Re: [math-fun] quantum theory foundational issues, my theory of how they should be resolved
Dan wrote:
Adam, what is the connection between emulating human behavior (or the vaguely defined Turing test) on the one hand, and Turing completeness on the other?
If any digital computer can emulate human behaviour, then so could any Turing-complete system.
Mike wrote:
Also, computers are provably not conscious, since they can be emulated by Turing machines, which are obviously not conscious.
It's not obvious to me.
Instead of thinking of the Turing machine as an electronic machine (which could indeed be consicous), think of it as a long tape of paper, with erasable symbols, and a reference book (the `rule table'). Every day, a different volunteer walks into the room, reads a single symbol, obeys the instruction given in the reference book (e.g. replace the symbol with another symbol, move left and change state), and leaves the room. An inanimate book cannot be conscious. Neither can the tape. Hence, if there is any consciousness, it must lie in the individual volunteers (assuming they're human, they definitely will be conscious). But they can be replaced with much simpler machines (say, mechanical automata made of LEGO), with no memory, which only exist for the few minutes it takes to enter the room, obey the instruction and leave. Now, these LEGO automata are *certainly* not aware of their own existence. And there is only one alive at any time, so we can't argue that they are forming some `super-brain'. The automata themselves cannot make decisions, nor be conscious, since they're far too simple. But the overall system I've described is a Turing machine, which advances its computation each day. Admittedly it's about 10^17 times slower than the average desktop computer, and will take billions of years to emulate just a single second of a large neutral network, but it's still capable of emulating one, nonetheless. Sincerely, Adam P. Goucher
Adam, it seems that you have reinvented Searle's Chinese Room Argument http://plato.stanford.edu/entries/chinese-room/ http://en.wikipedia.org/wiki/Chinese_Room_Argument and as noted in these link there have been many critical replies. That is, it is not a slam dunk arguments as you seem to think. --Edwin On Sun, Aug 4, 2013 at 11:31 AM, Adam P. Goucher <apgoucher@gmx.com> wrote:
Dan wrote:
Adam, what is the connection between emulating human behavior (or the vaguely defined Turing test) on the one hand, and Turing completeness on the other?
If any digital computer can emulate human behaviour, then so could any Turing-complete system.
Mike wrote:
Also, computers are provably not conscious, since they can be emulated by Turing machines, which are obviously not conscious.
It's not obvious to me.
Instead of thinking of the Turing machine as an electronic machine (which could indeed be consicous), think of it as a long tape of paper, with erasable symbols, and a reference book (the `rule table'). Every day, a different volunteer walks into the room, reads a single symbol, obeys the instruction given in the reference book (e.g. replace the symbol with another symbol, move left and change state), and leaves the room.
An inanimate book cannot be conscious. Neither can the tape. Hence, if there is any consciousness, it must lie in the individual volunteers (assuming they're human, they definitely will be conscious). But they can be replaced with much simpler machines (say, mechanical automata made of LEGO), with no memory, which only exist for the few minutes it takes to enter the room, obey the instruction and leave.
Now, these LEGO automata are *certainly* not aware of their own existence. And there is only one alive at any time, so we can't argue that they are forming some `super-brain'. The automata themselves cannot make decisions, nor be conscious, since they're far too simple.
But the overall system I've described is a Turing machine, which advances its computation each day. Admittedly it's about 10^17 times slower than the average desktop computer, and will take billions of years to emulate just a single second of a large neutral network, but it's still capable of emulating one, nonetheless.
Sincerely,
Adam P. Goucher
_______________________________________________ math-fun mailing list math-fun@mailman.xmission.com http://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
On 8/4/2013 8:31 AM, Adam P. Goucher wrote:
Dan wrote:
Adam, what is the connection between emulating human behavior (or the vaguely defined Turing test) on the one hand, and Turing completeness on the other? If any digital computer can emulate human behaviour, then so could any Turing-complete system.
Mike wrote:
Also, computers are provably not conscious, since they can be emulated by Turing machines, which are obviously not conscious.
It's not obvious to me. Instead of thinking of the Turing machine as an electronic machine (which could indeed be consicous), think of it as a long tape of paper, with erasable symbols, and a reference book (the `rule table'). Every day, a different volunteer walks into the room, reads a single symbol, obeys the instruction given in the reference book (e.g. replace the symbol with another symbol, move left and change state), and leaves the room.
An inanimate book cannot be conscious. Neither can the tape. Hence, if there is any consciousness, it must lie in the individual volunteers (assuming they're human, they definitely will be conscious). But they can be replaced with much simpler machines (say, mechanical automata made of LEGO), with no memory, which only exist for the few minutes it takes to enter the room, obey the instruction and leave.
Now, these LEGO automata are *certainly* not aware of their own existence. And there is only one alive at any time, so we can't argue that they are forming some `super-brain'. The automata themselves cannot make decisions, nor be conscious, since they're far too simple.
But the overall system I've described is a Turing machine, which advances its computation each day. Admittedly it's about 10^17 times slower than the average desktop computer, and will take billions of years to emulate just a single second of a large neutral network, but it's still capable of emulating one, nonetheless.
That's another form of Searle's Chinese room. Consciousness, if there is any, is in the computation being performed. It the Turing machine is simulating at a very low level (e.g. molecular) life on Earth, then it will include the conscious thoughts of the beings alive at the time simulated. Brent
participants (3)
-
Adam P. Goucher -
meekerdb -
W. Edwin Clark