[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Subscribe]
DM: [Robot-for-President] Machine Learning Taxonomy (fwd)From: Franklin Wayne Poley Date: Wed, 22 Mar 2000 23:11:42 -0800 (PST) I am trying to get more psychologists involved in machine learning but this may be helpful to our discussion on datamine-l. FWP. ---------- Forwarded message ---------- Date: Wed, 22 Mar 2000 18:05:19 -0800 (PST) From: Franklin Wayne Poley <fwpoley@vcn.bc.ca> Reply-To: Robot-for-President@onelist.com To: robot-for-president@onelist.com Cc: asap@spssi.org Subject: [Robot-for-President] Machine Learning Taxonomy From: Franklin Wayne Poley <fwpoley@vcn.bc.ca> Can we come up with a taxonomy which would incorporate enough CATEGORIES of learning such that all of the end results of learning in humans would be accounted for? That has to do with the "what" of human learning. The "how" is another matter and differs greatly in mice, men and machines even if the end result is the same. Maze learning is a good example, easily explained in these terms. Next, can we verbalize a theoretical "how", an abstract notion of how one might master a maze, extract the useful information from text, etc? If Skinner's dictum is correct ("If it can be verbalized it can be programmed"), next a program can be written for each category of learning. Is there a machinery which can then execute each of these categories of learning? It seems there is. That being so, we have the beginnings of a "general learning program" for robots which can extend their learning in co-operative relationships with humans as the Waseda University web site says. FWP. > ---------- Forwarded message ---------- > Date: Thu, 2 Mar 2000 12:42:20 -0800 (PST) > From: Franklin Wayne Poley <culturex@vcn.bc.ca> > Reply-To: FUTURE-CITIES@onelist.com > To: spsp@scp.rochester.edu, spssi@spssi.org, ahoffice@aol.com, > enquiry@bps.org.uk, public.affairs@apa.org, apa@psych.org > Cc: future-cities@onelist.com, humanoids@usc.edu, sschaal@usc.edu, > knoll@ti.uni-bielefeld.de > Subject: [FUTURE-CITIES] Humanoids 2000 Conference > > From: Franklin Wayne Poley <culturex@vcn.bc.ca> > > Dear Psychology Colleague: > I am writing this letter to bring to your attention the Humanoids 2000 Conference, http://humanoids.usc.edu . This is the world's first major international conference on humanoids and I believe it is a wonderful opportunity for psychologists in a number of specialties as well as general practitioners. The Humanoids 2000 web page invites multi-disciplinary participation. > Perhaps Category 2 for "Sessions" would be most suited for psychology participation: "Interaction With Humans And The Environment". > I think we have the beginning of a complete artificial psychological system. This is far more than traditional AI. How complete is this artificial personality? What would psychologists at the Society for Personality and Social Psychology say? How applicable are standard tests and profiles with respect to comparing humans and humanoids? I do know that psychologists who specialize in psychometric assessments of mentally and physically handicapped patients could contribute much to Humanoids 2000. I worked for almost a decade in an institution for mentally and physically handicapped people but that was a long time ago and psychometrics was not my specialty so I am not the best practitioner to address this topic now. However, I would be very eager to read a report from an up to date practitioner who does this work on a daily basis. > The wording of the Waseda University web site, http://www.humanoid.rise.waseda.ac.jp tells us how much the humanoid is a psychological system. Their goal is "to develop an anthropomorphic robot named 'humanoid' which will comprise sensing, recognition, expression and motion sub-systems to enable robots and humans to build common mental and physical spaces, co-operatively." Clinical and counselling psychologists typically have two sets of reports. One is technical and can be understood by their colleagues. The other set is no less meaningful but is presented in the vernacular for family, community etc. I believe the latter is needed to advance the building of those common mental and physical spaces as humanoids take their place in human society. I also think the mechanical engineers, electrical engineers and computer scientists who create the humanoids will be pleasantly surprised by how much psychologists can contribute to their field. This is, after all a "creative" unde! ! ! ! rtaking > as much as an engineering feat following draftsmans' plans. The next creative step for a humanoid R&D project could be foreseen by a psychologist as well as an engineer. > Psychologists who work in commerce, from personnel to marketing psychology will have much to say about humanoids. Putting a humanoid into the mass market capable of occupying a number of work roles is something no-one has done before! The objective of the Shadow Robot Project, http://www.shadow.org.uk/philo/manifesto.stm is "...to build a genuinely useful general-purpose robot, at a price which people, rather than just institutions, can afford." Perhaps it will take R&D funding comparable to the International Space Station to