("what it means" Ed ;-) that psychologists, being human, design their experiments about humans (and sociologists about society) with the handicap of being "biased" at birth. A good example of this effect is of the recent "scientific find" that high IQ correlates to myopia. "It's official" - Ed. Gordon Joly +44 71 387 7050 ext 3716 InterNet: G.Joly@cs.ucl.ac.uk UUCP: ...!{uunet,ukc}!ucl-cs!G.Joly Computer Science, University College London, Gower Street, LONDON WC1E 6BT
valis@athena.mit.edu (John O'Neil) (12/15/90)
In article <GREENBA.90Dec11111333@gambia.crd.ge.com> greenba@gambia.crd.ge.com (ben a green) writes:
I could attempt a precis of modern behaviorism, but it wouldn't be as
good as Skinner's book, _About Behaviorism_, already available in
paperback in your neighborhood. See especially the early chapter
listing myths about behaviorism. I'd enjoy further discussion with
anyone who takes the trouble to avoid these elementary errors.
I agree: one should read the literature in order to avoid elementary
errors.
For an excellent criticism of the behavioristic view of language and
semantics, see Chomsky's review of B. F. Skinner's ``Verbal Behavior''
in Language 35: 26-58. Although not as easily available, it is a
useful antidote to Skinner's work.
John O'Neil
Organlegger
"From head to toe, you know where to go."
Spleens a specialty.greenba@gambia.crd.ge.com (ben a green) (12/18/90)
In article <VALIS.90Dec14201209@e40-008-6.mit.edu> valis@athena.mit.edu (John O'Neil) writes:
For an excellent criticism of the behavioristic view of language and
semantics, see Chomsky's review of B. F. Skinner's ``Verbal Behavior''
in Language 35: 26-58. Although not as easily available, it is a
useful antidote to Skinner's work.
Chomsky's review was highly influential among those who were unfamiliar
with Skinner's work, and had the effect of making them think they could
dismiss it.
However, the review missed the point of the book. Chomsky's advice to
study "what language is" before "what language is for," doesn't work
well when you consider language as a tool. The essence of
understanding a tool is to understand first what it is for. Skinner
explains what language is for in many of its aspects.
Chomsky's formal grammars are useful for computers, but for human
languages, his approach has been discarded in favor of "pragmatics,"
a school which analyzes language explicitly in terms of its uses.
--
Ben A. Green, Jr.
greenba@crd.ge.com
Speaking only for myself, of course.valis@athena.mit.edu (John O'Neil) (12/18/90)
In article <GREENBA.90Dec17130607@gambia.crd.ge.com> greenba@gambia.crd.ge.com (ben a green) writes: In article <VALIS.90Dec14201209@e40-008-6.mit.edu> valis@athena.mit.edu (John O'Neil) writes: For an excellent criticism of the behavioristic view of language and semantics, see Chomsky's review of B. F. Skinner's ``Verbal Behavior'' in Language 35: 26-58. Although not as easily available, it is a useful antidote to Skinner's work. Chomsky's review was highly influential among those who were unfamiliar with Skinner's work, and had the effect of making them think they could dismiss it. As unlikely as it might seem to you, there are people who have read Skinner -- even understood him -- and do not agree with him. However, the review missed the point of the book. Chomsky's advice to study "what language is" before "what language is for," doesn't work well when you consider language as a tool. The essence of understanding a tool is to understand first what it is for. Skinner explains what language is for in many of its aspects. This is like carping that physicists study the properties of electrons outside of their use in computers and power systems. If Bohr, Einstein, and Planck _inter alia_ had not thought about ``what electrons are'', we would have a much poorer conception of ``what electrons are for''. Chomsky's formal grammars are useful for computers, but for human languages, his approach has been discarded in favor of "pragmatics," a school which analyzes language explicitly in terms of its uses. This will come as quite a surprise to the folks on sci.lang, and as for me -- you can only imagine my shock to discover that mathematical linguistics is as dead as the pyramids. Thanks, Ben, for clueing me in on this; I guess I'll have to change fields as soon as possible. Seriously, though, you might familiarize yourself with a field before pontificating about it. John O'Neil Organlegger "From head to toe, you know where to go." Spleens a specialty.
