Before I reply to Phil, a reminder: don't quote more than the reader
needs to understand your comment. Phil: You quoted way too much needless
text...
Now on to my own (economical) Q/C's...
On Mon, 15 May 2000, Blakemore, Philip wrote:
> are we any closer to T3 after 50 years for research? ...
> ...for T3, a robot capable of
> interacting with the real world, we are no closer.
Well, a good deal closer that 50 years ago, I think, but still very far.
It is the cinematic robots that are making us most impatient and
unappreciative of the progress we HAVE made, because they are so far
ahead of the game (but alas only fictionally).
Remember that even "toy" models of today do things nothing but people
could do 50 years ago.
Having said that, it's good to remember that these toys are all a far
cry from T3.
> In my opinion,
> consciousness is governed by the parts that we are made of; the exact
> atoms and combinations of them provide a unique way to create life, minds
> and consciousness. We will not be able to create consciousness until we
> understand every part of our brain and how it works.
That is indeed an opinion, and it may or not be right. It amounts to
conjecturing that the only way to pass T3 is by passing T4.
But how do we go about passing T4? The brain does not explain itself.
And, unlike liver function and heart function, which we might call
"vegetative," brain function is behavioural! That all brings as back to
T3. How are we to reverse-engineer the brain other than by trying to
model its performance power? It is not wearing its algorithms on its
sleeves...
And settle for T3 (performance) power, trusting that consciousness will
piggy-back on it, once you get there. (For if not, Turing points out that
you'll never know otherwise anyway...)
Harnad, S. (2000) Minds, Machines, and Turing: The Indistinguishability
of Indistinguishables. Journal of Logic, Language, and Information
(JoLLI) special issue on "Alan Turing and Artificial Intelligence."
http://www.cogsci.soton.ac.uk/~harnad/Papers/Harnad/harnad00.turing.html
> A robot cannot be [organic] by definition. If the
> functionality of a robot appeared to have consciousness, how can we prove
> that? We can't. We can either take the robot's word for it, or see how
> it was built. All technology, at the moment, can only achieve what has
> been programmed into it. A machine, toaster, video playing, stereo or
> robot can only do what a human has told it to do. For consciousness to
> exist in robots, a level of independent, high-level extraction of data is
> necessary.
Not sure what that last sentence means. The rest is certainly true: A
robot, being by definition built by us, is synthetic, hence not organic.
But otherwise, so what?
And you (and Turing) are right that we can't know whether it (or any
other entity but ourselves) is conscious one way or the other, so we
must settle for equivalence and indistinguishability.
And we've done several rounds about whether it makes sense to say that
the fact that a robot may be implementing a program someone wrote is a
problem, somehow: Our DNA is code; our neurons execute code. The code
was shaped by evolution (the "Blind Watchmaker"), but so what? Code's
code. Its history and origin is irrelevant; only its functionality is
relevant. Focus on T3, not on who made the candidate...
> An embryo is not an artifact. Whether artificially created or not, an
> embryo (by definition) has the capability of life itself.
A natural embryo is not an artifact, by definition. But what about a
synthetic embryo? And what is "life itself"? What if the synthetic
embryo can grow and develop into a grown-up (synthetic) adult, and can
reproduce more of the same. Is it not "alive" too (esp. if T3)?
> If we can't make an artificial ant have consciousness, we can't expect to
> build a machine with human consciousness.
There is unfortunately no ant T3 (or if there is one, we can't mind-read
ants well enough to tell if they are really Turing-Indistinguishable
from real ants; and real ants will mate with a caraway seed sprayed with
ant pheromone, so they are not good as ant Turing-Testers either).
So whereas we can try to approximate ant T3 on the way to human T3, only
the latter can be decisive.
Besides, have patience...
> > Edwards:
> > This will speed up the evolution of Cog, by tremendous amounts. A few years
> > in a lab may be akin to a million years of organic evolution.
>
> Blakemore:
> Where are these statistics from? Although machine's can run programs
> faster than humans can, Cog is designed to integrate with the real world;
> at our speed. Any learning or growth has to do achieved on our time line.
That is a correct point. Getting to T3 power can be accelerated, but T3
testing itself must take place in real time (and, in principle, for a
lifetime).
> We do not have the computational power to perform complete natural
> language processing (as admitted earlier). How is this advanced robot
> going to get this necessary power from?
T2 is a subset of T3, and (by my lights) grounded in it.
> For an understanding of Chicago, you need to define what level of
> understanding passes or fails the test. Is it where it is on the globe,
> what restaurants are there, what people live there or the experience of
> being there?
>
> Does it really matter? Yes, you need to know what you want a T3 robot to
> do before building it! A child isn't expected to know a great deal of
> knowledge, but probably would be enough to know the place exists.
According to T3, the robot has to be generically indistinguishable from
me, including in its relation to Chicago (which, by the way, I may or
may not know about, but, like a child, ought to be able to learn
about).
> For a T3 robot to pass, it surely must not lie. If it could Terminator
> could become a reality!!!!
Don't we lie? T3-equivalence means T3-equivalence...
Stevan
This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:36:28 GMT