[Avery Andrews 930105.10.00]
(Gary Cziko 930104.1730)
So I can only guess that today must be the latest sunset down under.
Not sure - we just got rid of all of our newspapers for recycling, & my 3D
visualization is to crummy to figure it out easily. Plus, of course,
the infinite confusing effect of the dateline ...
Not sure - we just got rid of all of our papers for recycling, & my 3d
visualization is to crummy to figure it out easily. Plus, of course,
the infinite confusing effect of the dateline ...
Switching topics, another problem with Gary's demo is that it doesn't
defuse the `standard brushoff' of PCT -- `we've known about this
feedback stuff for more than thirty years, and that it doesn't really
do very much for you'. I don't believe that any collection of `mere' demos
of the phenomenon of control can overcome this objection, & I'm not too
optimistic about the prospects for heaps of quantitative data either,
no matter how precise. People will just say `that's fine as far as it
goes, but it's just one step up from psychophysics, and doesn't address
the hard problems that people are concerned with now.' And I don't think
it's a good idea to dismiss these people as writeoffs, since it would
really much better if more of them were at least minimally sympathetic to PCT,
if not outright converts.
And I'm not too keen on `going to the public' a la Freud - after all,
there is a growing body of opinion that the public got gulled that
time, as well as by the astrologers, channellers, crystal dealers,
spoon-benders, ...
So I still think that the people to get onside with are those like the
MIT AI-Labbers, who certainly have their own agenda, but do know the
difference between stuff that works and stuff that doesn't, and whose
brains do not seem to be encased in concrete, from what I've see. Which
means more development of `big demos' like Arm.
So more on Arm: one possible difference between direct control of
velocity and proportional integration of position is that since the
former involves two levels of control systems, you get two distinct
error signals to play around with, positional error and velocity error.
Hence reorganization/self-tuning circuits would have more information
to play with, in particular, they can distinguish between average
positional error (caused perhaps by the target moving around a lot),
and average velocity error (caused by children tugging on your arm,
perhaps).
Avery.Andrews@anu.edu.au