Backward propagation

Kenneth Collins k.p.collins at worldnet.att.net
Mon Nov 4 16:15:07 EST 2002


If anyone 'wonders', I haven't read much in traditional 'artificial
intelligence'. I've tried to save all of what I have read, but
filled-up a dumpster when I had to give-up my apartment back in 1999.

I knew, before responding in this thread what "back-propagation" is
in AI, although I've never attempted programming any 'neural-network'
version of such, and I've not studied the Maths of it. I was
interested, back in the 70s, but, the then-emphasized 'expert
systems' approach turned-me-off, and when it quickly became clear
that 'neural networks' in AI were, at least then,
hopelessly-disconnected from real neural networks, which I had to
work-through, devoid of topological stuff like "decussation", I knew
that I had to 'look-elsewhere'. [The Hopfield ref cited in AoK was
cited in this spirit - as a "yes" to "emergence of collective
properties", but a "no" to the all-to-all approach of then-current
network calculations.] I ended up 'rolling-my-own', which is
described in AoK, and schematically-diagrammed in my 1983 NRL
poster-presentation paper coppies of which were distributed at that
conference and via mailings [my way of 'publishing'].

"Unlearning", "whittling"  [AoK, Ap5], "rendering useless", the
"prefrontal force" [AoK, Ap7, 8], many instances of neural activation
dynamics that are specifically-discussed throughout AoK, and
diagrammed in the NRL paper, and the stuff I've discussed online, but
which is not in AoK [but still Original to my own work - I'd
discussed it online long before I saw it discussed anywhere else - it
should be in my notes from the 70s - I got 'flak' for proposing it,
which is why, during the 8-10 weeks during which I wrote AoK, I
didn't include it], pertaining to the reiterative cross-correlation
that occurs during 'sleep-consciousness', are examples of
'back-bropagation' as it's integrated in NDT and my design for a
machine that can think [the "self-organizing system" presented in the
NRL paper].

But I don't do traditional AI, and I've avoided what I've reason to
expect would be the 'heart'-ache of reading in more-current AI refs.
It's one of those things I just 'shrug-off', because, although it
'hurts', it's not Killing folks - at least not directly - if NDT's
concepts've been 'co-opted' in modern AI, without public discussion
of its ramifications with respect to the understanding of Human
behavior, then that's ab-use that does Kill.

If anyone 'wonders' how what I'm doing in this thread correlates with
the stuff of AoK's Epilogue [which I Cherish], it is that, it's clear
to me that I must take a stand against what continues to be a
virtual-attack upon Humanity that takes the form of 'borrowing' and
'co-opting' stuff, in my work, that, if allowed existence in accord
with the First Ammendment Principles, would, at least to the degree
to which it were, in fact, communicated, be counter-balancing
'blindly'-automated Savagery.

One doesn't forgive a bullet when it's in the air. One tries to
deflect it's destructiveness, while, simultaneously, taking action to
end hostilities, so as to eliminate further bullets' getting into the
air.

Then, one Forgives.

Anyway, please breathe-Life into Freedom by voting, wisely, tomorrow.
Let the stuff of that action course-right-through-you. Feel it, and
all it means.

K. P. Collins





More information about the Neur-sci mailing list