On 30 Sep 2013, at 03:36, Pierz wrote:
If I might just butt in (said the barman)...
It seems to me that Craig's insistence that "nothing is Turing
emulable, only the measurements are" expresses a different
ontological assumption from the one that computationalists take for
granted. It's evident that if we make a flight simulator, we will
never leave the ground, regardless of the verisimilitude of the
simulation. So why would a simulated consciousness be expected to
actually be conscious? Because of different ontological assumptions
about matter and consciousness. Science has given up on the notion
of consciousness as having "being" the same way that matter is
assumed to. Because consciousness has no place in an objective
description of the world (i.e., one which is defined purely in terms
of the measurable), contemporary scientific thinking reduces
consciousness to those apparent behavioural outputs of consciousness
which *can* be measured. This is functionalism. Because we can't
measure the presence or absence of awareness, functionalism gives up
on the attempt and presents the functional outputs as the only
things that are "really real". Hence we get the Turing test. If we
can't tell the difference, the simulator is no longer a simulator:
it *is* the thing simulated. This conclusion is shored up by the
apparently water-tight argument that the brain is made of atoms and
molecules which are Turing emulable (even if it would take the
lifetime of the universe to simulate the behaviour of a protein in a
complex cellular environment, but oh well, we can ignore quantum
effects because it's too hot in there anyway and just fast forward
to the neuronal level, right?). It's also supported by the
objectifying mental habit of people conditioned through years of
scientific training. It becomes so natural to step into the god-
level third person perspective that the elision of private
experience starts seems like a small matter, and a step that one has
no choice but to make.
Of course, the alternative does present problems of its own! Craig
frequently seems to slip into a kind of naturalism that would have
it that brains possess soft, non-mechanical sense because they are
soft and non-mechanical seeming. They can't be machines because they
don't have cables and transistors. "Wetware" can't possibly be
hardware. A lot of his arguments seem to be along those lines — the
refusal to accept abstractions which others accept, as telmo aptly
puts it. He claims to "solve the hard problem of consciousness" but
the solution involves manoeuvres like "putting the whole universe
into the explanatory gap" between objective and subjective: hardly
illuminating! I get irritated by neologisms like PIP (whatever that
stands for now - was "multi-sense realism' not obscure enough?),
which to me seem to be about trying to add substance to vague and
poetic intuitions about reality by attaching big, intellectual-
sounding labels to them.
However the same grain of sand that seems to get in Craig's eye does
get in mine too. It's conceivable that some future incarnation of
"cleverbot" (cleverbot.com, in case you don't know it) could reach a
point of passing a Turing test through a combination of a vast
repertoire of recorded conversation and some clever linguistic
parsing to do a better job of keeping track of a semantic thread to
the conversation (where the program currently falls down). But in
this case, what goes in inside the machine seems to make all the
difference, though the functionalists are committed to rejecting
that position. Cleverly simulated conversation just doesn't seem to
be real conversation if what is going on behind the scenes is just a
bunch of rules for pulling lines out of a database. It's Craig's
clever garbage lids. We can make a doll that screams and recoils
from damaging inputs and learns to avoid them, but the functional
outputs of pain are not the experience of pain. Imagine a being
neurologically incapable of pain. Like "Mary", the hypothetical
woman who lives her life seeing the world through a black and white
monitor and cannot imagine colour qualia until she is released, such
an entity could not begin to comprehend the meaning of screams of
pain - beyond possibly recognising a self-protective function. The
elision of qualia from functional theories of mind has potentially
very serious ethical consequences - for only a subject with access
to those qualia truly understand them. Understanding the human
condition as it really is involves inhabiting human qualia.
Otherwise you end up with Dr Mengele — humans as objects.
I've read Dennett's arguments against the "qualophiles" and I find
them singularly unconvincing - though to say why is another long
post. Dennett says we only "seem" to have qualia, but what can
"seem" possibly mean in the absence of qualia? An illusion of a
quality is an oxymoron, for the quality *is* only the way it seems.
The comp assumption that computations have qualia hidden inside them
is not much of an answer either in my view. Why not grant the qualia
equal ontological status to the computations themselves, if they are
part and parcel? And if they cannot be known except from the inside,
and if the computation's result can't be known in advance, why not
say that the "logic" of the qualitiative experience is reflected in
the mathematics as much as the other way round?
Well enough. I don't have the answer. All I'm prepared to say is we
are still confronted by mystery. "PIP" seems to me to be more
impressionistic than theoretical. Comp still seems to struggle with
qualia and zombies. I suspect we still await the unifying perspective.
Dennett just put the mind under the rug, like he has, as he believe in
comp and in materialism.
Now, I do think that the intensional variant of the self-reference
logics does provide a genuine theory of qualia, which explain them as
knowable truth, having some special semantics (related to perception
field and imaging) and we do have the explanation why they seem
irreductibly not explainable in third person term.
For me, this is enough to believe that RA and PA is conscious, and has
already some qualia.
Computer science is sophisticated enough to explain why a self-
introspecting machine can understand why there are things about
herself which she cannot understand, yet can memorize, and even
described relatively to other entities having similar experience.
Consciousness is NOT Turing emulable, as it is a (self) selection on a
non computable domain, with the unification price: we have to recover
physics from number theology.
Bruno
On Thursday, September 26, 2013 8:17:04 PM UTC+10, telmo_menezes
wrote:
Hi Craig (and all),
Now that I have a better understanding of your ideas, I would like to
confront you with a thought experiment. Some of the stuff you say
looks completely esoteric to me, so I imagine there are three
possibilities: either you are significantly more intelligent than me
or you're a bit crazy, or both. I'm not joking, I don't know.
But I would like to focus on sensory participation as the fundamental
stuff of reality and your claim that strong AI is impossible because
the machines we build are just Frankensteins, in a sense. If I
understand correctly, you still believe these machines have sensory
participation just because they exist, but not in the sense that they
could emulate our human experiences. They have the sensory
participation level of the stuff they're made of and nothing else.
Right?
So let's talk about seeds.
We now know how a human being grows from a seed that we pretty much
understand. We might not be able to model all the complexity involved
in networks of gene expression, protein folding and so on, but we
understand the building blocks. We understand them to a point where we
can actually engineer the outcome to a degree. It is now 2013 and we
are, in a sense, living in the future.
So we can now take a fertilised egg and tweak it somehow. When done
successfully, a human being will grow out of it. Doing this with human
eggs is considered unethical, but I believe it is technically
possible. So a human being grows out of this egg. Is he/she normal?
What if someone actually designs the entire DNA string and grows a
human being out of it? Still normal?
What if we simulate the growth of the organism from a string of
virtual DNA and then just assemble the outcome at some stage? Still
normal?
What if now we do away with DNA altogether and use some other Turing
complete self-modifying system?
What if we never build the outcome but just let it live inside a
simulation? We can even visit this simulation with appropriate
hardware: http://www.oculusvr.com/. What now?
In your view, at what point does this break? And why?
Best,
Telmo.
--
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/groups/opt_out.
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/groups/opt_out.
