phil osborn wrote:
>
> Just a few thoughts: The other night I accidentally tuned in to "Red
> Dwarf." The computer convinced one of the characters to erase part of its
> "memory." Problem was that it had read every single book ever written -
> supposedly stored in its "library" - and was bored. So it wanted all the
> Agathe Christy novels erased so that it could have the pleasure of reading
> them again.
>
> Question: What is the precise difference between having the novels in
> "memory" and in "storage?" Presumably the computer could access the
> information in the library via any number of algorythmic searches or even
> spawn intelligent agents to do more sophisticated access to the actual
> informational content and meaning. What additional factor is involved in
> actually reading it that isn't present in intelligent transclusion?
Perhaps, for the computer, intelligent transclusion *is* "reading" the
book. Given the attitude of the builders of the ship Red Dwarf (for
humans, by humans - perfectly understandable, given that at the start of
RD, the Singularity is still a ways away), it makes sense that the
library would be stored as, say, ASCII text English. The AI's memory
must be in some other format, if only because ASCII text English is
unsuitable for perfectly representing the memory of a sentient being.
The computer was apparently made to think of itself as similar to a
human, so it could well use the metaphor of "reading" something by
translating text into whatever format its memory uses.
> In terms related to uploading, what would be the difference between my
> moving the library of congress - or just a single book - into my RAM or the
> equivalent and my actually "reading" the material in question?
The side effects. Wear and tear on the book, transportation to/from the
LOC, temporary inavailability of the material to other patrons of the
LOC, memory of having done one process or the other, and so forth
(probably including a siimilar litany for side effects of translating it
into your RAM, depending on how that is implemented). And this is
assuming your auto-reading software (which translates from text to your
memory format) perfectly duplicates whatever nuances you'd interpret the
book with when "actually" reading it. (Which is unlikely since the
book's contents would presumably go through a faster - and thus,
presumably different - path to get to your memory, unless the paths are
specifically engineerd to produce identical results. Unless the
differences are detrimental to your use of the contents - which is
probably true of imperfect "learned" skills like reading - then why
would you want to duplicate the nuances - and bugs - of the "natural"
approach exactly?)
This archive was generated by hypermail 2b29 : Mon Oct 02 2000 - 17:36:47 MDT