[ExI] Destructive uploading.

Adrian Tymes atymes at gmail.com
Sun Sep 4 18:58:32 UTC 2011


On Sat, Sep 3, 2011 at 10:55 AM, Ben Zaiboc <bbenzai at yahoo.com> wrote:
> I don't see how replacing a system's parts one-by-one with identically-functional replacements is any different in effect to replacing them all at the same time.
>
> In fact, I don't see any practical difference between any of the myriad states between system A (unmodified), system A1 (one part replaced by a functionally-identical component), A2.., An, and B (all parts replaced).  Therefore, by my perception of the logic here, there is no functional difference between A and B, whether B was attained via gradual or simultaneous replacement.
>
> Please explain the (presumed, by your above statement) error in my thinking.

In theory, you are correct.  The problem comes when you apply theory to
practice.

In practice, if you gradually change A through A1, A2, ...An, up to B, you'll
wind up with something (B') different than if you just tried to
replace A with B.

It is theoretically possible to avoid this, and to wind up with exactly B.  It's
just that it happens so rarely, that common speech and perceptions make the
assumption that it never happens - and, again, that's practically always a
correct assumption.  It's so frequent that it infiltrates into
concepts of the "self",
and what it means to be "the same", like the example of "the same" boat I gave.

As far as I know, the reason this happens is because, with gradual replacement,
the goal is not simply to arrive at exactly B.  (If it was, then
replacing A with B
directly would usually be the way to go.)  Rather, the goal is to wind up
somewhere around B, adjusting and tweaking the target during the process -
often, giving time to identify any properties of A that might best be preserved.
In this analogy, artificial intelligence B would be a one-shot emulation of
biological intelligence A without any fine-tuning, while B' would be
arrived at by
tweaking A1, A2, and so on to make sure the desired properties are retained.
Obviously, B would be lower quality than B' (as judged by those doing the
tweaking) - but the point is, B is not B'.




More information about the extropy-chat mailing list