Re: When will GOFAI die, for good?
From: Brian Hulley (brianh_at_metamilk.com)
Date: 03/12/05
- Next message: alex goldman: "Re: Minsky still posting"
- Previous message: Traveler: "Re: When will GOFAI die, for good?"
- In reply to: Traveler: "When will GOFAI die, for good?"
- Next in thread: Traveler: "Re: When will GOFAI die, for good?"
- Reply: Traveler: "Re: When will GOFAI die, for good?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: 11 Mar 2005 18:59:13 -0800
Traveler wrote:
> Reification is just another word for symbol processing. The very
> notion that one needs a special programming language to do AI is a
> sign of cluelessness. It's a cluelessness that got its start with the
> lame ideas of Alan Turing and the subsequent rise of symbolic AI in
> the fifties. When will you people accept that GOFAI is and has been
an
> abject failure? When will you people stop worshipping Turing? It's
> pathetic and it's getting tiresome.
>
[cut]
> The Silver Bullet: Why Software Is Bad and What We Can Do to Fix it
> http://users.adelphia.net/~lilavois/Cosas/Reliability.htm
I've looked at your COSA pages.
Expressing a computation as a dataflow diagram with concurrent
synchronous execution may indeed make certain kinds of problems easier
to solve. However this does not change the fact that you are still
constructing algorithms and expressing them, since the textual
representation of the graph + the emulator is just a program like any
other.
If your COSA graphical language has a power much less than a UTM, it
may well be possible to automatically verify the programs thus
constructed.
However if you are claiming that COSA has more power, or at least as
much power, as a UTM, then *obviously* you can't verify them
automatically - otherwise we could just convert all existing software
into COSA and verify it (and burn a few centuries of mathematics :-)).
Did you really believe that anyone reading this newsgroup would fall
for that???
Hardware can be verified because you can verify all machine
instructions for a CPU by exhaustively enumerating propositional states
over a finite number of clock cycles. This is feasible in practice (not
just theory) because the number is also very small. However, your
example of a while loop already requires 100 cycles to verify
completely. What if you'd used a variable instead of a constant? How
would you verify it then? And what happens when you have a larger
system with multiple loops interacting with each other?
If you stay within Newtonian physics (as opposed to quantum physics
etc) I think you will find that it is not possible to construct any
machine more powerful than a TM. In particular, concurrency is
completely irrelevant: any parallel computation can be expressed as a
serial computation.
Your distinction between emulation and simulation is void.
No doubt I've sounded a bit negative, but my main question to you is:
What do you want from us???
Regards - Brian.
"The fox provides for himself; God provides for the lion" - William
Blake
- Next message: alex goldman: "Re: Minsky still posting"
- Previous message: Traveler: "Re: When will GOFAI die, for good?"
- In reply to: Traveler: "When will GOFAI die, for good?"
- Next in thread: Traveler: "Re: When will GOFAI die, for good?"
- Reply: Traveler: "Re: When will GOFAI die, for good?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
