Re: Disguised AI bots in social platforms

Rich <rich@example.invalid> writes:

Anton Shepelev <anton.txt@g{oogle}mail.com> wrote:

Hello, all.
 
No sooner was Usenet purged of the plague of GoogleGropus SPAM, than
another has appeared on the horison.  Since AI in general and LLMs in
particular are developing at break-neck speed, social platforms may
soon be infested by intelligent bots that will be rather hard to
distinguish from humans (e.g.  when the LLM is uncensored).  Will it
be the end of online group-based communication?  Is there any hope of
preventing or at least staving off this new apocalypse?

>
Removing the profit motive from the spammers.  So long as gullible
users buy the wares offered, or hand money over to the scams, the
spammers have a profit motive to continue to work around all attempts
to thwart them.

Easy to say, very hard to do...

I for one have only one idea: a heterarchical redundant mutual cross
verification of users by each other via off-line meetings.

>
I.e., the pgp web-of-trust.  It technically worked well.  In reality
it did not live up to its true value due to the need for those
"off-line" meetings to truly make it workable.
>
So I see no reason to expect a new variant will fare any better.

The PGP implementation is pretty bad. Actually the in-person key
confirmation is one of the few features to have survived (generally in
more user-friendly form) into other designs.

--
https://www.greenend.org.uk/rjk/