get to that stage but society at large may decide it is worth the cost. That cost could be $100 b. over 10 years. Those providing the funding are seldom engineers or computer scientists and they are represented by very shrewd professionals called politicians and administrators. The psychologist can serve as liason between these two factions in society, the faction supplying the grants and those using them to develop humanoids. Statements to the publi! ! ! ! c must > sum up as "This is the state of the art, now; here are some expected steps in the near future; and here is where we expect to be after spending x dollars and y years." > I think the Association for Humanistic Psychology and psychologists involved in ethics, self-actualization and transpersonal psychology will be concerned about the extent to which a humanoid constitutes a living system. Ian Pearson from British Telecom writes in the latest issue of The Futurist that by 2020 we can expect "electronic life-forms (to be) given basic rights". Certainly we cannot dismiss humanoid rights as a joke when we have a complex, intelligent, human-like creature with much autonomy, capable of learning ad self-improvement. > Next we come to the 'holy grail' of robotics, the "general learning program". Pearson writes that by 2011, "computers (will) surpass human learning and logic abilities." That seems surprising but it looks to me as if we can find computers or robots now which are capable of every category of learning found in humans. I say category, not process. How humans and humanoids learn may differ greatly. Consider this list of learning categories: (1) maze learning; (2) hand-over-hand learning as is used to train small children and some robots in factories; (3) modelling/imitation/mimicry as described on the Kawato Humanoid web site; (4) one-trial learning which is how we could describe the acquisition of knowledge by direct downloading; (5) learning from text (refer to some recent research by Seeung and Lee); (6) trail-and-error learning (see Christiansen et al); (7) object recognition (Holland's classifiers; face, fingerprint and iris recognition programs); (8) voice recognition l! ! ! ! earning > (standard office dictation programs); (9) mapping (mapping from an artificial vision system and range finders); (10) reinforcement learning (robot responses to reward and punishment); (11) self-improvement learning (including evolutionary robotics, generating new software). What have I left out? Given that humanoids now manifest all categories of learning found in humans could they not be assembled into that "general learning program"? After it is clearly written in layman's language, psychology language and computer language, anyone is free to suggest ways in which it can be improved upon, large or small. Of course psychologists who specialize in conditioning, learning and educational psychology will be able to advance this technology a great deal. The dictum of the late B.F. Skinner is as relevant now as it was a half century ago: "If it can be verbalized, it can be programmed." > If Pearson is correct and we are only a decade away from humanoids/robots which can learn better than humans then we need to hear from social philosophers and political psychologists as to what they think the social-political-economic implications will be. How much incentive will humans have to learn when humanoids can learn better and tell us what we need to know? Will we then have problems maintaining control over the humanoids? How fast might humanoids learn? Is there any way humans can predict humanoid growth in learning with reasonable accuracy? Will it be fast enough to keep ahead of human populations increasing with geometric growth and expanding across the near galaxy? > A starter project which could be initiated some time in this century might be to send humanoids out to the near-by asteroid belt to learn how to mine, mill and ship the billions x billions of dollars in mineral wealth. > I am hoping this letter will encourage psychologists of many backgrounds to get involved in Humanoids 2000. As valuable as their contributions have been it seems to me that AI psychologists have been co-opted to the way of thinking of computing science experts. In my opinion, society would be better served if psychologists would approach humanoids as they would "very unusual humans". They should maintain their techniques, concepts, jargon, analyses etc. as much as possible. Humanoids 2000 is very exciting. It opens up opportunities for psychologists to contribute to the development of this new technology in the service of humankind. I hope many will avail themselves of the opportunities. > Please circulate this letter/information freely. > Sincerely-Franklin Wayne Poley, Ph.D. > > http://users.uniserve.com/~culturex/Machine-Psychology.htm ------------------------------------------------------------------------ GET A NEXTCARD VISA, in 30 seconds! Get rates as low as 2.9% Intro or 9.9% Fixed APR and no hidden fees. Apply NOW! http://click.egroups.com/1/936/5/_/433155/_/953777121/ ------------------------------------------------------------------------ *** The Era of Total Automation is Now ***
|
MHonArc
2.2.0