Online political arguments are highly stereotyped. As an AI problem, it’s completely different from that addressed by @OpenAI.
If you want to generate political spam, there are much better ways of doing it, taking advantage of the restricted problem domain.
-
-
Show this thread
-
As far as general spam goes, similar statistical generation techniques have been in routine use for many years, and have long been 100% effective, in the sense that actually existing spam filters don’t try to guess human/bot based on grammar or content.
Show this thread -
Here’s some typical spam. The bot spliced up bits of text it found on the internet, and injected some drug names, which it hopes you are searching for. It is plenty good enough to defeat human/bot filters. You reject this based on keywords (the drug names) or IP behaviors.pic.twitter.com/C4LzxFJhH2
Show this thread
End of conversation
New conversation -
-
-
Have you seen the Deepak Chopra quote generator? (Just keep clicking on 'Receive more wisdom' link to get another quote). http://wisdomofchopra.com
-
That’s great!
End of conversation
New conversation -
-
-
In your opinion, are any of these threats significant and novel enough to justify withholding the full model?
-
I don’t know any details about what the supposed threat model is, so I can’t have a definite opinion. Offhand, I can’t think of a threat model which the technique, or similar techniques, would render significantly more dangerous. But that doesn’t mean there isn’t one.
End of conversation
New conversation -
-
-
the scientific interest claimed in the paper is its ability to generalize to many other tasks in zero-shot fashion; it even does pretty well at translating to other languages despite the fact that they tried to filter out non-english pages from the dataset
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.