Sea Wasp wrote:
> Graham Woodland wrote:
>
>> You can't really simulate *that* one. Kim is defined by his
>> extraordinary mind - and Clarissa's is, we know, just as good as his,
>> along complementary lines. The first time they go into a wide-open
>> two-way, if not sooner, the jig is going to be up, unless the
>> 'simulated' mind has all the qualities of the real one. In which case
>> it's still a real mind, which happens to be instantiated in a virtual
>> environment. But it's precisely their mental capacities that make
>> them 'superhuman' - you have, to all intents and purposes, still got
>> two superfolks there.
>>
>> Conversely, if you *can* fool 'Kimball Kinnison' into believing that
>> some lesser entity is his Chris for more than about a skillionth of a
>> second, I'm pretty sure you've failed to simulate him.
>>
>>
>
> Part of the goal is to get as close as possible -- but possibility
> will limit you. In this case you could make a super-competent AI that
> THINKS it's Clarissa, and which is at least as capable as the Kinnison
> design.
Sure, that disposes of that problem - they really are peers. And
I'll grant you that pure perspective effects will work in your
favour wrt the rest of us: the difference between
'super-competent' and 'super-uber-competent' will look much
smaller to us than it really is, since we're not equipped to
appreciate a lot of the higher-level stuff anyway.
However...
>
> An example of one of these adaptations would be, oh, Wolverine from
> the X-Men. I can use the Near-Unobtainium (which is carbon, so easily
> integrated) for the "adamantium" skeleton and as the main component of
> his claws, I can enhance his toughness and speed and strength with the
> right modifications and give him regeneration to some extent with the
> nanotech -- but I'm not going to actually reach the "Healing Factor"
> level of the comics, or the true indestructibility of the adamantium,
etc.
>
Approximating Wolverine is fine - he's not a particularly
analytical guy. I still think you're going to have trouble with
Big Brains, and triply so with telepaths of any kind. Consider:
a big part of the essence of the Kinnison simulation is that
they're both much smarter than you are. Worse - Kim is
inquisitive, inventive, and gets around. Granted he isn't Mentor
by a long chalk, I bet he's going to notice sooner or later if
his apparent environment is really a simulation at a much coarser
grain than it pur****ts to be. Probably sooner, since I suspect
that merely human-level statistical analysis would start ringing
alarm bells sooner or later. The alternative seems to be that
your Nearly Sufficient AI can simulate a substantial chunk of a
real-type universe, along with its horde of native humanish-level
intelligences, so well that the inhabitants can't detect it.
You could still get away with lower-tech or pre-scientific mental
supermen, I guess - or with narrower-chanelled geniuses who take
a more purely engineering approach to things.
The only way I can think of right now to avoid the problem for
even a cut-price Kinnison is to have an even more cut-rate
Mentor-process hacking into his brain whenever his train of
thought threatens to become embarrassing. It's not like that's
completely alien to the setting, what?
--
Cheers,
Gray
---
To unmung address, lop off the 'be invalid' command.
|
