David Fokkema wrote:
> On Mon, Jul 12, 2004 at 05:26:42PM -0700, Brian Nelson wrote:
>>It doesn't need to be available. If C-R gains enough popularity, of
> Of course it does. For any statement to be held true, it has to be
> corroborated by some evidence, with the exception of statements which
> can be logically or mathematically proved. The above statement is too
> vague for that.
That is not true. If we look at the pattern of Bayesian classifiers we do
see a good case of being able to say "C-R doesn't work because spammers will
do this..." Many people who were dubious of Bayesian classification and some
C-R zealots pointed out that spammers could defeat it by simply throwing out
random words. Dunno about you but several months ago email started throwing
out random words to bypass the classifiers. Of course unlike the dire
predictions of those people back then, esp. the zealots, the classifiers
didn't just crumble. Many still work admirably and now others are advancing
to new forms which make it harder for the spammer to just toss out random
words. So, yes, we can look at this problem and identify the weaknesses.
Just because those weaknesses haven't been exploited yet doesn't mean they
don't exist.
> Probably spammers will invest a lot of time. BUT, so will C-R system
> designers. Challenges may require more and more human logical thought,
> like little riddles or something, or requiring to actually _read_
> something or whatever. If spammers will invest more and more time to
> mimic human thought and answer those challenges, maybe this will result
> in a great advancement in artificial intelligence.
We're already to that point. Spamarrest requires people to hit their
page, view a graphic which has some random noise applied to it, and enter the
word they see there. Sad part is I've had to do this dozens of times to
people who wrote me first. :(
--
Steve C. Lamb | I'm your priest, I'm your shrink, I'm your
PGP Key: 8B6E99C5 | main connection to the switchboard of souls.
-------------------------------+---------------------------------------------
Attachment:
signature.asc
Description: OpenPGP digital signature