> The "interesting exercise" would be the exercise of trying to write in
such
> a way as to get the statistical measuring instrument (usually a computer
> nowadays) to guess wrong - in other words, to bring the gender of the text
> back under conscious authorial control (and from thereon to manipulate
it).
This was {partly} what I was thinking of when I flagged the problem of the
reverse-Turing-box.
Actually, it wouldn't be (I assume) all that difficult to fool a statistical
program -- all you need to know are the criteria the program uses, and then
(over)load the program with the appropriate signals.
The old race of cannon against armour ...
There's a piece by Hoffstadder somewhere on how even he almost got caught on
the wrong side of a Turing box.
But I've yet to meet an AI I could truly love.
As far as I know, that is ...
Robin
|