Re: Grandmother objections

From: Harnad, Stevan (harnad@cogsci.soton.ac.uk)
Date: Wed Dec 13 1995 - 19:35:19 GMT


> From: "Lucas, Melody" <MFL93PY@psy.soton.ac.uk>
> Date: Wed, 13 Dec 1995 18:00:37 GMT
>
> > Hi Mel, there's hope because there are only about 6-8 grandmother
> > objections to computers/machines in all; once you've heard why
> > each of them is invalid, it's over!
>
> I feel as if I must have nearly exhausted them. What about a
> straight-forward kid brother list of the 8, or is that cheating?

Ok, but I'm not sure I can remember them all:

(1) Computer only does what it's programmed to do.

[Answers: (a) Program is just a set of formal rules; many of our traits
are programmed in that sense, e.g., the DNA code in our genes. (b)
If a programmer writes a programme it doesn't mean he knew in advance
everything it would/could do: A programmer could invent the DNA code.
(c) Programmes can be self-modifying, just as we are. (d) Codes can be
generated without a programmer (e.g., DNA).]

(2) Computers can't do anything new.

[Yes they can, for example, because chance effects from outside the
computer (inputs), or from inside the computer, or pseudo-random
effects produced by the code itself, or through effects of the code
that the programmer did not expect, or through self-modification.]

(3) Computers can't be creative.

[Yes they can, see above; and if you mean REALLY creative, like
Einstein, most of us can't either.]

(4) Computers can't make mistakes.

[Yes they can; reply similar to reply about doing something new.]

(5) Computers are mechanical, we are flexible.

[Programmes can be extremely flexible, adapting to inputs, or to changes
in their own code; and if we adopt another scale, we're pretty
mechanical -- predictable, repetitive, limited -- too.]

(6) People have real-time histories; computers only have a pseudo-past.

[If you were duplicated, molecule-for-molecule, at this moment, your
double would not have a real history either: so what? It doesn't matter
whether the current state was reached through real time or otherwise:
If it's the right state, it's the right state.]

(7) Computers can't choose; they can only do what they are programmed to
do.

[See above, for freedom, flexibility, and error; besides, it's not clear
whether we can really choose either.]

(8) Computers don't/can't have feelings.

[Whether or not that's true is what this is all about; it cannot simply be
assumed to be true.]

(9) We're not mere machines.

[What's a machine? -- Till further notice, it is any system that operates
according to the causal laws of physics. And what are we?]

(10) I don't want to know how a computer does it, I want to know how *I*
do it.

[Till further notice, the clearest theory of how anyone or anything does
certain kinds of intelligent things is: through computation. So until a
better theory comes along, we have no basis for rejecting computation.]

There IS a version of (10) that is not a grandmother-objection, and that
is that computers can only do little BITS of what we can do. There are
many ways to do the little bits, so there's not point taking them
seriously. The answer to this is that it's correct, but if/when the
models start scaling up toward human capacity (as the Turing Test
dictates), this objection loses its force.

(11) Computers are isolated from the world; we are not.

[Computers can be as interactive with the world as their input/output
devices make them.]

A nongrandmotherly version of this objection, however, points to the
symbol grounding problem: The symbols in a computer are ungrounded; our
brains are not.

> > Grandmother-objections are the ones that everyone always
> > raises at first; rare objections whatever their direction.
>
> Does that mean I'm EXCEPTIONALLY stupid then?
> I'm getting a tad worried. Hope you can help, Mel.

No grounds for worry; everyone means everyone: smart and stupid.
(You're smart, but you are one of us all...)

Chrs, S



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:56 GMT