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

RE: Learning Paradox



So I don't know which direction to go with this - so I'm just going to forge ahead with something interesting I read related to this.  And sticking to this whole Pragmatism trip I'm on I thought it might be interesting to pose it as a thought experiment (not by me, but by Daniel Dennett -did I get the name right).  It seems this is a big argument among the cognitive scientists themselves.  With the Pragmatic AI cognitive scientists (Dennett lists them all but I can't remember - but he lists Rorty who's not AI or a cognitive scientist, but always a good ally in a pinch I suppose) against the more nativist cognitive scientists such as Fodor and Searle.
 
Before I copy the thought experiment, and it is a little long, and can hurt the head under some circumstances, I do want to say that Dennett makes the distinction between simple physiological systems such as plants and humans (sort of following on what Geoff says here - which sort of made me think of this whole thing).
 
Here is the thought experiment with apologies to Professor Dennett (I hope this is legal).  I will offer a couple of lines on the end concerning my own thinking.  If there are no responses I will assume everybody's plate is too full, or everybody went to the beach (can't do that in Ohio).  Oh, one more thing, which I think is really interesting thinking about this,
 
Von Glaserfield (again, possible apologies about the name, but it's late and I don't want to look it up) suggests that the Learning Paradox focuses on the benefits of inductive logic, while Joe Glicks accomodatioin, assimilation, and adaptation focuses more on abductive logic (I got a C- in logic in college so I'm not going any farther with that).
 
Here goes,
 

Suppose you decided, for whatever reasons, that you wanted to experience life in the 25th century, and suppose that the only known way of keeping your body alive that long required it to be placed in a hibernation device of sorts, where it would rest, slowed down and comatose, for as long as you liked. You could arrange to climb into the support capsule, be put to sleep, and then automatically awakened and released in 2401. This is a time- honored science fiction theme, of course. 

Designing the capsule itself is not your only engineering problem, for the capsule must be protected and supplied with the requisite energy (for refrigeration or whatever) for over 400 years. You will not be able to count on your children and grandchildren for this stewardship, of course, for they will be long dead before the year 2401, and you cannot presume that your more distant descendants, if any, will take a lively interest in your well-being. So you must design a supersystem to protect your capsule, and to provide the energy it needs for four hundred years. 

Here there are two basic strategies you might follow. On one, you should find the ideal location, as best you can foresee, for a fixed installation that will be well supplied with water, sunlight, and whatever else your capsule (and the supersystem itself) will need for the duration. The main drawback to such an installation or "plant" is that it cannot be moved if harm comes its way--if, say, someone decides to build a freeway right where it is located. The second alternative is much more sophisticated, but avoids this drawback: design a mobile facility to house your capsule, and the requisite early-warning devices so that it can move out of harm's way, and seek out new energy sources as it needs them. In short, build a giant robot and install the capsule (with you inside) in it. 

These two basic strategies are obviously copied from nature: they correspond roughly to the division between plants and animals. Since the latter, more sophisticated strategy better fits my purposes, we shall suppose that you decide to build a robot to house your capsule. You should try to design it so that above all else it "chooses" actions designed to further your best interests, of course. "Bad" moves and "wrong" turns are those that will tend to incapacitate it for the role of protecting you until 2401--which is its sole raison d'tre. This is clearly a profoundly difficult engineering problem, calling for the highest level of expertise in designing a "vision" system to guide its locomotion, and other "sensory" and locomotory systems. And since you will be comatose throughout and thus cannot stay awake to guide and plan its strategies, you will have to design it to generate its own plans in response to changing circumstances. It must "know" how to "seek out" and "recognize" and then exploit energy sources, how to move to safer territory, how to "anticipate" and then avoid dangers. With so much to be done, and done fast, you had best rely whenever you can on economies: give your robot no more discriminatory prowess than it will probably need in order to distinguish what needs distinguishing in its world. 

Your task will be made much more difficult by the fact that you cannot count on your robot being the only such robot around with such a mission. If your whim catches on, your robot may find itself competing with others (and with your human descendents) for limited supplies of energy, fresh water, lubricants, and the like. It would no doubt be wise to design it with enough sophistication in its control system to permit it to calculate the benefits and risks of cooperating with other robots, or of forming alliances for mutual benefit. (Any such calculation must be a "quick and dirty" approximation, arbitrarily truncated. See Dennett forthcoming.) 

The result of this design project would be a robot capable of exhibiting self-control, since you must cede fine-grained real-time control to your artifact once you put yourself to sleep). Endnote 2 As such it will be capable of deriving its own subsidiary goals from its assessment of its current state and the import of that state for its ultimate goal (which is to preserve you). These secondary goals may take it far afield on century- long projects, some of which may be ill-advised, in spite of your best efforts. Your robot may embark on actions antithetical to your purposes, even suicidal, having been convinced by another robot, perhaps, to subordinate its own life mission to some other. 

But still, according to Fodor et al., this robot would have no original intentionality at all, but only the intentionality it derives from its artifactual role as your protector. Its simulacrum of mental states would be just that-- not real deciding and seeing and wondering and planning, but only as if deciding and seeing and wondering and planning. 

 

All right, now away from Dennett's brilliance to my more mundane questions.  Is it possible that the robot cannot, will not have any original intentionality beyond what we have created for it, and if so is it possible for the robot, and us as the creators to survive?  Isn't extinction inevitable if we follow the whole idea of the learning paradox.  By the way, according the Dennett, Fodor seems to hate evolutionary theory (or is at the very lead annoyed by it).

Hey, did anybody see Obama tonight?  He rocked!  And a great Pragmatist!

Michael


________________________________

From: Geoff Hayward [mailto:geoff.hayward@edstud.ox.ac.uk]
Sent: Tue 7/27/2004 6:03 PM
To: xmca@weber.ucsd.edu
Subject: RE: Learning Paradox



Physiological metaphors one and all and a physiological system can but
react according to the set parameters (thank you Claude) and that is the
learning paradox. But if you move beyond the physiological individual
you find some bootstrapping devices, albeit limited by our collective
intelligence - which begs another question ....... an additional unit of
analysis - the activity system. But how does this arise and how well
does Activity Theory deal with issues of identity ... grey moments in an
English summer.

Geoff

Dr Geoff Hayward
Associate Director SKOPE
OUDES
15 Norham Gardens
Oxford
OX2 6PY UK

Phone: +44 (0)1865 274007
Fax: + 44 (0)1865 274027
e-mail: geoff.hayward@edstud.ox.ac.uk


-----Original Message-----
From: Glick, Joseph [mailto:JGlick@gc.cuny.edu]
Sent: 27 July 2004 18:47
To: 'xmca@weber.ucsd.edu'
Subject: RE: Learning Paradox

Assimilation, accommodation, adaptation, organization anyone?






<<winmail.dat>>