From: Lee Corbin (lcorbin@rawbw.com)
Date: Mon Mar 24 2008 - 07:27:50 MDT
Stathis writes
> Lee wrote:
> > [Stathis wrote]
>> > The physics of the machine is just there to determine that the right
>> > state transitions consistently occur. A model of an AND gate does not
>> > have to involve a simulation of semiconductor physics; the only
>> > purpose of the physics is to ensure that the state transitions in a
>> > simple lookup table are followed.
>>
>> Then we disagree, as follows. Your last paragraph conveys, though
>> naturally too briefly, one standard way of saying what a computation is.
>> That your views lead directly to beliefs I consider faintly absurd does
>> not bother you at all, since you endorse the final results of my
>> reduction anyway.
>
> Do I take that as meaning you don't agree that a person following a
> truth table - such and such an input leads to such and such an output
> - could simulate a logic gate?
No, because sufficiently low-level table lookups are just fine. Not
as any kind of estimate to take to the bank, suppose me to be
claiming that when you start looking up bit patches of 10^6 or so
---or in the inimitable example of a Life Board, a region 1000x1000
---then a very small diminution of consciousness occurs.
(By that last phrase, if anyone else is reading this, it does not mean
any event or condition that the subject is later to be able to report
upon. For example, if you were killed at 5pm today, but then a
duplicate of you were created at 6pm and given all the memories
you would have acquired anyway during that hour, then you too
cannot report on "not having really been there" during that hour.
The diminution of consciousness of which I speak has the same
quality.)
I had a great idea while talking to a friend last night on the phone.
Question: would you consider that a given program had passed
the Turing Test under these conditions?
A. All Earth's six hundred million English speaking people interact
by teletype with it, some specialist computer folks up to several
thousand interactions each.
B. Each interaction consists of a question to the program/human
of less than 10,000 characters. The reply of the program (or
human, the interviewer does not know) is always of about
the same length and complexity of that of the question, i.e.,
around 10^4 characters.
C. Each interviewer is convinced that as of 2008, the entity typing
back responses must be a human being.
My guess is that your answer would be "yes, if it is a program then
it has passed the Turing Test", and that, moreover, the remote
entity (whether human or not) was conscious and had internal
experiences.
Now, what if it is revealed to you than an Alien intelligence had
deposited this program/machine in the possession of an Earth
person collaborator, and that the computronium based machine
was merely doing a table lookup on all the possible 26^10000
possible keyboard inputs? In other words, the Alien had somehow,
perhaps by making untold trillions of copies of himself, had personally
answered all possible 10,000 character questions, and then had
just left the device to parrot his responses? You think it's still
conscious?
Lee
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:02 MDT