Are you serious?

Lynn H. Maxson
Wed, 12 Jul 2000 00:54:21 -0700 (PDT)

"Are you saying we want to make machines think?  
That's a very attractive proposal, to tell the 
truth, but also one which doesn't seem to me to 
offer much hope of success."

Look, I'm almost sorry I brought it up.  My point 
was that there was reflective programming (as 
defined by Fare for Tunes) and then there was 
reflective programming (as defined within 
cybernetics).  Having become thoroughly enraptured 
by D. Ross Ashby's "Design for a Brain" and 
"Introduction to Cybernetics", I'm prepared to go 
where computer scientists fear to tread.<g>

While I slowly pour through some of Fare's 
writings, the most recent being "Reflection, 
Non-Determinism, and the lambda-Calculus", I am 
coming more and more to the conclusion that he 
should call it something other than "reflection".  
He has yet to introduce an implementation which is 
other than (externally) rule-based.  He is very 
careful to point this out.

For me the reading is like piercing a fog and thus 
I may miss some important connection.  I accept the 
responsibility for any misstep.  I may be all alone 
in this.  What he says and how he says it may be 
absolutely clear to everyone else.  In which case I 
will pursue operational definitions on my own until 
my understanding is somewhere close to that of 

To say that a system reflects when it has no choice 
in the matter seems stretching it to me.  I can 
only reflect when it is my choice to do so.  When 
an external agent demands that I reflect on my 
behavior, thinking, or whatever, I do so within a 
range which they have defined.  Otherwise they will 
continue to make the same demand until I do.<g>

In my view to do reflection to the degree that Fare 
suggests exists beyond our rule-based means.  I 
have no more idea of what this is than I do of what 
thinking is or that anything corresponds to it.  We 
have no means that I am aware of a system 
establishing its own rules of reflection through 
spontaneous generation, that which occurs in 
humans.  Neither computers nor software with their 
100% basis in logic have this capability.

In that respect "reflection" as defined by Fare is 
extremely limited.  Certainly not that envisioned 
by cybernetics nor by that actually demonstrated by 
Ashby.  That's not a crime, but it is confusing 
unless proscribed by an operational definition, a 
metric, which sets not only to what it does pertain 
but also to that which it does not.

I would say the same exists for the other 
requirements.  It's very difficult to evaluate any 
Tunes HLL candidate in any "objective" (and thus 
scientific) manner with qualitative measures 
exclusively.  We should have a quantitative basis 
on which we can all agree on the measures, even if 
we disagree on what we are measuring.<g>