12 Dec
2010
12 Dec
'10
9:14 a.m.
Richard wrote:
As I've already said, you do an experiment by
issuing work items at
various price points to determine how much you need to pay in order to
get reliable quality.
I can get on average significantly less than one character in error per
page, *including* the line number, address, object code, source
statement, and comments. I'd be willing to wager that using "unskilled"
labor (people not familiar with the assembly language in question) will
not be able to do it with fewer than five errors per page no matter how
much you offer, since they won't even be able to tell what constitutes
an error when the source material is a poor-quality reproduction from
microfilm or a bad photocopy. Sometimes it is difficult even with a
good copy!
So far, there isn't any way that people can be
made to do perfect work
that contains no errors.
I never claimed otherwise.
Using mechanical turk doesn't bring about utopia.
It just means that
we pay someone else to do most of the drudgery instead of doing it
ourselves.
It is *more* drudgery to correct someone else's errors doing this than
to type it myself. Speaking from experience, dealing with data entered
*and* proofread by supposedly skilled data entry operators.
I would be absolutely delighted if someone could prove me wrong.
Eric