Spam filters

29 views
Skip to first unread message

Aaron Meurer

unread,
Jul 19, 2021, 5:58:26 PM7/19/21
to sympy
I've noticed that Gmail has recently been incorrectly categorizing
quite a few messages on this list as spam. I don't know if this
problem affects others as well, but if you follow this list and don't
want to miss any messages, you may want to consider creating a filter
that prevents any messages from sy...@googlegroups.com from being
marked as spam.

Aaron Meurer

Aaron Meurer

unread,
Jul 19, 2021, 7:37:14 PM7/19/21
to sympy
It's been pointed out to me that at least one of the recent messages
to this list actually *is* spam. It contains a hidden link at the end
of the message. Whoever let it through moderation didn't notice this,
because the message otherwise looks legitimate. I've already removed
the message from the online view. If anyone notices such a message
getting through to this list, please let me know so I can mark it as
spam in the Google Groups interface.

I guess the spammers are getting smarter, and probably using more
advanced neural network techniques to generate convincing sounding
text. So to people moderating messages to this group, be on the
lookout for this sort of thing. You may need to read the messages a
little closer and if they don't seem very high quality or coherent,
check closer if they contain things like hidden links. Hopefully
Google's own spam filters will get smart enough to prevent this sort
of thing.

Aaron Meurer

Aaron Meurer

unread,
Jul 19, 2021, 7:39:36 PM7/19/21
to sympy
On Mon, Jul 19, 2021 at 5:36 PM Aaron Meurer <asme...@gmail.com> wrote:
>
> It's been pointed out to me that at least one of the recent messages
> to this list actually *is* spam. It contains a hidden link at the end
> of the message. Whoever let it through moderation didn't notice this,
> because the message otherwise looks legitimate.

To be fair, the message only looks legitimate at a glance. If you read
it carefully, it's actually nonsensical. So I think the moderators
just need to take a little closer look at messages before they let
them through.

Aaron Meurer

Oscar Benjamin

unread,
Jul 19, 2021, 7:49:18 PM7/19/21
to sympy
I think I let that one through. Looking at it now it doesn't seem like
a great conversation starter but that's not really a criterion. I
wonder if it was created by an AI somehow or if someone spent a little
time crafting a vaguely believable but not quite sensical message.
This list seems an odd target for marketing (what I presume is)
porn...
> --
> You received this message because you are subscribed to the Google Groups "sympy" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sympy+un...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CAKgW%3D6JaD_RkH1LoPbFBgqo1xwcTw7y2z_G7-rkJd3RxhBT_Hw%40mail.gmail.com.

Aaron Meurer

unread,
Jul 19, 2021, 8:06:53 PM7/19/21
to sympy
On Mon, Jul 19, 2021 at 5:49 PM Oscar Benjamin
<oscar.j....@gmail.com> wrote:
>
> I think I let that one through.

No worries. I'm not trying to assign blame. I think in the past a
cursory glance would be fine, but that's sadly not the case anymore. I
think in general we as humans are increasingly going to have to get
better at administering Turing tests.

> Looking at it now it doesn't seem like
> a great conversation starter but that's not really a criterion. I
> wonder if it was created by an AI somehow or if someone spent a little
> time crafting a vaguely believable but not quite sensical message.

I think it was definitely generated by AI. If you've seen the sorts of
things that modern algorithms like GPT-3 can generate, it's just like
this. It all looks grammatically correct and seems to be coherently
talking about some subject. It's only when you read the actual content
that you realize that it makes no sense, because the algorithm doesn't
actually understand the meaning of the words it is using. The
suggestion that "1/2" gives an "error message like my calculator" is
the giveaway.

> This list seems an odd target for marketing (what I presume is)
> porn...

I think the idea is also to fool things like search engine algorithms
which might assume that links on mailing lists are somehow more
legitimate.

Aaron Meurer

>
> On Tue, 20 Jul 2021 at 00:39, Aaron Meurer <asme...@gmail.com> wrote:
> >
> > On Mon, Jul 19, 2021 at 5:36 PM Aaron Meurer <asme...@gmail.com> wrote:
> > >
> > > It's been pointed out to me that at least one of the recent messages
> > > to this list actually *is* spam. It contains a hidden link at the end
> > > of the message. Whoever let it through moderation didn't notice this,
> > > because the message otherwise looks legitimate.
> >
> > To be fair, the message only looks legitimate at a glance. If you read
> > it carefully, it's actually nonsensical. So I think the moderators
> > just need to take a little closer look at messages before they let
> > them through.
> >
> > Aaron Meurer
> >
> > --
> > You received this message because you are subscribed to the Google Groups "sympy" group.
> > To unsubscribe from this group and stop receiving emails from it, send an email to sympy+un...@googlegroups.com.
> > To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CAKgW%3D6JaD_RkH1LoPbFBgqo1xwcTw7y2z_G7-rkJd3RxhBT_Hw%40mail.gmail.com.
>
> --
> You received this message because you are subscribed to the Google Groups "sympy" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sympy+un...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CAHVvXxQOhL0C0xeRRTfKOZOENe1B04GO7xof_8zxPXyhd27bvw%40mail.gmail.com.

Oscar Benjamin

unread,
Jul 19, 2021, 8:24:09 PM7/19/21
to sympy
On Tue, 20 Jul 2021 at 01:06, Aaron Meurer <asme...@gmail.com> wrote:
>
> On Mon, Jul 19, 2021 at 5:49 PM Oscar Benjamin
> <oscar.j....@gmail.com> wrote:
> >
> > I think I let that one through.
>
> No worries. I'm not trying to assign blame. I think in the past a
> cursory glance would be fine, but that's sadly not the case anymore. I
> think in general we as humans are increasingly going to have to get
> better at administering Turing tests.
>
> > Looking at it now it doesn't seem like
> > a great conversation starter but that's not really a criterion. I
> > wonder if it was created by an AI somehow or if someone spent a little
> > time crafting a vaguely believable but not quite sensical message.
>
> I think it was definitely generated by AI. If you've seen the sorts of
> things that modern algorithms like GPT-3 can generate, it's just like
> this. It all looks grammatically correct and seems to be coherently
> talking about some subject. It's only when you read the actual content
> that you realize that it makes no sense, because the algorithm doesn't
> actually understand the meaning of the words it is using. The
> suggestion that "1/2" gives an "error message like my calculator" is
> the giveaway.

I think I need to raise my expectations of humanity. I've seen enough
low-grade communication from actual humans that I no longer expect a
human on the internet to "understand the meaning of the words [they
are] using"!

--
Oscar
Reply all
Reply to author
Forward
0 new messages