On 05 Feb 2015, at 19:25, meekerdb wrote:
On 2/5/2015 6:07 AM, Bruno Marchal wrote:
...
To simplify, you can consider a competence as an ability to follow
a program P_i or to compute the corresponding partial or total
function phi_i.
Learning can then be described as the inverse: finding i (or j ...)
when you are presented with a sample (perhaps infinitely growing)
of input and output of phi_i (= phi_j, ...).
There is a general theory of such learning (for both machine and
non-machine). There is no universal learner (Putnam), unless you
weaken very strongly the criteria of identification of phi_i. If
you allow unbounded finite errors, infinite changes of mind,
including changes of mind when the learning machine has already
find a correct program (modulo the finite number of errors), then
you have a (very) theoretical notion of universal learner (Case &
Smith), but it can be shown necessarily impractical (Harrington).
So learning is usually considered as a different kind of
competence, even if it will be itself related to the code of an
inference inductive machine, and so is a form of competence too.
Exactly my point.
But intelligence is not a competence per se: it is the natural
state of all universal machine, before someone installs Windows or
Christian Dogma, or ...
That entails that intelligence is not only different from
competence, it is complimentary to competence.
OK.
The universal machine is maximally intelligent (or the Löbian one?)
The universal machine is maximally incompetent.
Intelligence=incompetence.
Locally, but without intelligence, you can't acquire and develop
competences, which have in general incomparable degrees.
I think that is a reductio. As you agreed above, the ability to
learn is a kind of competence
Assuming computationalism. This means some k compute function phi_k
capable of identifying some class of phi_i from their samples of input-
output (x, phi_i(x)). Of course such k exists in arithmetic, and in
our reality if we are machine.
and intelligence requires that much competence, even if it doesn't
require any other specific competences.
Competence degrees theory is necessarly non-constructive. We can prove
the existence of the right numbers, but we can prove also there is no
tools to recognize or construct those number.
...
It will be conscious at the place where it confuses itself with
the (relatively real) environment. OK. It depends also on its
abilities, and you can make it self-conscious by adding enough
induction axioms. Don't put to much induction axioms, as Mars
Rover will get stuck in self dialog about its consciousness and
how to convince those self-called [censored] humans!
So without the to-many induction axioms it will be conscious, but
not self-conscious.
Yes, but I have still difficulties on this (not to mention my
dialog with salvia which makes things more weird and counter-
intuitive. I feel I still miss something. I might be unconsciously
restrained by some prejudices and probably don't yet push the logic
of comp far enough).
Thus you agree that consciousness is not all-or-nothing.
"Being conscious" is all-or-nothing, but the content, intensity and
atmosphere can vary greatly.
I identify "being conscious" with "there is consciousness", and
this is binary, as there is no degree of unconsciousness (Quentin).
But that's just like saying "being a Belgian" is binary. You can
pick any predicate and say it is "binary" because it divides the
world into x such that P(x) or ~P(x).
In classical logic. Not in constructive or intuitionist logic.
It's a meaningless tautology and it obfuscates the fact that there
are different kinds and degrees of consciousness.
May be you made an argument which was contradicting that tautology. I
don't remember right now. Obviously, a theory of consciousness has to
explain the infinitely many kinds and sorts of conscious states, and
their relations with truth/reality.
But some state of consciousness, (like in sleep), remembered in
other state of consciousness (like after awaken) can seem (and
perhaps be) less intense, blurred, fuzzy, etc.
It is generally very difficult to compare a first person experience
with another one, including our own. The intimacy of that state
precludes any direct comparison. Memories are eliminated, colors
can be added, ...
But can they be added by any consciousness? I don't think my
thermostat can add colors to the temperature. I doubt that color
blind persons can add color.
I remind you that consciousness starts with Turing universal machine
(technically, it starts from less, but so close that I don't want
introduce technical nuances: Turing universality is conceptually very
cheap.
This excludes thermostats. It has obviously not enough senses and
neurons (cabled amoebas) to treat the information needed to get the
color experience.
And I was actually talking about humans or Löbian machines.
But you can, by gentleness confer the consciousness of the universal
machine to all non universal machine, it will not change many things,
because the consciousness of the universal machine is typically
disconnected from the contingent realities, it "lives" at different
levels (given by the arithmetical interpretations of the hypostases).
I give the math to proceed.
We live a constant tyranny of the here-and-now present state, which
can embellish or impoverish the re-enacting of a past experience.
Yes and isn't that a defining characteristic of consciousness: it's
here-and-now. That's one of the difficulties of models like UD
computation; how do you recover here-and-nowness from a timeless
model.
By the indexical, which for machines can be produced by the second
recursion theorem of Kleene. The hypostases define somehow a universal
"here and now", conscious, before the alarming senses tells it that
something has to be done.
Of course the same problem occurs in physics because we want our
theories to be timeless and even time-symmetric.
Not sure we want. We infer this from observation, but the physicist in
me bet this is right, and classical computationalism also indicates
that the physical bottom is linear, symmetrical, reversible. (It is
given by the variants []p & <>t (& p) restricted on p sigma_1 (= the
arithmetical UD).
A priorii this is still quite mysterious, like the importance of low
dimension (around 24?), the role of simple groups, and rings.
The physicists works bottom up. The theologian works top down.
They have to meet in between. If serious enough.
Bruno
Brent
--
You received this message because you are subscribed to the Google
Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it,
send an email to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
http://iridia.ulb.ac.be/~marchal/
--
You received this message because you are subscribed to the Google Groups
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
