I haven't previously encountered the spam phenomenon we're seeing in several new threads. The users all have realistic, albeit formulaic, names. They post something that seems pretty on-topic and doesn't contain any spam links. Sometimes it even generates conversation! The language is pretty correct; it seems like either it's written by a human or a very good facsimile
Then, a couple of days later, they come back and edit the message to contain some links. And the links aren't to typical spam-type websites (like 'buypiratewindows10.com' or 'freestreamingmovies.net' or something), but I haven't followed them to see where they actually go.
Posted by Mr. Y (Member # 11590) on :
Bots be learning...
Posted by steven (Member # 8099) on :
quote:Originally posted by Mr. Y: Bots be learning...
They’re not bots. They’re my alts
Posted by ricree101 (Member # 7749) on :
Language generation models have gotten pretty good lately (see GPT-2 for instance). So long as you don't actually care what gets outputted, it's possible to produce a large amount of fairly plausible looking text.
I suppose it was only a matter of time before spammers started using them.
Posted by JanitorBlade (Member # 12343) on :
I very nearly pounced on ricree101's post by habit.