From: Emlyn (emlyn@one.net.au)
Date: Tue Mar 27 2001 - 14:45:15 MST
> Sure, you'd apparently get the right behavior as output.
> But the real Eugene would no longer be with us. Right now,
> ask him a question, and the wheels really turn. A great
> deal of logic and calculation occurs, where, significantly
> a lot of information flows from state to subsequent state.
> But in this lookup table, there is no actual calculation.
> At each point in time, you merely form the enormous 10^30
> bit long address; say 10^10 bits for the input this milli-
> second, and 10^20 bits for his current state, and voila!
> You fetch from the computronium his next state. But no
> information flows. It's not a real calculation, I think.
>
> Lee Corbin
>
Just to bring an implicit assumption to the surface... this is the point of
view that the act of calculation, iterative process, imperative methods, are
the basis of consciousness. That consciousness does not come from the
substrate itself (wetware, hardware, magicware), or from some ineffable
eternal soul, or some other such construct, but as a by-product of the act
of calculating. Personally, I don't think that's as clearly true as is
implied by many.
Something to consider about lookup tables; the giant Eugene Leitl lookup
table cannot learn. If you give him an entirely foreign set of inputs,
you'll go looking for actions in your lookup table, and either come up blank
(Eugene blue-screens), or get some default behaviour(s), which would only
be appropriate initially. Eugene is a smart guy... he'd adapt to and learn
about his new environment, which would ultimately affect his behaviour.
The model of the lookup table needs to be altered to allow it to be altered.
You possibly need a meta-lookup table, with the rules about how Eugene
learns, and all input and output, including from Eugene's own actions, and
modeled thoughts (they must be included as actions in the lookup table) must
be processed through the meta-table, which would produce instructions for
altering the original table on the fly, so that Eugene can change and learn.
Of course, these actions must also be able to alter the meta table.
But perhaps we are only talking about putting the tabulated Eugene into
situations identical to those that the real Eugene has experienced. In this
case, you wouldn't need learning possibly, as long as an input to the lookup
table is time...
At time x, in situation y, Gene does Z
At time x', in situation y, Gene does Z'
That starts to make the whole table construct very brittle indeed...
But, assuming that it can still work... is it conscious? I've seen no
argument to suggest that it is any more or less conscious than the original
Gene.
Emlyn James O'Regan - Managing Director
Wizards of AU
http://www.WizardsOfAU.com
emlyn@WizardsOfAU.com
"Australian IT Wizards - US Technology Leaders
Pure International Teleworking in the Global Economy"
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:06:43 MST