mmt@client2.DRETOR.UUCP (Martin Taylor) (12/19/90)
John O'Neil writes: In article <16562@cgl.ucsf.EDU> brianc@labmed.ucsf.edu (Brian Colfer) writes: - ->I think that a systematic development of Skinner's ideas concerning ->verbal behavior will result in an effective general model for semantic ->processing. - -Please then explain the verbal behavior contained in "War and Peace" -by Tolstoy. Make reference, if you wish, to Napoleon as the stimulus -for a book written several decades later. - To understand the sense of this exchange, let us consider a possible parallel exchange: Person A " I think a systematic development of Gell-Mann's ideas concerning the interactions among quarks and photons and gluons will result in an effective general model for physical and chemical structures." Person B " Please explain then the colour patterns in the tail of a peacock" Not that I necessarily accept Skinner's ideas, but the manipulation of simple concepts related in simple ways can lead to some wildly unexpected results. How about 1 and 0, "and" "or" and "not". -- Martin Taylor (mmt@ben.dciem.dnd.ca ...!uunet!dciem!mmt) (416) 635-2048 There is no legal canon prohibiting the application of common sense (Judge James Fontana, July 1990, on staying the prosecution of a case)
G.Joly@cs.ucl.ac.uk (Gordon Joly) (12/29/90)
A.N.Other said
* This is like carping that physicists study the properties of electrons
* outside of their use in computers and power systems. If Bohr,
* Einstein, and Planck _inter alia_ had not thought about ``what
* electrons are'', we would have a much poorer conception of ``what
* electrons are for''.
*
* Chomsky's formal grammars are useful for computers, but for human
* languages, his approach has been discarded in favor of "pragmatics,"
* a school which analyzes language explicitly in terms of its uses.
*
* This will come as quite a surprise to the folks on sci.lang, and as
* for me -- you can only imagine my shock to discover that mathematical
* linguistics is as dead as the pyramids. Thanks, Ben, for clueing me
* in on this; I guess I'll have to change fields as soon as possible.
*
* Seriously, though, you might familiarize yourself with a field before
* pontificating about it.
It was Milikan who first measured the charge on the electron, just
before "Bohr, Einstein, and Planck _inter alia_" and he falsified his
records. A close examination of one hundred odd experiments in his
notebooks showed that he only reported about half. If he had reported
the full results his thesis would have been "the electron, as a single
discrete unit, does not exist".
A very pragamatic scientist, Milikan.
Gordon Joly +44 71 387 7050 ext 3716 Internet: G.Joly@cs.ucl.ac.uk
UUCP: ...!{uunet,ukc}!ucl-cs!G.Joly Computer Science, University
College London, Gower Street, LONDON WC1E 6BTjmc@Gang-of-Four.stanford.edu (John McCarthy) (12/30/90)
Curiously enough, Millikan came up with approximately the correct charge on the electron in the oil drop experiment that Gordon Joly claims he faked. Picking over old notebooks is an uncertain methodology, because not all the considerations for what measurements to report are likely to have been written down. By the 1930s, the Millikan oil drop experiment was doable in undergraduate laboratories. Now those are the notebooks to look at!
selmer@hpcuhc.cup.hp.com (Steve Elmer) (01/11/91)
Whatever happened to the "comp.ia.philosophy" part of this notes string?
greenba@gambia.crd.ge.com (ben a green) (01/16/91)
In article <113250001@hpcuhc.cup.hp.com> selmer@hpcuhc.cup.hp.com (Steve Elmer) writes:
Whatever happened to the "comp.ia.philosophy" part of this notes string?
Steve, my email to you and to Carl Turner bounces.
I would like to post a specification for a robot whose program
imitates that of humans in a way that reflects what we understand
today about learning and maintenance of behavior. It would result in a
robot that could be described not only as intelligent but also
intuitive and subject to certain human frailties, such as getting
hooked on horse-race betting.
My problem with doing this is lack of time and some fear of the explanation
task, considering that the line in psychology that I would follow is rather
out of fashion among readers of comp.ai.
Reinforcement is a central idea in this approach. Neural nets are not
so central but offer a practical way around certain computational
hurdles, such as rapid updating of a conditional probability table.
Cellular automata are not part of the scheme of things at all. I think
of cellular automata as things analogous to cells -- maybe even nerve
cells -- that operate at an even lower level that that I am interested
in. They are surely interesting in their own right. (One of my
cellular automaton pastimes is written up in Algo/rithms, Winter,
1990.)
Thanks for the show of interest. Maybe that will help me get up the
energy and courage to post my design.
--
Ben A. Green, Jr.
greenba@crd.ge.com
Speaking only for myself, of course.G.Joly@cs.ucl.ac.uk (Gordon Joly) (01/25/91)
In article <JMC.90Dec29161242@Gang-of-Four.stanford.edu> John McCarthy said > Curiously enough, Millikan came up with approximately the correct > charge on the electron in the oil drop experiment that Gordon Joly > claims he faked. Picking over old notebooks is an uncertain methodology, > because not all the considerations for what measurements to > report are likely to have been written down. By the 1930s, > the Millikan oil drop experiment was doable in undergraduate > laboratories. Now those are the notebooks to look at! Yup; indeed. But suppose Millikan was a team and suppose the names of the team members were Fleischmann and Pons and suppose they had come up with Cold Fusion.... would you buy a used Cadillac from them? Or trust their theory? Gordon Joly +44 71 387 7050 ext 3716 Internet: G.Joly@cs.ucl.ac.uk UUCP: ...!{uunet,ukc}!ucl-cs!G.Joly Computer Science, University College London, Gower Street, LONDON WC1E 6BT Email: Les jeux sans frontiers du monde